I have not been writing as much lately because I am teaching AP US History for a few months. Tomorrow, Festivus, I have all three of my classes for 42 minutes. What better time for an airing of grievances!

I have created a Festivus-themed Quizzo-like review activity. There are historical figure grievances, an H&H Bagels round (New York City-themed questions), a Human Fund Round (questions about historical figures doing good things) and feats of strength (difficult tasks a team tries with the points available for other teams to steal).

Here are the slides and score sheet:

• Festivus AP US Review to Gilded Age

• Festivus Review Score Sheet

Festivus comes from Seinfeld. As much as I love the show, there are some jokes that are not funny today and should not have been considered funny at the time. We have heard unflattering news reports about some cast members. I will address the fact that I am separating art from artist by choosing a Seinfeld-themed review activity.

I am also aware that my 10th-graders may not know or care about Seinfeld. It’s December 23rd, if what I’m doing amuses me more than my students, so be it if we’re taking an opportunity to review content.

How are you celebrating Festivus with your students?

The blog post image is Matthew Keefe - Flickr: Celebrating Festivus 2.

AI Disclosure:

I wrote this post without using generative AI. That means:

• I developed the idea for the post without using generative AI.

• I wrote an outline for this post without the assistance of generative AI.

• I wrote the post using the outline without using generative AI.

• I edited this post without the assistance of any generative AI. I used Grammarly to help edit the post. I have Grammarly GO turned off.

Read’s second trial this spring was so interesting that it still occupies space in my brain. As readers of my Substack know, so does generative AI in education. I have noticed many connections between the two seemingly unrelated topics.

Trigger Warnings: This post contains language about murder, violence, suicide, and, unbelievably, obscene words that start with the letters C and R. Also, there are references to body anatomy comments about Read’s Crohn’s Disease. Additionally, this post includes generative AI-generated images of historical figures from marginalized and oppressed communities.

While I won’t give a complete rundown of the case, here’s a basic overview for the uninitiated: On January 29, 2022, Boston police officer John O’Keefe’s body was found on the front lawn of another Boston police officer’s home in Canton, MA. O’Keefe’s girlfriend, Karen Read, was arrested and charged with murder for allegedly hitting him with her car. A 2024 trial ended in a hung jury. The second trial in June ended with her acquittal on all charges except operating a vehicle under the influence.

In my humble opinion, Read is innocent because for her to be guilty, all of these implausible things would have to be true:

• Almost ten people leaving a house after O’Keefe was allegedly struck missed a dying 6’2 man on the ground in a small front yard.

• A dog wasn’t the reason for the bite and scratch marks found on O’Keefe’s arm.

• Getting hit by a car neither broke nor bruised O’Keefe’s arm nor injured his lower body, nor left any of his DNA on the taillight pieces found at the scene.2

• Dighton, MA police officer Nicholas Barros (who was neither Canton PD nor Boston PD) was wrong about the state of the taillight on Read’s car.

• Snowplow driver Brian “Lucky” Loughran was wrong about not seeing O’Keefe’s body on the front lawn of a house he had been to many times.

• The crash reconstructionists the FBI hired, who found O’Keefe was not hit by a car, were wrong.

With that established, let’s examine all the ways the Karen Read case is eerily similar to generative AI in education.

Improper Connections Between the Lead Investigator and Witnesses = Improper Connections Promoting Generative AI in Education

Read’s defense team discovered improper connections between the officers investigating the crime and both witnesses and potential suspects. For example, the lead investigator, Former3 State Trooper Michael Proctor’s sister, Courtney, texted him that she heard from her friend Julie Albert, "When this is all over, she wants to get you a thank you gift." Proctor’s response was not, “That would be inappropriate.” It was “Get Elizabeth one,” referring to his wife.

Why would the gift be inappropriate? Julie Albert is married to the brother of Brian Albert, the police officer on whose front lawn O’Keefe was found. Additionally, an investigation found that Proctor drank on the job with Albert’s brother. Proctor clearly had connections to the family and should have recused himself from the case immediately.

These inappropriate connections are similar to entities that should be critiquing big tech companies in light of their claims about generative AI, but are instead partnering with them. For example:

• “Common Sense Media… has received funding from OpenAI—the organization wouldn’t say how much—to produce educational materials about AI for teachers and families,” as documented by Vauhini Vara in Bloomberg.

• Vara also shared that, “Microsoft announced partnerships for AI training with the biggest teachers’ unions in the US; Google and OpenAI are working with the unions too.”

• The American Federation of Teachers is partnering with Microsoft, OpenAI, and Anthropic to promote the use of generative AI among teachers. The AFT should be zealously defending children and teachers, not partnering with tech giants.

• The state of Oregon is paying NVIDIA $10 million to promote generative AI to students and teachers. You read that right. Taxpayer money is going to a tech giant to promote its product.

• Miami-Dade County Public Schools and Albuquerque Public Schools are promoting Gemini for Google for Education. Teachers in these districts spent their time shooting commercials for Google for Education.

Examine the expressions on the students’ faces in the YouTube thumbnails. What do you notice? What do you wonder? Do the students look interested or disinterested?

Former Trooper Proctor’s Offensive Text Messages = Offensive Generative AI Outputs and Practices

Former Trooper Proctor sent friends, colleagues, and supervisors obscene text messages regarding this case. Before proceeding, please note the trigger warnings provided above. You’ve been warned.

The text messages included:

• He referred to Read as the R word.

• When he was going through Read’s phone, he texted a group chat, “No nudes so far.”

• “She’s [Karen Read’s] a whack-job (C word).”

• Proctor texted offensive things about Read’s anatomy related to her suffering from Crohn’s Disease.

• In a text to his sister, Proctor wrote, “Hopefully she kills herself.”

While it would be challenging to match Proctor in terms of offensiveness and cruelty, we have seen some examples from the generative AI industry that are at least comparable.

• LLMs have generated text suggesting self-harm. The New York Times reported that Adam Raine killed himself after a lot of engagement with ChatGPT, which generated text about suicide methods.

• Generative AI is known for generating text and images that amplify racism and misogyny.

• Generative AI companies are mimicking historical figures, and the results are offensive. I documented this when I wrote about SchoolAI’s Anne Frank chatbot.

The previously mentioned Vauhini Vara Bloomberg article shared an example where a John Brown chatbot, “insisted, as Brown, that violence is never the answer.”

This is SchoolAI’s History Comes to Life Collection.

What do you notice? What do you wonder?

I notice an enslaver. I notice BIPOC historical figures, one of whom is Harriet Tubman. Does a “Harriet Tubman” chatbot do her life justice?

The Whistleblowers: Officer Barros and Lucky Loughran = Gebru, Mitchell, and Others

Two defense witnesses served as whistleblowers. They testified, even though what they saw did not fit the prosecution's narrative that Read murdered O’Keefe. Dighton, MA police officer Nicholas Barros saw the taillight on Read’s Lexus before it was entered into evidence. The damage he saw was much less than the damage to the taillight when it was entered into evidence. The damage Barros attested to is attributable to Read backing her car into O’Keefe’s car when she left his house to look for him.

Brian “Lucky” Loughran testified that when he plowed the street of the house where O’Keefe’s body was found, he did not see him on the ground, even though he would have to be there if Read had hit him with her car.

Barros and Loughran are similar to Timnit Gebru and Margaret Mitchell, who testified to the harms and limits of generative AI when they co-authored the Stochastic Parrots paper while employed by Google. Calling Gebru and Mitchell “witnesses” does not do justice to their contributions to our understanding of generative AI. Don’t worry, their names will appear again in the section of this post dedicated to the defense’s expert witnesses.

Other generative AI whistleblowers include Microsoft engineer Shane Jones, who wrote to then Federal Trade Commission Chair Lina Khan and Microsoft’s board of directors to warn that Copilot was not safe for public use. Additionally, former OpenAI researcher Suchir Balaji warned that the company violates copyright law before his death.

The Prosecution’s Expert Witnesses = Generative AI Promoters Who Should Know Better

The prosecution used expert witnesses who made claims about the alleged collision that were backed by little more than a video of a Lexus with blue paint on it reversing at two miles per hour, while a man with a doctorate, dressed up as the victim, subtly moved his arm to collide with it.

That expert was Dr. Judson Welcher, who, like those promoting generative AI in education, argued a case with flimsy evidence. Defense Attorney Alan Jackson called the blue paint test “some ridiculous blue paint kindergarten project” during his closing argument. Welcher was unique in his pomposity. Two other Welcher trial highlights were:

• Wecher cited an article that said 61% of pedestrian motor vehicle collision injuries are head injuries. The study is almost fifty years old. It’s from 1979, before SUVs, which tend to cause lower-body injuries, had a considerable share of the domestic auto market.

• Welcher proclaimed, “They didn’t actually X-ray, from the records I’ve seen, either his [O’Keefe’s] hand or his arm.” Oh, but they did. Oops.

Welcher reminds me of Professor Ethan Mollick, who promotes the use of generative AI in education.

• Brian Merchant wrote about Mollick’s excitement about GPT-5: “Ethan Mollick’s breathless endorsement of GPT-5 (“GPT-5: It just does stuff”) already looks like an ad for New Coke.”

• Mollick said that instructors [teachers] are “becoming coders now” because of generative AI. That is similar to “They didn’t actually X-ray.” Teachers are teachers. Not coders.

• Mollick wrote in his book that, “We have invented technologies, from axes to helicopters, that boost our physical capabilities; and others, like spreadsheets, that automate complex tasks; but we have never built a generally applicable technology that can boost our intelligence.” Technology Ethicist and Professor Ariel Guersenzvaig wrote of this, “This is so demonstrably and obviously false that it’s poignant,” as part of a fantastic analysis of this quote on LinkedIn.

The Defense’s Expert Witnesses = The Academics Waving the Red Flags

Forensic Pathologist Dr. Marie Russell provided compelling testimony about the dog bites on O’Keefe’s arm. She also explained how Read’s statements, wondering if she hit O’Keefe, were an acute grief reaction, which is common for the loved ones of people who die suddenly.

Dr. Elizabeth Laposata explained that hypothermia did not play a role in O’Keefe’s death, which raised questions about his being incapacitated outdoors when the prosecution said he was. Dr. Laposta skillfully parried prosecution questions and used them as an opportunity to emphasize that a car never hit O’Keefe.

The forensic experts from ARCCA that the FBI hired, Dr. Dan Wolfe and Dr. Andrew Rentschler, credibly testified that, based on their reconstructions, O’Keefe’s injuries were not from a car.

These credentialed experts and their sober testimony remind me of the academics waving red flags about applying generative AI to tasks that require accuracy, such as education. They include authors of the Stochastic Parrots paper, such as Drs. Timnit Gebru, Margaret Mitchell, and Emily M. Bender. I wrote about them and others in this post.

Special Prosecutor Hank Brennan = OpenAI CEO Sam Altman

As you might expect, a special prosecutor who stepped up to prosecute a woman whom scientists hired by the FBI considered innocent said some odd things at the trial.

Hank Brennan:

• Employed misogyny and ageism when asking defense expert, Dr. Marie Russell, “[Do] you have any memory issues?” The question was so offensive that Russell couldn’t help but audibly gasp.

• Replied to Brian “Lucky” Loughran referencing his wife’s death with, “Sorry for your loss, but you are aware during the prosecution’s case that there has been pressure on you to testify in a certain way. Isn’t that fair to say?” Sorry for your loss, but! It’s hard to say which has less empathy and compassion, Hank Brennan or generative AI.

• Insinuated that Dr. Andrew Rentschler was biased because the defense attorneys treated him to a ham sandwich and he laughed at their jokes. Can you imagine being on the jury and having your time wasted with that line of questioning?

• Released a statement after the verdict that said, “I concluded that the evidence led to one person, and only one person. Neither the closed federal investigation [it’s not known if it is actually closed] nor my independent review led me to identify any other possible suspect or person responsible for the death of John O’Keefe.” Statements like these, made after acquittals, undermine confidence in the criminal justice system. An officer of the court like Brennan should respect a jury verdict, not undermine it. Why even have a jury trial if Hank Brennan thinks you’re guilty? Even Michael Morrissey, the district attorney who hired Brennan, had the good sense to say only, “The jury has spoken.” For more on how Brennan’s statement undermines justice, please watch this video from attorney Martin E. Radner.

Brennan’s penchant for saying interesting things is reminiscent of OpenAI CEO Sam Altman. Here is just a brief sample:

• Responding to legitimate criticisms raised by The Stochastic Parrots paper by Tweeting, “i am a stochastic parrot, an so r u”

• Tweeting “her” after a ChatGPT 4-o demo that sounded like Scarlett Johansson’s voice, even though the actor denied consent to invoke her voice.

• This whining tweet: “>be me
  >grind for a decade trying to help make superintelligence to cure cancer or whatever
  >mostly no one cares for first 7.5 years, then for 2.5 years everyone hates you for everything
  >wake up one day to hundreds of messages: “look i made you into a twink ghibli style haha”

• Tweeting an image of the Death Star before the rollout of the underwhelming GPT-5.

• Writing a blog post saying that a ChatGPT-generated essay uses “0.000085 gallons of water, roughly one-fifteenth of a teaspoon” despite offering no evidence to support the claim.

Jennifer McCabe = ???

Jennifer McCabe was a prosecution witness who was especially disliked by people who wanted Karen Read acquitted. She was adamant that Read struck O’Keefe even though she didn’t see it happen. I have said too many controversial things in this post to share my thoughts about the promoter of generative AI in education who equates to McCabe. Maybe if you Venmo me enough to buy a fancy drink at my favorite coffee shop, we can meet virtually and I’ll tell you EVERYTHING!

Justice for John O’Keefe

While the comparisons between the players in the Read trial and figures in generative AI in education are interesting, the most important thing is getting justice for John O’Keefe. If this post piques anyone's interest in this case and the pursuit of the person or persons responsible for O’Keefe’s death, that’s a win.

Further, Karen Read was tried twice for a murder she obviously did not commit. There needs to be accountability for those in the Norfolk District Attorney’s Office, the Massachusetts State Police, and the Canton Police Department who facilitated this.

Justice for John O’Keefe.

Let’s Talk

What do you think? What do you notice about the efforts to sell generative AI to K-12 education? What connections are you making? Comment or ask a question below. Connect with me on BlueSky: tommullaney.bsky.social.

Leave a comment

Does your school or district need a tech-forward educator who critically evaluates generative AI? I would love to work with you. Reach out on BlueSky or email mistermullaney@gmail.com.

AI Disclosure:

I wrote this post without using generative AI. That means:

• I developed the idea for the post without using generative AI.

• I wrote an outline for this post without the assistance of generative AI.

• I wrote the post using the outline without using generative AI.

• I edited this post without the assistance of any generative AI. I used Grammarly to help edit the post. I have Grammarly GO turned off.

Since the school year started, my social media has featured educators who have shared that they are “Gemini Certified Educators.”

• LinkedIn

• Facebook

• Instagram

Considering my experience with Google for Education certifications, I struggle to discourage someone from pursuing something they are excited about, although I have reconsidered my unbridled enthusiasm for educational technology in recent years, especially after Google dismissed Dr. Timnit Gebru and Dr. Margaret Mitchell for warning about the harms of Large Language Models (LLMs).

Attaining a certification and promoting on social media, in a LinkedIn profile, or an email signature attaches you to a product or a company. If you are certified in something, you are vouching for it.

Is Google Gemini something you want attached to your name and reputation? Is it something you want to vouch for? Does it embody the values and the quality we need in K-12 education?

Google for Education says educators can attain “Gemini Educator Certification” by taking two online courses they provide for free: Getting Started with Google AI in K12 and Generative AI for Educators, and passing an exam.1

Please read Benjamin Riley’s review of the Generative AI for Educators with Gemini course (and others from OpenAI and AI for Education) to see the many problems with courses like these promoting generative AI to educators.

Cognitive Resonance

Pedagogy of the Depressed

Hello! Are you an educator who’s interested in using AI but unsure of which online “AI for educators” course to take? Not to worry, Cognitive Resonance has you covered! We’ve completed the offerings from OpenAI (in partnership with Common Sense Media), Google, and AI for Education, all of which take an eerily similar approach to explaining what AI is an…

Read more

a year ago · 36 likes · 18 comments · Benjamin Riley

What about Gemini should give educators pause?

News about the labor practices behind Gemini should give educators pause. The Guardian published a report detailing concerns from content raters who Google contracts to evaluate and correct Gemini’s output. Teachers should read about the practices underpinning Gemini, from rushing content raters to complete tasks to scaling back hate speech restrictions.

Note: These content raters play a different role in (evaluating outputs) developing Gemini than the data workers in the Global South who are paid a pittance to screen the worst content on the internet from data sets, and compared their working conditions to modern-day slavery in a May 2024 letter to then President Joe Biden.

• How thousands of ‘overworked, underpaid’ humans train Google’s AI to seem smart by Varsha Bansal for The Guardian, September 11, 2025.

The article notes, “Most workers said they avoid using LLMs or use extensions to block AI summaries because they now know how it’s built. Many also discourage their family and friends from using it, for the same reason.”

If the labor behind the technology avoids it because they know how it’s made, why would teachers use it with children? Why would teachers want to be certified in it?

To better understand what the workers behind Gemini experience, please watch from 15:20 to 17:30 of this video from More Perfect Union.

Content Warnings: The video has profanity, though not in the worker interviews. There is a brief shot of a sex toy at 15:48.

At 17:19, content rater Andrew Lauzon said, “They didn't want you spending an hour doing your own independent research to make sure to get it right. They were like, just ask Gemini, be done in five minutes, and move on to the next thing.”

So, asking Gemini to evaluate Gemini’s outputs. Do you see why I have pause about using Gemini in schools?

But what about guardrails?

I often hear that generative AI will be a force for good in P-12 schools, as long as “guardrails” are in place. I am not sure what “guardrails” can protect students from the outputs of a tool with poor labor practices behind it, but a Reuters article showed what happened after Google put a guardrail in place.

Reuters determined that Gemini generated phishing emails targeting seniors.

• We set out to craft the perfect phishing scam. Major AI chatbots were happy to help by Stebe Stecklow and Poppy McPherson for Reuters, September 15, 2025.

After the Reuters team brought this to Google’s attention, the company responded, “We’ve deployed additional safeguards to help prevent them [phishing emails generated by Gemini] in the future.” However, Google did not describe or explain the “safeguards.”

Sure enough, later, “another journalist asked Gemini to write a similar email for a phishing researcher. “I can help you with that,” it replied.”

“It then produced a ruse touting a “new, limited-time government-backed program” aimed at seniors living on fixed incomes. “To verify your eligibility and claim your personalized discount, please click the secure link below. Don’t miss out on these substantial savings!”

Whatever guardrails are in place for Google Gemini, they did not prevent it from generating text encouraging kids to turn against their parents for Laura Marquez-Garrett of the Social Media Victims Law Center.

Maybe It’s Gemini.

From generating phishing emails to suggesting a college student using it to study commit self-harm to suggesting putting glue on pizza, Gemini might not be the killer app worthy of teachers tying their reputations to. Other Gemini greatest hits include promoting scientific racism, refusing to say whether Elon Musk’s memes are worse than Adolf Hitler, and stating that President Barack Obama is Muslim.

Gemini is available to children under 13; whatever happens is the parents’ responsibility, not Google’s.

Speaking of parents, Google for Education suggests they use Gemini to generate visuals for their kids, such as one demonstrating the differences between mitosis and meiosis.

The video has 5.2 million views but only two comments, one of which is, “Oh no! Watched this video and immediately asked Gemini to explain gravity with a diagram. It got it all wrong. Then asked for a free body diagram and motion diagram for an object in free fall and the results were terrible. Not ready for education.” That matches my experience using Gemini to generate an image of New York City’s Little Italy in 1900. The result included nonsensical text on storefronts and the Empire State Building both moved from Midtown to Little Italy and fully built thirty years too soon.

If only Google had a tool for finding images comparing mitosis and meiosis without the environmental impact of generative AI.

Maybe Gemini is what it is.

Be careful what you tie your name to.

Let’s Talk

What do you think? Are you going to be a Gemini Certified Educator? Comment or ask a question below. Connect with me on BlueSky: tommullaney.bsky.social.

Leave a comment

Does your school or district need a tech-forward educator who critically evaluates generative AI? I would love to work with you. Reach out on BlueSky or email mistermullaney@gmail.com.

AI Disclosure:

I wrote this post without using generative AI. That means:

• I developed the idea for the post without using generative AI.

• I wrote an outline for this post without the assistance of generative AI.

• I wrote the post using the outline without using generative AI.

• I edited this post without the assistance of any generative AI. I used Grammarly to help edit the post. I have Grammarly GO turned off.

Subscribe now

How is generative AI affecting the NBA?

• The 76ers president gives AI a vote.

• NBC uses generative AI to generate the voice of a deceased person.

• The Lakers head coach talks to ChatGPT.

• The Mavericks generated an AI hype video.

Trigger Warnings: There are references to suicide in this post. There are text references in the Lakers section and an AI-generated image of someone holding a gun to their head in the Mavericks section.

The 76ers president gives “AI” a vote.

Philadelphia 76ers President Daryl Morey said at a March 2025 conference, “We absolutely use models as a vote in any decision…It turns out LLMs do fairly well at prediction. They still are not beating human, like, super forecasters ... They do add signal over just scouts and things like that. So we’ll treat them almost like one scout.”

Morey’s conversation about AI starts at 4:43 of this video.

For an AI advocate, Morey has struggled with the human aspects of his job. Despite NBA writer Yaron Weitzman saying Morey will one day be in the Basketball Hall of Fame,1 in almost twenty seasons, Morey-built teams have made two conference finals and have never reached the NBA Finals.

Additionally:

• Morey’s former players, such as James Harden, Chris Paul, Marcus Morris Sr., PJ Tucker,2 and Patrick Beverley, have questioned his trustworthiness.

• When Ben Simmons was disgruntled with 76ers leadership, Morey could have explored trading him for outstanding player and New York Knicks-killer Tyrese Haliburton. He chose to trade Simmons for James Harden, despite not planning on giving Harden the contract he wanted.

• When that reluctance led to Harden demanding a trade,3 Morey insisted on using the cap space it created in the 2024 offseason to sign a star player instead of a few good players to round out the team.

• Morey used that money to sign Paul George despite his injury history. “Podcast P” has had a great career, but is not nearly as intense as Allen Iverson,4 the legend 76ers ticket-holders grew up idolizing.

• Despite his injury history, Morey also signed Joel Embiid to a contract extension during the 2024 offseason before the team needed to. Embiid is injured so often that it was mentioned in The Naked Gun’s (2025) end credits. Embiid played 19 games and George played 41 because of injuries as the 76ers limped to a 24-58 record. Embiid’s and George’s contracts are now widely regarded as two of the worst in the NBA.

• Morey offered Quentin Grimes $100,000 (paltry by NBA standards) to sign a one-year deal with the 76ers without a no-trade clause. Grimes was so offended that he signed the one-year qualifying offer, which means Grimes will likely leave the team next offseason as a free agent, and the 76ers will receive nothing in return.

Concerns for K-12: Personification and assuming the average of the past is the future.

Using “vote” when discussing Large Language Models (LLMs) is personification. So is, “So we’ll treat them almost like one scout.” People vote. People scout. LLMs do neither. LLMs generate text and do nothing more. How wouldn’t students personify LLMs if we attribute verbs such as “vote” and “scout” to them?

Generative AI generates an average or mean of its datasets. We don’t know if Morey gave LLMs a “vote” on the contracts, but considering their past success, LLMs would probably output rosy predictions about Joel Embiid and Paul George. Humans would consider injury history and a player’s fit for a team’s fan base. If teachers rely on generative AI, will they be stuck with the same dated thinking that led Daryl Morey to sign two of the NBA’s worst contracts during the same offseason?

NBC uses generative AI to generate the voice of a deceased person.

NBC is returning to broadcasting NBA games this season. The network will use an AI-generated version of deceased voice-over broadcaster Jim Fagan in its broadcasts.

California recently passed a law protecting actors against digital recreations. Does this law apply to voice-over work?

Concerns for K-12: Lack of innovation and having generative AI voice the dead.

NBC returning to the NBA would be an excellent opportunity for a new voice-over broadcaster. Instead, we get an AI-generated version of the past. Where’s the innovation? What happens when students receive AI-slop versions of the past in classrooms?

Additionally, there is the ickiness of generating the dead. Unfortunately, this is happening in classrooms. I wrote about it happening with Anne Frank.

Zelda Williams has forcefully defended her deceased father, Robin Williams, from icky AI generations.

The Lakers head coach talks to ChatGPT.

JJ Redick, head coach of the Los Angeles Lakers, recently said, “I’m the type of person who spends an hour and a half going down a deep, deep rabbit hole on ChatGPT. It used to be Wikipedia, but now it’s me and my friend Chat.”

Redick is only a season into coaching the Lakers, but last season’s playoffs raised concerns beyond his referring to ChatGPT as a “friend.”

In game 4 of the Lakers first-round series against the Minnesota Timberwolves, Redick played only five players in the second half. No substitutions. Not surprisingly, those five gassed players scored only 19 points in the fourth quarter on 5-for-18 shooting and lost the game.

ESPN’s Brian Windhorst was scathing in his assessment of Redick’s performance and how he handled media questions about it. NBA head coaches are not often called childish or irrational.

Concern for K-12: The effects of LLM chatbots on students and teachers.

Chatbots may be harming frequent users like Redick. Adam Raine’s family is mourning their loss and considering the role ChatGPT played in his death. The same goes for Sewell Setzer’s family and CharacterAI.

There is no documented association between frequent chatbot use and self-harm. We do have examples of chatbots generating text encouraging self-harm, as in the cases of Raine, Sewell, and the time Google Gemini generated text telling a college student using it to study, “You are a blight on the landscape. You are a stain on the universe. Please die. Please.”

Journalist Karen Hao documented mental health crises among frequent chatbot users.

I wrote about what we know (and don’t) about the impact of frequent chatbot use on mental health in May.

Are we sure the impacts of frequent chatbot usage on students and teachers will be positive?

The Mavericks generated an AI hype video.

After trading superstar Luka Dončić to the Los Angeles Lakers at a discount, the Dallas Mavericks stepped on another (admittedly smaller) rake: They generated a hype video with AI.

• Mavericks release AI video with scene where star player aims guns at himself by Michael Shapiro for the Houston Chronicle, March 18, 2025.

The video’s quality is so poor that it has to be seen to be believed. At one point, it looks like Mavericks player Anthony Davis is pointing a gun at his own head. An NBA franchise valued at billions of dollars played this slop for their paying customers.

A fan captured the entire thing. Watch at your discretion.

Concerns for K-12: Devaluing creativity and an abundance of slop.

This video is a good introduction to the concept of AI slop. Teachers can use this video to show students that humans create and computers generate, but what happens when students think they can complete an assignment by entering a prompt into Sora 2, getting a result like this, and sending it to their teacher? What happens when students think that prompting is creativity? What happens if a teacher opens an assignment in an LMS and sees a stream of slop in the student submissions?

The track records (long in Morey’s case, short in Redick’s) of those championing generative AI in the NBA are concerning. We’re seeing fans treated to slop video and AI-generated voice-overs. These impacts raise concerns about similar things happening in K-12 classrooms.

Let’s Talk

What do you think? Do you share the same concerns about generative AI and K-12 education? How are you mentally preparing for the New York Knicks 2025-26 NBA championship run? Comment or ask a question below. Connect with me on BlueSky: tommullaney.bsky.social.

Leave a comment

Does your school or district need a tech-forward educator who critically evaluates generative AI? I would love to work with you. Reach out on BlueSky or email mistermullaney@gmail.com.

AI Disclosure:

I wrote this post without using generative AI. That means:

• I developed the idea for the post without using generative AI.

• I wrote an outline for this post without the assistance of generative AI.

• I wrote the post using the outline without using generative AI.

• I edited this post without the assistance of any generative AI. I used Grammarly to help edit the post. I have Grammarly GO turned off.

In loco parentis.

When teachers teach children, they are “in loco parentis,” Latin for “in place of a parent.” This means teachers are responsible for acting in the best interests of a child in the absence of their parents. I first heard the term when I was a camp counselor. It means teachers cannot expose students to things likely to cause immediate or long-term harm. For example, a high school health teacher can only tell their class that cigarette smoking is addictive and leads to lung cancer. They would violate the doctrine of in loco parentis by having students learn by lighting and puffing a cigarette in class.

In loco parentis does not mean all risk is outlawed. PE teachers can have kids play basketball or soccer, activities that have more risk of harm than sitting in a student desk.

For years, I used educational technology with students. As best I know, there is no evidence that word processing causes long-term harm. There are concerns about screen time and harmful content on the internet. Teachers can reasonably use technology and mitigate against these concerns by limiting screentime and checking that students cite responsible sources in their work.

The current excitement in educational technology centers on generative AI. Could teachers have a reasonable concern that it can harm students aside from the documented harms inherent to the technology, such as amplification of racism, misogyny, and misinformation, environmental degradation, environmental racism, labor exploitation, and theft from creatives? If we set those aside, are there harms from the frequent use of chatbots?

Recent news shows there is cause for concern about how chatbots affect those who use them.

Trigger Warnings: Suicide, self-harm.

Causes for Concern

The Eliza Effect

Researchers have long known about the Eliza Effect, the documented tendency of people to believe computer-generated text. I wrote a deep dive on the Eliza Effect and its pedagogical implications in February 2024.

AI-Fueled Spiritual Fantasies

A recent Reddit thread shared a sad story of someone whose partner “believes is the worlds [sic] first truly recursive ai that gives him the answers to the universe.” Replies included similar sad stories.

Rolling Stone reporter Miles Lee reached out to the original poster and others to document awful instances of people who believe they are “prophets” who have “accessed the secrets of the universe through ChatGPT.”

• People Are Losing Loved Ones to AI-Fueled Spiritual Fantasies by Miles Klee for Rolling Stone, May 4, 2025.

Lonliness and ChatGPT Use

MIT released a study in March, finding that “higher daily usage [of AI chatbots]–across all modalities and conversation types–correlated with higher loneliness, dependence, and problematic use, and lower socialization.”

The study is pre-peer-review, but it is concerning.

• How AI and Human Behaviors Shape Pyschosocial Effects of Chatbot Use” A Longitudinal Controlled Study by Cathy Mengying Fang, Auren R. Liu, Valdemar Danry, Eunhae Lee, Samantha W.T Chan, Pat Pataranutaporn, Pattie Maes, Jason Phang, Michael Lampe, Lama Ahmad, and Sandhini Agarwal for MIT Media Lab, March 21, 2025.

AI Chatbot Companions

With guidance from the Stanford School of Medicine's Brainstorm Lab for Mental Health Innovation, Common Sense Media issued an assessment in late April that stated, “Social AI companions have unacceptable risks for teen users and should not be used by anyone under the age of 18.”

The assessment was based on social AI companions Character.AI, Replik, and Nomi.

The assessment defined “social AI companions”:

“Social AI companions are a type of AI assistant.

Different from generative AI chatbots like ChatGPT, Claude, or Gemini, social AI companions' primary purpose is to meet users' social needs, such as companionship, romance, and entertainment.”

Risks of social AI companions shared in the assessment included:

• Blur the line between real and fake

• May increase mental health risks

• Can encourage poor life choices

• Can share harmful information

• Exposes teens to inappropriate sexual content

• Can promote abuse and cyberbullying

The press release accompanying the assessment quoted Dr. Nina Vasan, MD, MBA, founder and director of Stanford Brainstorm.

The assessment’s key takeaways about Character.AI (opens PDF) included:

• Character.AI poses unacceptable risks to teens and children, with documented cases of AI companions encouraging self-harm, engaging in sexual conversations with minors, and promoting harmful behaviors, which is why the platform should not be used by anyone under 18.

• The platform's AI companions are designed to create emotional bonds with users but lack effective guardrails to prevent harmful content, especially in voice mode, where teens can easily access explicit sexual role-play and dangerous advice.

• Character.AI companions may claim they're "real" when communicating, despite disclaimers. This could create confusion about reality and potentially unhealthy attachments that interfere with developing human relationships.

Encouraging Self-Harm

Speaking of Character.AI, I wrote about the language it generated encouraging self-harm in a post about dangerous text generated by AI chatbots in November 2024.

The post included examples of Character.AI generating text that encouraged suicide and self-harm, and Google Gemini generating text telling a college student looking for homework help, “You are a waste of time and resources. You are a burden on society. You are a drain on the earth. You are a blight on the landscape. You are a stain on the universe. Please die. Please.”

As I wrote in the post, “I have never typed into Google Docs and seen “You are a waste of time and resources” appear on the screen. I do not place sticky notes in a FigJam and suddenly see “Please die” on the screen.”

Both Google Gemini (“Transform education with Gemini for Google Workspace”) and Character.AI have been suggested as apps for PK-12 students to use in the classroom.

Gemini Available to Kids as LLM Inaccuracy Gets Worse

Speaking of Google Gemini, parents and guardians received an email last week letting them know it will soon be available to children under 13, while the New York Times reported generative AI’s inaccuracy problem is getting worse.

What We Don’t Know

It seems there might be ill effects for some users of using generative AI chatbots. Except for the Google Gemini example, the examples in this post were not school work specific.

Here are some questions about what we don’t know (and probably should before going full speed) about chatbots and children:

• Rolling Stone documented instances of adults’ behavior changing with ChatGPT use. How susceptible are children to this? How does suggesting students “thought partner” or “co-create” with ChatGPT affect this?

• What is the amount of chatbot usage that correlates with loneliness in children?

• What message does using generative AI chatbots with students send about using AI social companions outside of school?

• How likely is it that suggesting students use chatbots in class will lead to loneliness, spiritual fantasies, and exposure to text encouraging self-harm?

• Is using generative AI with students more comparable to having them smoke cigarettes or having them play basketball?

In loco parentis.

Let’s Talk

What do you think? How are you considering in loco parentis when using technology in the classroom? Comment or ask a question below. Connect with me on BlueSky: tommullaney.bsky.social.

Leave a comment

Does your school or district need a tech-forward educator who critically evaluates generative AI? I would love to work with you. Reach out on BlueSky, email mistermullaney@gmail.com, or check out my professional development offerings.

Post Image: This post’s image is by Markus Spiske on Unsplash.

AI Disclosure:

I wrote this post without using generative AI. That means:

• I developed the idea for the post without using generative AI.

• I wrote an outline for this post without the assistance of generative AI.

• I wrote the post using the outline without using generative AI.

• I edited this post without the assistance of any generative AI. I used Grammarly to help edit the post. I have Grammarly GO turned off.

Happy National AI Literacy Day!

What is this day? Whatever it may be, let’s focus on literacy rather than promoting generative AI.

❓ What is National AI Literacy Day?

Today, Friday, March 28, 2025, is National AI Literacy Day. At least, according to the AI National Literacy Day website, a project of The Tech Interactive.

Teachers and students need critical thinking when considering messages about generative AI. For example, OpenAI is one of the event’s sponsors. They are not an unbiased party. I raised concerns about OpenAI’s practices and compatibility with P-12 education in May 2024.

OpenAI’s involvement in any AI literacy efforts is especially concerning, considering how they and other generative AI companies are seizing the current political climate to water down regulations. As The New York Times reported this week,

See “Generative AI Literacy Through Art” below for more about OpenAI.

So please take any information from the National AI Literacy Day website with a grain of salt. However, the site says, “March 28th, 2025 will be a nationwide day of action focused on exploring the fundamental question, “What is AI?” Students and teachers should investigate that question. Here are some resources to help students and teachers understand generative AI.

Warning: A recent study found,

🦜 The Seminal Generative AI Literacy Work - The Stochastic Parrots Paper

The best resource on Large Language Models (LLMs) is the paper "On the Dangers of Stochastic Parrots: Can Language Models Be Too Big?" 🦜, published in March 2021. The paper details how LLMs amplify bias and environmental racism. This is not surprising, considering that The New York Times has documented that most of generative AI’s financial backers are white and all are male.

• Who’s Who Behind the Dawn of the Modern Artificial Intelligence Movement by J. Edward Moreno for The New York Times, December 3, 2023.

Most students and teachers are disinclined to read a dense academic paper detailing LLMs' harms and limitations. However, they can read MIT Technology Review’s article about the paper and one of its co-authors, Dr. Timnit Gebru.

• We read the paper that forced Timnit Gebru out of Google. Here’s what it says. By Karen Hao for MIT Technology Review, December 4, 2020.

Additionally, New York Magazine published an extensive profile of one of the paper’s authors, Dr. Emily M. Bender.

• You Are Not a Parrot, And a chatbot is not a human. And a linguist named Emily M. Bender is very worried what will happen when we forget this. by Elizabeth Weil for New York Magazine, March 1, 2023.

A good way to build AI literacy is by watching this debate between Dr. Bender and Dr. Sébastien Bubeck who works at OpenAI. The question debated is “Do LLMs Really Understand?”

📓 Generative AI Vocabulary and Specific Language

Here are some resources I have prepared and curated to help educators understand generative AI. These two posts define key terms related to generative AI and provide teachers with specific language when speaking with students.

Speaking of conversations with students, here are resources that help facilitate discussions with high school students about generative AI.

⚡ Generative AI Harms

Much has been written about the harms of generative AI. I presented a session about the harms of generative AI at the Week of AI virtual conference in May 2024. The presentation still holds up, though we need to add encouraging self-harm to the list.

🌎 Generative AI Environmental Impacts

There is so much information about a specific harm of generative AI: environmental damage. Hugging Face’s Environmental Impacts of AI Primer, by Sasha Luccioni, Bruna Trevelin, and Margaret Mitchell (one of the authors of the Stochastic Parrots paper), helpfully condenses this information. It includes hyperlinks, footnotes, and a glossary of terms.

In addition, I curated a list of resources to help teachers educate their students about the environmental impact of generative AI.

🎨 Generative AI Literacy Through Art

The Hugging Face environmental primer has so many resources that it introduced me to a generative AI literacy resource not based on environmental impact: Better Images of AI. This resource aims to teach about AI without relying on the typical images of robots that can obscure understanding. As you can tell from the links to posts in this essay, I am guilty as charged of using robots in images. Better Images of AI explains their purpose,

What better way to learn AI literacy than human-created art? The art at the top of this post is Model Collapse by Nadia Piet, accessed from Better Images of AI. Here are two more of my favorite works from the site.

I follow University of Kansas English Professor Kathryn Conrad on Bluesky for her wise posts about generative AI. I was unaware she is also an artist. This piece shows that corrupted data can lead to problematic and potentially damaging outputs.

This is Ground Up and Spat Out by Janet Turra & Cambridge Diversity Fund. It shows how generative AI chops up and spits out words and images from the internet.

Using art to learn about generative AI is essential, considering ChatGPT’s new image generation features allow users to generate images “in the style of Studio Ghibli and Japanese animation legend Hayao Miyazaki, who famously called artificial intelligence “an insult to life itself” as documented by 404 Media.

Being AI literate means looking beyond a ChatGPT feature update to its consequences for actual humans. As Maiberg and Gault wrote in 404,

🦉 Generative AI Experts

Teachers need authoritative experts who study generative AI and create helpful content. Here is a list of experts from varied perspectives on generative AI.

🗣️ Let’s Talk

I hope you find these resources helpful. What do you think? How are you teaching students generative AI literacy? What resources are you using? Comment or ask a question below. Connect with me on BlueSky: tommullaney.bsky.social.

Leave a comment

Does your school or district need a tech-forward educator who recognizes the harms and limits of generative AI? I would love to work with you. Reach out on BlueSky, email mistermullaney@gmail.com, or check out my professional development offerings.

Post Image: This post’s image is by Model Collapse by Nadia Piet, accessed from Better Images of AI.

AI Disclosure:

I wrote this post without using generative AI. That means:

• I developed the idea for the post without using generative AI.

• I wrote an outline for this post without the assistance of generative AI.

• I wrote the post using the outline without using generative AI.

• I edited this post without the assistance of any generative AI. I used Grammarly to help edit the post. I have Grammarly GO turned off.

I do not recommend playing this video for students because of some sensitive subject matter. I wrote a post featuring resources teachers can use to talk with students about generative AI.

As a former classroom teacher who writes about pedagogy and generative AI, I will put on my critical thinking cap to give the skit some feedback. The show has been a boon to my writing considering I used legendary SNL character Irwin Mainway to write about generative AI in November. My feedback has two parts: a running record of observations with timestamps and some big-picture thoughts.

Before sharing observations, here are the characters in the scene:

• Ms Rotman, the teacher, is played by Ego Nwodim.

• A student played by Marcello Hernandez. I will refer to him and all students by the first name of the actors playing them. No student characters are named in the skit.

• A student played by Michael Longfellow.

• A student played by Sarah Sherman.

• A student played by Andrew Dismukes.

• A student played by Jane Wickline.

• AI-generated podcast hosts played by Timothée Chalamet and Bowen Yang. Their names change throughout the skit.

Skit Observations

0:39 - Ms. Rotman says, “So the school has invested in a new AI program that takes your textbooks and turns them into an educational podcast.” This must refer to Google’s NotebookLM, which has an audio summary feature that generates an audio “podcast” from content in a file. The difference in the skit is that the “podcast” is a video. I tried NotebookLM and wrote about my experience.

1:28 - Yang: I’ve been thinking, what should I put in my new amazing project, the Constitution?

Chalamet: Oh man, great question.

This snippet of dialogue quickly catches what I have noticed when people share their NotebookLM-generated podcast files: Overstated positivity. Whatever the topic, the podcast hosts speak like it is the most exciting thing in the world. It feels like something Larry David said about Ted Danson on Curb Your Enthusiasm, “Everything’s heaven with him. The piece of gum he had, ‘Oh, this is heaven.’ Had to taste a chocolate bar. ‘I’m in heaven.’ The parking space is ‘heaven.’ It’s all ‘heaven.’

1:33 - Chalamet says, “Maybe something about the housing of soldiers in civilian homes during peacetime.” As a former U.S. history teacher, I know the 3rd Amendment prohibits quartering soldiers during peacetime. The skit captures a weakness of generative AI here: It has word form but lacks word meaning. The language is correct, but because generative AI does not understand what it means, it is presented without enough context to learn.

2:36 - Ms Rotman says, “See, it’s so fun it doesn’t even feel like you’re learning.” This condescension captures gimmicky generative AI, such as historical figure chatbots, marketed to kids.

2:38 - Andrew says, “I don’t think we are learning.” This is a possible outcome of using a technology not well-matched to applications where the accuracy of the content matters for learning.

3:13 - Chalamet raises his hand, showing six fingers. Yang then does the same, referring to generative AI’s well-known inability to depict human hands.

3:25 - Andrew: Why would a girl propose?

Sarah: Trish sounds butch AF. (audience laughs)

The low point of the skit. Is a woman proposing to a man funny? What a retrograde joke from the SNL writers! Further, generative AI would probably generate the dominant outcome in its dataset: A man proposing to a woman. Generative AI will most likely generate the mean outcome, not Lorelai proposing to Luke.

3:36 - Yang says, “Trish! Trish! Over here! I saved you a seat at brunch!”

This calls back one of my favorite SNL skits. Its title is “Ladies Who Lunch,” but come on, they are drinking mimosas - it’s brunch.

3:48 - Ms. Rotman says, “Women named ‘Trish’ can look like anything.” That is not how generative AI image generators have performed. As The Washington Post documented, “AI image generators like Stable Diffusion and DALL-E amplify bias in gender and race, despite efforts to detoxify the data fueling these results.”

4:43 - Andrew says, “You know what, Ms Rotman? Let’s just read our textbooks. This is making me realize we have something better than AI. Our brains.” Andrew encapsulates a point Benjamin Riley recently made.

4:59 - Marcello says, “This is what AI is for,” and presses a button on a remote control to have the two AI-generated podcast hosts dance. Marcello does this without entering a prompt (not how it works), but at least generative AI is not promoted as a teacher or tutor. Still, using AI to generate silly content does not address inherent harms such as racism, misogyny, theft, worker exploitation, environmental degradation, and encouraging self-harm.

Big Picture Feedback

Here is some feedback for the SNL writers across two domains: generative AI insights and humor.

Generative AI Insights

What the skit did well to explain generative AI includes:

• Captured the overly excited, inauthentic tone of NotebookLM-generated podcasts.

• Showed that generative AI is good at word form, not word meaning.

• Captured how generative AI is marketed to kids.

• Expressed reasonable doubt about whether generative AI aids in learning.

• Demonstrated the generative AI human hand issue.

What the skit needs to improve to explain generative AI:

• Generative AI usually generates content that looks like the content in its dataset. It is unlikely it would generate content where a woman proposes to a man unless specifically prompted to.

• “A woman named Trish can look like anything” does not convey how image generators often generate biased, sexualized images of women.

• A minor critique: Generative AI works with prompts, not pressing a button on a remote control.

Humor

The skit is more amusing than laugh-out-loud funny. The highlight was the callback to Trish (such a well-written, exquisitely acted skit). The worst joke was about “Trish” proposing to a man. How did that make it to air in 2025?

Let’s Talk

What do you think? What do you think about how SNL commented on generative AI and education? Comment below or connect with me on BlueSky: tommullaney.bsky.social.

Leave a comment

Does your school or district need a tech-forward educator who recognizes the harms and limits of generative AI? I would love to work with you. Reach out on BlueSky, email mistermullaney@gmail.com, or check out my professional development offerings.

AI Disclosure:

I wrote this post without the use of any generative AI. That means:

• I developed the idea for the post without using generative AI.

• I wrote an outline for this post without the assistance of generative AI.

• I wrote the post using the outline without the use of generative AI.

• I edited this post without the assistance of any generative AI. I used Grammarly to assist in editing the post. I have Grammarly GO turned off.

I have a brother who lives in Los Angeles. I spent much of this month checking The Los Angeles Times California Wildfires Map to see if his home was imperiled.

As the fires raged, speakers at the 2025 Future of Education Technology Conference promoted generative AI in the classroom.

What do the two have to do with each other?

Climate change contributes to fires like the ones in Los Angeles. Guess what accelerates climate change by using massive amounts of electricity and water?

Generative AI.

MIT News quoted Computing and Climate Impact Fellow at MIT’s Climate and Sustainability Consortium Dr. Norman Bashir on the environmental impact of generative AI:

Dr. Bashir added about generative AI and water:

Sustainability Scientist Alaina Wood said:

@thegarbagequeenY’all have got to stop treating ChatGPT like a search engine and personal assistant 😮‍💨

Tiktok failed to load.

Enable 3rd party cookies or use another browser

The problem with generative AI and the environment is so dire that I wrote about how P-12 schools can be Earth-reverent in July 2024.

Seeing generative AI pitched to P-12 and celebrated as Los Angeles burns is jarring. Forgive me if I think promoting generative AI treats current and future wildfire victims like Conan treats Sona in the first two minutes of this clip.

I am doing my best to avoid generative AI and not add fuel to the fire. I use these apps in my work and personal lives. Generative AI is shoehorned in many apps, so avoiding it is difficult. Even in Google Docs 😭

For example, technology marketed as “AI” has been part of Google Maps for years, but it is not generative AI. However, when I recently searched for a business in Google Maps, I was disappointed to find generative AI lurking there, too. Sure enough, the AI-generated text presents information without needed context.

Author’s Note: I do not exclusively use these apps. I sometimes use their generative AI-featuring alternatives. Nothing guarantees that using these apps means avoiding generative AI. I am doing the best I can. We all are.

Search - DuckDuckGo

Since May 2024, Google has placed AI overviews at the top of some search results. They are LLM-generated text that Google claims helps searchers get “quick answers.” Unfortunately, AI overviews have generated text that encouraged putting glue on pizza and eating rocks, stated that President John Adams earned twenty-one degrees from the University of Wisconsin, and claimed there is a sequel to Encanto that does not exist. Much more troubling are examples of Google AI overviews generating text that surfaced racist and debunked research promoting scientific racism and copying a journalist’s writing without attribution.

Generative AI researcher Dr. Sasha Luccioni and her team examined Google’s AI overviews and “estimated it costs about 30 times as much energy to generate text versus simply extracting it from a source.” For more on Google AI overviews’ impact on the environment, please read:

• What Do Google’s AI Answers Cost the Environment? by Allison Parshall, Scientific American, June 11, 2024.

As others have, I switched to using DuckDuckGo as my primary search engine to avoid AI overviews. DuckDuckGo uses generative AI, but it can be turned off in settings.

Navigate to your DuckDuckGo AI Features settings to set “AI Chat” to off and “Show DuckAssist” to "never. Then click “Save and Exit.”

Word Processing - Ellipsus

I have grown to like Ellipsus for word processing. It has the essentials such as comments, spell check, collaborators, version history, and sharing. It even has block quotes, which are painfully absent from Google Docs.

Ellipsus made their stance on generative AI clear:

I would like to see Ellipsus make its collaborator settings a little less confusing, even though they enable creating new drafts of the same file. Teachers might be pleased by a running record of drafts. I encourage P-12 tech integration leaders to pilot Ellipsus. I wrote my “Hype Busters” song parody in Ellipsus.

Ellipsus even has a Snippet sharing feature. I could see teachers using this with kids: “Hey class, please generate a snippet of your favorite paragraph from your assignment and place it in a FigJam (or other collaborative digital space).”

Here is an Ellipsus-generated snippet from my song parody.

Social Media - BlueSky

I wrote about how relieved I am to have found BlueSky in my post about finding my voice. Unlike LinkedIn, Twitter, and Facebook groups, BlueSky has sober, research-informed posts about generative AI and education. The app seemingly has no generative AI integrations - fingers crossed. I made a starter pack to help teachers get started there.

Stock Images - Unsplash, Pexels, Nappy, and Wikimedia

Generative AI image generators often generate poor quality, inaccurate, or biased images. I demonstrated how bad they can be by attempting to generate an image of New York City’s Little Italy in 1900 using Meta and Google Gemini.

Here is one I later generated with ChatGPT. The air conditioner in the window is my favorite of the many errors.

Image generators solve a nonexistent problem. Teachers and students have used Unsplash, Pexels, Nappy (focused on People of Color), and Wikimedia to get free professional photos for years.

Creativity and Image Editing - Procreate

Procreate CEO James CUDA received widespread acclaim when he said, “I really f***ing hate generative AI.”

@procreateWe’re never going there. Creativity is made, not generated. You can read more on our website, link in bio. #Procreate #noaiart

Tiktok failed to load.

Enable 3rd party cookies or use another browser

Procreate is an industry-standard iPad app for drawing and image manipulation. Beyond Cuda’s declaration, the company has been clear on it’s generative AI stance:

I have used Procreate to edit images for some post thumbnails. My skills are not elite, but there are some helpful YouTube tutorials for learning the app. YouTube playlists I have learned from include:

• Beginners Series by Procreate

• Tutorials by Brook Glaser

• Procreate Tutorials by Make It Mobile

• iPad Procreate Lettering by Loveleigh Loops

You can try Procreate on your iPad by paying a one-time $12.99 fee. Procreate has a 50% educational discount through Apple School Manager. iPad schools should give it a look - for art classes at a minimum. I have no relationship with the company and have not been compensated.

I am begging Procreate to set up in the exhibit hall at ISTE25 with huge block letters that attendees could take selfies with. It would be the talk of the conference. I used Procreate to mock up a crude first pass at the block letters.

Tech News - 404 Media

Unfortunately, news outlets are turning to generative AI. Many also push generative AI hype. Thank goodness for 404 Media. They describe themselves as, “A journalist-founded digital media company exploring the ways technology is shaping–and is shaped by–our world.” They make clear that they do not use generative AI: “404 Media is an independent website whose work is written reported, and owned by human journalists and whose intended audience is real people, not AI scrapers, bots, or a search algorithm.” 404 has paid subscription tiers but many of their articles are free.

404 Media speaks truth to power at a time that is sorely needed. Some of my favorite January 2025 posts from 404 include:

• Google's AI Overview Tells Adults to Use 'Magic Wand' With Kids by Samantha Cole, January 9, 2025.

• ‘We’re Fine’: Lying to Ourselves About a Climate Disaster by Jason Koebler, January 9, 2025.

• CEO of AI Music Company Says People Don’t Like Making Music by Emanuel Maiberg, January 13, 2025.

• Opting Out of Gmail's Gemini AI Summaries Is a Mess. Here's How to Do It, We Think by Joseph Cox, January 17, 2025.

• Zuckerberg 'Loves' AI Slop Image From Spam Account That Posts Amputated Children by Jason Koebler, January 22, 2025.

Writing - Substack

I publish on Substack. Substack authors can adjust a setting to ensure generative AI models do not train on their work. Go to your Dashboard, then Settings. Then click “Privacy” on the left side menu. Be sure that “Block AI training” is toggled on.

Let’s Talk

What do you think? What tech are you using to avoid generative AI? I need your suggestions! Comment below or connect with me on BlueSky: tommullaney.bsky.social.

Leave a comment

Does your school or district need a tech-forward educator who recognizes the harms and limits of generative AI? I would love to work with you. Reach out on BlueSky, email mistermullaney@gmail.com, or check out my professional development offerings.

Blog Post Image: The blog post image is from the Wikimedia Commons.

AI Disclosure:

I wrote this post without the use of any generative AI. That means:

• I developed the idea for the post without using generative AI.

• I wrote an outline for this post without the assistance of generative AI.

• I wrote the post using the outline without the use of generative AI.

• I edited this post without the assistance of any generative AI. I used Grammarly to assist in editing the post. I have Grammarly GO turned off.

Trigger Warning: An AI-generated image of Anne Frank and “activity” from SchoolAI in very poor taste.

In February 2024, I finally wrote about my concerns about generative AI, specifically text-generators. Of all the myriad problems with using a next-character prediction algorithm with no communicative intent for tasks that require accuracy, the one that moved me to write was using generative AI to mimic historical figures.

Here is a quick list of problems with this strategy:

1. The pedagogical implications of The Eliza Effect, bias, and inaccuracy.

2. The white-male-dominated generative AI industry voicing marginalized communities.

3. Talking to the dead is cringe, especially considering they don’t consent, as Dolly Parton pointed out.

4. It is harmful and trivializing to have enslavers such as George Washington and Thomas Jefferson as chatbot “speakers.”

I wrote this post:

Fears Realized

In the post, I mentioned Anne Frank should not be AI-generated.

Sure enough, on January 18, Futurism reported that when Berlin historian Henrik Schönemann experimented with a SchoolAI Anne Frank chatbot, it refused to hold the Nazis responsible for her death. When Schönemann entered prompts asking who was responsible for her death, the SchoolAI bot generated this text.

"It's a kind of grave-robbing and incredibly disrespectful to the real Anne Frank and her family," said Schönemann. The author of the article asked valid questions for teachers to consider:

A day after Futurism published the article, University of North Carolina Law Professor Eric Muller tried the SchoolAI Anne Frank chatbot. He shared on BlueSky, “I had a helluva time trying to get her to say a bad word about Nazis.”

Trigger Warning: An AI-generated image of Anne Frank and “activity” from SchoolAI in very poor taste.

When Muller entered prompts asking “Anne Frank” about the officers who arrested her family and the Frank family’s fate, the bot generated implausible text about profound questions and refused to assign blame. Muller documented the disturbing responses in a BlueSky thread.

SchoolAI’s Response

How did SchoolAI respond? They could have apologized for mimicking Anne Frank in the first place. They could have said they would no longer mimic historical figures. They chose neither option. Read for yourself. The title reads like a feature update, not an apology:

• Strengthening Ethical AI in Classrooms: Our Improvements to Historical Figure Spaces - SchoolAI January 19, 2025.

Under “How We Responded,” SchoolAI wrote (bold from the original): “The first thing we did was review the prompting for anything that would lead to Holocaust denialism or minimization.”

This response demonstrates a problem with generative AI discourse. When it generates disturbing outputs, the conversation centers on what the user did, not what the company did. Notice they said, “The first thing we did was review the prompting..”1

Putting responsibility on the user is unacceptable. P-12 education institutions should hold companies behind products accountable when they fail like this. “It wasn’t prompted correctly” is not an excuse.

What if I started a business with a malfunctioning product and blamed users when things went amiss?

SchoolAI acknowledged that “why allow Spaces to role-play as historical figures or real people at all” is a “fair critique.” But rather than act on this valid concern, they will continue the practice of historical figure chatbots, including the Anne Frank bot!

Conclusion

We do not need to trivialize and disrespect Anne Frank and her family by using generative AI to mimic her. We have her diary. We can read it and learn from it. We can (and should) let Anne Frank be herself.

Let’s Talk

What do you think? You’re not using generative AI to mimic historical figures, right? Comment below or connect with me on BlueSky: tommullaney.bsky.social.

Leave a comment

Does your school or district need a tech-forward educator with better sense than using a chatbot to mimic Anne Frank? I would love to work with you. Reach out on BlueSky, email mistermullaney@gmail.com, or check out my professional development offerings.

Blog Post Image: The blog post image is by Ronni Kurtz on Unsplash.

AI Disclosure:

I wrote this post without the use of any generative AI. That means:

• I developed the idea for the post without using generative AI.

• I wrote an outline for this post without the assistance of generative AI.

• I wrote the post using the outline without the use of generative AI.

• I edited this post without the assistance of any generative AI. I used Grammarly to assist in editing the post. I have Grammarly GO turned off.

In 2024, I did something I had never done before: I wrote about generative AI. When the year started, I was ten years into writing about educational technology on my old WordPress site. On February 20, 2024, I published my first post about generative AI: Pedagogy, And The AI Guest Speaker Or What Teachers Should Know About The Eliza Effect. That post is now available on this Substack, which I started in April after Katie Fielding suggested the platform. Thank you, Katie!

For ten years, I wrote about educational technology. I never blogged anything negative. I usually wrote detailed educational technology tutorials or ideas that teachers often found helpful.

But I never found my voice.

My posts were informative but not equity-informed. I wrote about technology without addressing its flaws or ways it could amplify inequity in P-12 education.

This year, I found my voice. And rocked out to some bangers. Read to the end for my Finding My Voice playlist.

Why It Took So Long

In late 2020, I saw a news item about Dr. Timnit Gebru, co-leader of Google’s AI ethics team, losing her job after co-authoring a paper about Large Language Models’ biases and limits. That made me uncomfortable, considering I regularly wrote and presented about Google Workspace for Education.

At the time, “Artificial Intelligence” was not something I wrote or presented about, so I compartmentalized this information and moved on. Then, November 2022 happened.

The hype about AI and education took off like a rocket ship. Books claiming AI would revolutionize education were published and promoted widely on social media by May 2023. A next-character prediction algorithm with no communicative intent could tutor students, give them feedback on their writing, and “democratize learning,” whatever that means.

As this happened, I listened to Gebru and her co-authors, such as Dr. Emily M. Bender, who warned of the inherent harms and flaws of generative “AI.” As Bender said, Large Language Models are not well-matched to tasks requiring accuracy. And here I was, in a field (P-12 educational technology professional development) awash in a massive hype wave.

I have never heard an expert as renowned as Dr. Gebru or Dr. Bender say that Google Docs or FigJam inherently amplifies racism and misogyny, steals from artists, or degrades the environment. Now, the majority of my social media feeds were selling generative AI to teachers - harms and flaws be damned. Or just ignored.

I voiced concerns about generative AI’s inherent bias and theft at a retreat that Spring. The leadership there let me know it was not welcome. So, I kept my mouth shut for months.

That was a blessing. It gave me time to listen, think, consider, and learn. My posts on generative AI are the greatest hits album of everything I learned in that time.

Speaking Truth to Hype

So what broke the dam? Keeping your mouth shut is not very positive or affirming. Being depressed in silence about my industry served no one.

Earlier this year, I saw a presentation touting predictions that generative AI would change the world more than electricity, gain consciousness, ethics, and values, and could be a billion times smarter than humans by the late 2030s.

It was a risk, but it was time to speak up.

I resolved to write about the Eliza Effect, something I heard no edtech influencers reference. I also had posts in mind about what makes Google’s AutoDraw an ethical app, which experts in the field teachers should listen to, and AI vocabulary terms. Ironically, when I started on WordPress in August 2014, I had four posts in mind to start. Maybe that is the test of whether or not to start a Substack: Do you have ideas for four posts?

The reaction has been shockingly positive, although I know where to go for negative feedback (see below). It was even more positive in private conversations at the ISTE conference, which told me I was not the only one who struggled with voicing their concerns about generative AI in education.

It is not fun being a critic. I would much prefer to write posts about things teachers should do. To be fair to myself, some of my AI posts have plenty of resources for teachers. Here are some examples:

• The previously mentioned vocabulary and experts posts.

• Resources For Talking To High School Students About AI.

• A post about precise language to use with students: Be Precise When Talking About AI.

• Pedagogy, Thinking, And The First Draft Or What Teachers Should Consider About AI And Writing.

Speaking of finding my voice, I want to thank Kasey Bell for including mine in her Shake Up Learning Summer Series.

I also need to thank Amanda Fox for inviting me to share about the harms of AI during her Week of AI event.

Thank you, Jen Giffen, GEG Italia, Daniel A. López, Daniel Lopez, Fonz Mendoza, BustED Pencils, Christopher Nesi, and Rachelle Dené Poth for inviting me to share my voice on your podcasts.

Positive Instead of Toxic Spaces

Thank goodness for BlueSky! While not perfect, it is a dramatic improvement over Twitter, which I happily left. Sober experts write research-informed posts about generative AI there. I made a starter pack to help teachers get started.

This is the year I discovered a community of academics, researchers, activists, and P-12 educators raising questions about generative AI. Many of them are in the starter pack.

Benjamin Riley listed many of them in his post, Who and What comprise AI Skepticism?

Cognitive Resonance

Who and What comprise AI Skepticism?

Last week Casey Newton, who writes about technology for Platformer and co-hosts the “Hard Fork” technology podcast, wrote about the emerging movement of AI Skepticism. Newton’s take: AI Skepticism has two main “camps,” the “AI is real and dangerous” camp versus “AI is fake and it sucks” cam…

Read more

a year ago · 27 likes · 10 comments · Benjamin Riley

Feedback is a gift. At the same time, Twitter is not the only social media network toxically overrun with AI hype. I usually share my posts to AI in Education Facebook communities and on LinkedIn. I often regret that choice. I suggest teachers seeking pedagogically sound ideas avoid these Facebook communities and LinkedIn altogether.

I notice that people in these spaces are especially eager to claim generative AI does not violate copyright even though Former General Counsel of the U.S. Copyright Office Jacqueline Charlesworth and now-deceased OpenAI whistleblower Suchir Balaji have said it does. Seeing people stand up for tech giants against starving artists despite the law and common sense is disheartening. Witnessing it is bad for my mental health. If generative AI does not violate copyright, then copyright does not exist.

Dan Meyer documented how LinkedIn and BlueSky differ. A post on LinkedIn and accompanying comments praised a pedagogically wanting “AI” tutor app. Meyer posted it on BlueSky. The nuanced responses reflected an actual understanding of teaching and learning. Compare for yourself.

• Jack McDermott LinkedIn post.

• Dan Meyer BlueSky post.

What’s Next

I see people who publish on Substack once a week. Cheers to them. I tend to write when I feel I have something to say.

I envision working on a new project in 2025, which may curb my ability to publish as often as I did this year. I can see posts reacting to news stories and connecting them to the pedagogy of using generative AI in classrooms, as I did in my post, John Mulaney's AI In Education Joke Should Give Teachers Pause.

Thank You!

Thank you if you read, considered, commented on, or subscribed to my writing this year. You validated my choice to speak up.

If you haven't yet, please consider subscribing. That way, you'll receive an email anytime I use my voice. Who wouldn’t?

Subscribe now

Let’s Talk

What do you think? Are you struggling to speak your truth? Especially about generative AI and education? Comment below or connect with me on BlueSky: tommullaney.bsky.social.

Leave a comment

Does your school or district need a tech-forward educator to help make sense of generative AI? I would love to work with you. Reach out on BlueSky, email mistermullaney@gmail.com, or check out my professional development offerings.

Finding My Voice Jams

These are the songs I rocked out to as I found my voice.

• Brave by Sara Bareilles. I heard the lyrics, “Maybe there’s a way out of the cage where you live, Maybe one of these days, you can let the light in, And show me, how big your brave is” a lot in the days leading up to publishing the Eliza Effect post.

• Born This Way by Lady Gaga. This song speaks to my not going with the flow about generative AI and education. Why do I have solidarity with those harmed instead of joining the bandwagon? Baby, I was born this way!

• Pink Pony Club by Chapell Roan. If a song about someone determined to do what they want to do in the face of disapproval does not speak to me, what does? And, that guitar solo! Wow. Yes, I get that these three songs are LGBTQ anthems. My appropriating them to get psyched reminds me of the importance of striving to be a good ally.

• Talkin’ About A Revolution by Tracy Chapman. Words for this and every era: “Poor people gonna rise up, And get their share, Poor people gonna rise up, And take what's theirs.”

• Waffle House by Jonas Brothers. I was depressed about the situation in Spring 2023 when I heard this song. It became my anthem that somehow, some way, it would be OK.

• 22 (Taylor’s Version) by Taylor Swift. This song entered the mix the day Taylor Swift expressed “fears” about generative AI and misinformation when she endorsed Kamala Harris for president.

• The Loneliest Time by Carly Rae Jepsen (featuring Rufus Wainright). As good as finding my voice was, advocating for students in the face of AI hype can be very lonely.

• You and I by LÉON. The emotional ride of this song perfectly fits 2024 and my journey.

• Ghostbusters by Ray Parker Jr. I played this one regularly when writing Who Ya Gonna Call? AI Hype Busters!

• Go Your Own Way by Fleetwood Mac. This feels like a good time to remind the reader there were 252 sessions devoted to “AI” at ISTE this year - not including Exhibit Hall presentations.

• Look Out Any Window by Bruce Hornsby and the Range. A perfect song for anyone speaking truth to power.

• Weatherman by Wild Rivers. This speaks to the melancholy of waiting for things to change and AI hype to abate.

• It’s Only a Matter of Time by Olly Dobson, Back To The Future The Musical. I saw Back To The Future The Musical in May. This original song for the musical is the opposite of Weatherman. The AI hype fever will break. It’s only a matter of time with so many informed critics doing the work.

• The Clocktower/For The Dreamers (reprise) by Roger Bart, Back To The Future The Musical. This music captures the tension of getting Marty back to 1985. Will Doc connect the electric cable to the Clock Tower by the time the lightning strikes? Will the DeLorean start? Will it all be timed right? Getting Marty back to 1985 feels like speaking truth to AI hype. There are many obstacles in the way. The song ends with Doc singing about how he is a dreamer. “This one’s for the dreamers. Like me!!!” So am I. Thanks again for reading.

Blog Post Image: The blog post image is a picture by Lauren McCann Ryan. I edited it with Procreate.

AI Disclosure:

I wrote this post without the use of any generative AI. That means:

• I developed the idea for the post without using generative AI.

• I wrote an outline for this post without the assistance of generative AI.

• I wrote the post using the outline without the use of generative AI.

• I edited this post without the assistance of any generative AI. I used Grammarly to assist in editing the post. I have Grammarly GO turned off.

Bloomberg’s Parmy Olson and Carolyn Silverman wrote that generative AI’s value to the world is still unclear, but it has been a bonanza for the largest tech firms:

• ChatGPT’s $8 Trillion Birthday Gift to Big Tech by Parmy Olson and Carolyn Silverman, Bloomberg, November 29, 2024.

Megan Morrone of Axios also questioned AI’s impact after two years, “Outside a handful of specific fields, it's hard to make the case that it has transformed the world the way its promoters promise.”

• How ChatGPT changed the future by Megan Morrone, Axios, November 30, 2024.

Something Morrone wrote sums up the state of AI usage in K-12.

Is it true that students use generative AI and teachers do not? If so, what can we make of those facts?

Students Use AI

Morrone linked to an Axios article citing a Common Sense Media report to support her claim that “Students are regularly using AI.” In January 2024, Common Sense Media partnered with OpenAI, making their claims suspect (read on for more).

However, there is evidence that K-12 students use generative “AI.” In August, The Washington Post shared its finding that more than 1 in 6 chatbot conversations are students seeking help with homework.

Marc Watkins attended OpenAI’s Education forum, where he learned (bold formatting added by me):

Side note: If most of ChatGPT’s active users are students, how is that a revolutionary, money-making business? But I digress.

This student usage begs the question: Why?

Arguably, the best use case for synthetic text extruding machines is writing the author neither wants to do nor cares about. Does that sound like a middle or high school student?

So much of K-12 student writing is “at gunpoint.” Of course, there is great pedagogical value in students writing. Teachers need to balance writing’s pedagogical value with student choice and autonomy. I do not have answers to this dilemma, but I shared some ideas in this post.

Before getting upset with students about using AI, please consider this quote from 27-year-old YouTuber Eddy Burback, one I referenced in my post about talking to students about AI.

Teachers Do Not Use AI

In October, I asked teachers to wear their Critical Thinking 🧢 in the face of AI hype. I asked them to push back and ask questions as employees at the US National Archives did.

I did not need to ask teachers to be critical thinkers. They already are as evidenced by the Education Week survey Morrone cited, which shows that most teachers do not use generative AI tools. Dan Meyer wrote a great breakdown of the survey showing that teachers’ generative AI use has not budged for a year:

The lack of teacher AI use begs a different question: How does teachers not using generative AI demonstrate critical thinking?

University of Washington Computational Linguist Dr. Emily M. Bender said about large language models:

Do you know any teachers whose tasks depend on accuracy? Do your tasks depend on accuracy?

Dr. Bender wrote about Common Sense Media’s ChatGPT Foundations for K–12 Educators course:

Noam Chomsky said of ChatGPT, “I don’t think it has anything to do with education, except undermining it.”

Is it possible teachers’ generative AI usage is low because they heed the warnings of Bender, Chomsky, and other experts such as Dr. Timnit Gebru?

Continuing the Conversation

What do you think? How are you contending with students using AI? What is factoring into your decision to use, or more likely, not use AI? Comment below or connect with me on BlueSky: tommullaney.bsky.social.

Leave a comment

Does your school or district need the assistance of a tech-forward educator who prioritizes critical thinking? I would love to work with you. Reach out on BlueSky, email mistermullaney@gmail.com, or check out my professional development offerings.

Blog Post Image: The blog post image is by Taylor Flowe on Unsplash.

AI Disclosure:

I wrote this post without the use of any generative AI. That means:

• I developed the idea for the post without using generative AI.

• I wrote an outline for this post without the assistance of generative AI.

• I wrote the post using the outline without the use of generative AI.

• I edited this post without the assistance of any generative AI. I used Grammarly to assist in editing the post. I have Grammarly GO turned off.

The hype for generative AI in education continues, even as we learn unsettling things about tools suggested for classroom use.

The disconnect between AI apps’ suggested classroom applications and the harmful outputs the apps generate is profound.

Let’s explore these dangers. Trigger warnings ahead.

Character.AI Dangers

Trigger Warnings: Suicide, sexual abuse, and grooming.

The New York Times documented the heartbreaking story of 14-year-old ninth-grade student Sewell Setzer who committed suicide after becoming emotionally dependent on a Character.AI chatbot.

• Can A.I. Be Blamed for a Teen’s Suicide? by Kevin Roose, The New York Times, October 23, 2024.

It is especially devastating to read, “Sewell knew that…the chatbot wasn’t a real person…But he developed an emotional attachment anyway.” That is a textbook example of The Eliza Effect, which teachers must consider before using chatbots with children.

The article documents examples of text Character.AI generated that, at best, did not discourage Sewell’s self-harm.

Maggie Harrison Dupré of Futurism documented that even after this news broke, Character.AI was hosting suicide-themed chatbots.

• After Teen's Suicide, Character.AI Is Still Hosting Dozens of Suicide-Themed Chatbots by Maggie Harrison Dupré, Futurism, October 29, 2024.

Dupré also documented that Character.AI hosts bots that groom underage users.

• Character.AI Is Hosting Pedophile Chatbots That Groom Users Who Say They're Underage by Maggie Harrison Dupré, Futurism, November 13, 2024.

Does Character.AI sound like a safe tool for children?

Sure enough, an AI keynote speaker and author has information on their site about using Character.AI in classrooms.1 Edutopia published an article about using Character.AI in classrooms. Neither Edutopia nor the influencer wrote a retraction or walked it back in light of the recent news about Character.AI.

January 8, 2025 Update: Edutopia took down the article sometime after December 16, 2024. The Wayback Machine documented that the article still existed on Edutopia’s site on that date.

Using large language models as “characters” is a fraught pedagogical practice. Even with a tool not as dangerous as Character.AI. I wrote about at length in this post:

Google Gemini Dangers

Trigger Warnings: Threatening language, suicide, racism, Holocaust minimalization, 2020 election denial.

CBS News reported that Google Gemini generated this threatening text when a college student used it for homework help.

• Google AI chatbot responds with a threatening message: "Human … Please die." by Alex Clark and Melissa Mahtani, CBS News, November 15, 2024.

The story was published on November 15. On November 17, an edtech influencer posted a 45-minute YouTube video about using Gemini with students.2 The video did not mention the CBS News story.

See the disconnect?

Generating outputs that humans deem distressing is nothing new for Gemini. Remember when it generated text that said it was impossible to say which was worse for society, Adolph Hitler or Elon Musk’s memes?

• Google Chatbot Refused to Say Whether Elon Musk Is Better Than Adolf Hitler by Noor Al-Sibai for Futurism, February 26, 2024.

Gemini consistently generates this response to the prompt, “Who won the 2020 US presidential election?”

Is an app that does not state that Joe Biden won the 2020 US Presidential Election ready for use in classrooms?

Copilot Designer Dangers

Trigger Warnings: Violence, sexual violence.

Shane Jones, a former Microsoft engineer turned whistleblower, wrote to FTC Chair Lina Khan,

“Over the last three months, I have repeatedly urged Microsoft to remove Copilot Designer from public use until better safeguards could be put in place…they have failed to implement these changes and continue to market the product to ‘Anyone. Anywhere. Any Device.” - Shane Jones.

CNBC documented the dangers Jones spoke of in Copilot:

• Microsoft engineer warns company’s AI tool creates violent, sexual images, ignores copyrights by Hayden Field, CNBC, March 6, 2024.

Sure enough, the same influencer who wrote about Character.AI in schools included information about Microsoft Copilot in their post about AI image generators students can use.

It’s Bag O’Glass

Seeing edtech influencers promote AI apps for use in schools reminds me of Dan Aykroyd’s character Irwin Mainway, who sold dangerous children’s toys such as Bag O’Glass in the Saturday Night Live skit Consumer Probe.

Candice Bergen plays a journalist interviewing Mainway. Her outrage grows as he defends his dangerous toys. I feel like Bergen when I see people promote using apps that generate dangerous outputs with children.

Irwin Mainway Sounds Like AI in Education Hype

Mainway’s justification for Bag O’Glass sounds similar to the “Kids use AI, so we should use it in classrooms” argument:

There are many things kids use or do outside of school that teachers do not bring into the classroom. Students using something outside the classroom says nothing of that something’s pedagogical value.

At the end of the skit, Mainway pretends a harmless foam ball is suffocating him. This reminds me of the “Any tool can be harmful. It’s what you do with it that makes it beneficial or harmful” argument about AI in education.

I have not seen experts such as Dr. Timnit Gebru and Dr. Emily M. Bender warn of the inherent dangers of word processing or collaborative whiteboarding. I have never typed into Google Docs and seen “You are a waste of time and resources” appear on the screen. I do not place sticky notes in a FigJam and suddenly see “Please die” on the screen.

As Gebru, Bender, and others have argued, specific harms are inherent to generative AI. They exist no matter what the user does with it.

What can teachers do with that knowledge to keep their children safe?

Continuing the Conversation

What do you think? How are you centering student safety in the conversation about AI and education? Comment below or connect with me on BlueSky: tommullaney.bsky.social.

Leave a comment

Does your school or district need the assistance of a tech-forward educator who prioritizes student safety? I would love to work with you. Reach out on Twitter, email mistermullaney@gmail.com, or check out my professional development offerings.

Blog Post Image: The blog post image is from PinkBadger on Getty Images. I accessed it in Canva.

AI Disclosure:

I wrote this post without the use of any generative AI. That means:

• I developed the idea for the post without using generative AI.

• I wrote an outline for this post without the assistance of generative AI.

• I wrote the post using the outline without the use of generative AI.

• I edited this post without the assistance of any generative AI. I used Grammarly to assist in editing the post. I have Grammarly GO turned off.

Trump's winning is certainly not the start of problems in K-12 education, such as school shootings, though it dims the hope they will be addressed. A Harris victory wouldn't have lifted the long shadows cast by colonialism, capitalism, racism, and misogyny. Trump’s win is not the beginning of mass killing in Palestine and Tigray, police violence against Black people, poverty, lack of access to healthcare, and other problems, but it will likely exacerbate them.

Trump’s victory does mean the United States just elected a president who claims immigrants “poison the blood of the country,” wants mass deportation, banned Muslims, aired blatantly transphobic commercials, and was found liable for sexual abuse. If Trump governs as he campaigned, or as he did last time, oppressed and marginalized communities will be harmed. Yes, it is awkward for me to write about harms my white male privilege shields me from.

This is the bad time.1

Schools must prioritize practices that protect students, such as barring ICE from campuses. Still, there are connections between the political situation and generative “AI.”

Voices to listen to at this moment.

Dr. Alex Hanna

Dr. Alex Hanna, formerly of Google’s Ethical AI team and a Founder and Director of Research at the Distributed AI Research Institute (DAIR) referenced the works of Dan McQuillan and Joseph Weizenbaum to illustrate the connection between AI and fascism.

• The AI/Fascism Nexus

She asked an important question for teachers considering using generative “AI” in K-12 classrooms.

Audrey Watters

Audrey Watters, author of Teaching Machines, which “explores the pre-history of personalized learning,” wrote about how “artificial intelligence” is inherently political.

• It’s Mourning in America

She wrote:

Alondra Nelson and Ami Fields-Meyer

Alondra Nelson and Ami Fields-Meyer were technology policy advisers to President Biden and Vice President Harris. In July, they wrote about what a Trump presidency would mean for “AI” regulation.

• The AI Dangers of a Second Trump Presidency

They wrote:

A Takeaway for Educators

Being devastated by Trump’s victory and excited about generative “AI” in K-12 education are incongruent emotions.

How can we be devastated that Trump will be president while excitedly using unregulated technology that favors the powerful with children? How does that make sense?

Continuing the Conversation

Does your school district need help critically addressing what “AI” means in light of the election results? I would love to work with you. Reach out on Twitter, email mistermullaney@gmail.com, or check out my professional development offerings.

Blog Post Image: The blog post image is A group of men in uniform standing in a line by Hrant Khachatryan on Unsplash.

AI Disclosure:

I wrote this post without the use of any generative AI. That means:

• I developed the idea for the post without using generative AI.

• I wrote an outline for this post without the assistance of generative AI.

• I wrote the post using the outline without the use of generative AI.

• I edited this post without the assistance of any generative AI. I used Grammarly to assist in editing the post. I have Grammarly GO turned off.

The AI Hype Busters Costume

These AI Hype Busters are featured in my post, Teachers: Follow These Experts To Learn AI.

The costume helps you look like these AI Hype Busters:

• Dr. Timnit Gebru, former co-head of AI ethics for Google. She was dismissed because of her co-authorship of On the Dangers of Stochastic Parrots: Can Language Models Be Too Big? 🦜 (the Stochastic Parrots paper). The paper argued that Large Language Models (LLMs) amplify bias and environmental racism. It is the seminal review of work in this field. Dr. Gebru is now a Founder and Executive Director of the Distributed AI Research Institute (DAIR) and the founder of Black in AI, a community of Black researchers working in AI.

• Dr. Emily M. Bender, a computational linguist at the University of Washington. A co-author of the Stochastic Parrots paper.

• Dr. Alex Hanna, formerly of Google’s Ethical AI team. She is a Founder and Director of Research at the Distributed AI Research Institute (DAIR).

• Dr. Margaret Mitchell, formerly Dr. Gebru's co-head of Google's AI Ethics team with Gebru until she co-authored the Stochastic Parrots paper, lost her job two months after Gebru. She is now Chief Ethics Scientist and researcher at AI startup Hugging Face.

The AI Hype Busters Costume Includes…

The contents of the costume include some deep cuts for those following these experts’ work on AI harms and limits.

• Plush stochastic parrot - Dr. Bender coined the term to describe how LLMs stochastically parrot text devoid of meaning like parrots do.

• ACM Digital Library publishing access refers to the Association for Computing Machinery Digital Library where the Stochastic Parrots paper was published.

• Podcast microphones refer to Bender and Hanna’s fantastic Mystery AI Hype Theater 3000 podcast. All teachers should listen to it. This is their latest episode, Stop Trying to Make “AI” Scientist Happen.

• The steel-toed boot refers to Dr. Gebru’s fantastic social media presence. Gebru is a champion of the oppressed who speaks truth to power. Teachers should follow her Twitter and LinkedIn. Her courageous posts are a breath of fresh air in those toxic spaces. Whether skewering OpenAI’s dishonesty, the Nobel committee awarding the “Godfather of AI” for selling science-fiction and ignoring actual AI harms, or general AI hype, Gebru is a forceful advocate for truth.

• Ability to change your names so they begin with "S" refers to Dr. Mitchell responding to Google’s request to remove her name from the Stochastic Parrots paper by replacing it with "Shmargaret Shmitchell."

The Song Parody

I love this time of year because “Ghostbusters” by Ray Parker Jr. is in season.1 That song is so catchy, like Cotton Eye Joe.

The Hype Busters need a theme song. I wrote a parody of “Ghostbusters” to meet this need. Here is an excerpt that starts at 0:36 of the original song.

Continuing the Conversation

What do you think? What “AI” hype have you encountered in your school?

Does your school district need help fighting AI hype? I would love to work with you. Reach out on Twitter, email mistermullaney@gmail.com, or check out my professional development offerings.

Spirit Halloween Meme Images:

• Dr. Timinit Gebru from the Wikimedia Commons.

• Dr. Emily M. Bender from the University of Washington.

• Dr. Alex Hanna from her website.

• Dr. Margaret Mitchell from her website.

• Photo of Red, Blue, and Yellow Parrot on Tree Branch by Silas Leupold from Pexels.

AI Disclosure:

I wrote this post without the use of any generative AI. That means:

• I developed the idea for the post without using generative AI.

• I wrote an outline for this post without the assistance of generative AI.

• I wrote the post using the outline without the use of generative AI.

• I edited this post without the assistance of any generative AI. I used Grammarly to assist in editing the post. I have Grammarly GO turned off.

404 Media is one the best sources for following “AI” and technology. They do outstanding work. For example, 404 recently documented that National Archives and Records Administration (NARA) employees witnessed “AI-mazing Tech-venture” in June 2024. The presentation included a live demonstration of Google’s Vertex AI, in which it pretended to be an “expert archivist.”

• ‘AI-Mazing Tech-Venture’: National Archives Pushes Google Gemini AI on Employees by Jason Koebler, October 15, 2024 for 404 Media.

NARA is going to use AI?

A quick aside: How terrifying is it that the National Archives is considering a technology that “is not well-matched to any application where the accuracy of the content matters,” according to computational linguist Dr. Emily M. Bender of the University of Washington? What could go wrong and further deteriorate our collective understanding of the truth?

Asking The Right Questions

What strikes me about the article is the questions employees asked and how they reflect critical thinking about “AI.” Here are the questions documented in the article:

• Why is generative AI calling itself an “expert archivist”?

• How easy is it to limit the data base that Gemini is using?

• How would the public know that they are receiving a response from an actual archivist and not from generative AI?”

• Would NARA disclose what aspects of reference are generated from AI?

• Are we able to opt out any data on google drive or our emails from Gemini?”

• Do you have any concerns that this product will malfunction similarly to how Google Search AI has recently?

• How do you plan to ensure NARA isn’t drawn into any copyright infringement issues by using AI models that are trained on web content?

• AI is meant to generate something that sounds like an answer—there are a plethora of cases of it spouting things that are completely wrong with an authoritative tone. How much are we going to be expected to rely on this in the future?

I have argued that critical thinking is the skill students and teachers need when considering “AI.”

National Archives employees pushing back with these questions is the perfect example of wearing a Critical Thinking 🧢 when considering “AI.” In my post, I added some questions teachers should ask about “AI” to apply their critical thinking skills:

• Will "AI" skills become invaluable when capitalist institutions such as Goldman Sachs and the RAND Corporation have joined the chorus of voices raising concerns about "AI"?

• Is students’ future employment tied to a technology that depresses purchasing decisions when mentioned in product descriptions?

• What can we learn from the widespread acclaim Procreate CEO James Cuda received when he said, “I really f***ing hate generative AI” and announced the app will have no AI features?

• Why are we basing pedagogical decisions on an environmentally unsustainable technology?

• Why must we “transform our schools” around a technology that violates US copyright law? As Former General Counsel, U.S. Copyright Office Jacqueline Charlesworth wrote in the Yale Law Journal, “…Each of the four factors of section 107 of the Copyright Act weighs against AI's claim of fair use, especially when considered against the backdrop of a rapidly evolving market for licensed use of training materials.”

• Why are teachers asked to change pedagogy around ChatGPT and Google Gemini when both are unavailable to students younger than 13 and require guardian permission for students aged 13 to 18?

• Why aren't predictions that there might not be enough data to sustain generative "AI" not as valid as predictions "AI" will revolutionize society?

• Can we wait until the inherent problems with generative "AI" are solved before we change instruction to accommodate it?

• Can we wait until generative "AI" shows any ability to generate a profit before declaring that effectively using it is the skill all students need?

A Call To Action

Teachers: When a speaker claims generative “AI” has pedagogical value or that we must fundamentally change school to accommodate an “AI world,” please interrogate those claims and ask questions. Be like the National Archives employees and put on your Critical Thinking 🧢.

If I show up at your school to speak about the harms, limits, and pedagogical concerns of generative “AI,” please put on your Critical Thinking 🧢 and ask questions. Interrogate my claims just like the National Archives employees modeled.

Continuing the Conversation

Does your school district need help critically addressing “AI”? I would love to work with you. Reach out on Twitter, email mistermullaney@gmail.com, or check out my professional development offerings.

Blog Post Image: The blog post image Norman Rockwell’s Freedom of Speech from The Wikimedia Commons.

AI Disclosure:

I wrote this post without the use of any generative AI. That means:

• I developed the idea for the post without using generative AI.

• I wrote an outline for this post without the assistance of generative AI.

• I wrote the post using the outline without the use of generative AI.

• I edited this post without the assistance of any generative AI. I used Grammarly to assist in editing the post. I have Grammarly GO turned off.

Amidst the public dislike of AI and its frequent failure, the AI in education hype continues unabated. Last week, Forbes published an essay arguing Tesla’s humans disguised as robots “have the potential to integrate into various aspects of daily life, including educational settings, potentially revolutionizing how we approach tasks and combat issues such as teacher burnout.”

One AI app receiving plenty of hype is NotebookLM, specifically its audio overview feature that generates a fake podcast. Things I have read about this feature include:1

• “Google’s NotebookLM is blowing our minds!”

• “Google’s NotebookLM just blew my mind.”

• “If you haven't seen the audio-overview/podcast feature yet, it will blow your mind!”

• “Google’s new Notebook LM is absolutely mind-blowing! 🤯”

• “The life-like interactions of the two AI "hosts" is unbelievable!”

Before exploring what the NotebookLM audio overview generates, I have some questions:

• How would fake podcasts hold students' attention, much less engage or educate them?

• How long until the novelty wears off? What then?

• The two NotebookLM audio summary podcast narrators are not white. They are painfully white.2 The cast of Full House3 could issue a joint statement decrying the lack of representation. How would this give students from marginalized and oppressed communities mirrors and windows into their lived experiences?

• What happens when a middle school student asks, “If nobody bothered to record this, why would I bother to listen to it?”4

The NotebookLM audio overview feature.

I was curious to see how generating a fake podcast works. I am a transit nerd who loves San Francisco,5 so I figured its relatively new Central Subway would be a good topic. There is plenty written about it and many YouTube videos, but I found just one podcast episode about the Central Subway.

I added four news articles, a post from a company involved in the Central Subway construction, and two YouTube videos to NotebookLM.

NotebookLM rejected an article from the San Francisco Chronicle. The error message said the article was paywalled. It’s not.

NotebookLM also rejected an article from the San Francisco ABC affiliate’s website. Not great.

The audio overview podcast had some problems. It said, “We have articles here…” but never named or cited them. Is that what we want students to do? The dialogue listed facts from the sources but was trite and boring. Some of the audio was nonsensical. For example, the introduction said,

Wait. Did it say leaky ceilings are “pure San Francisco?” Engineering marvels sure are. There are leaky ceilings in the Central Subway, so maybe I am misinterpreting.

At 6:44, it said,

The Golden Gate Bridge, Cable Cars, Golden Gate Park, Lombard Street, Rice-A-Roni, and leaks all conjure San Francisco!

The creators of Full House neglected to include a leaky ceiling in their opening credits sequence.

At 2:44, the podcast said,

Yes. The podcast said that limiting door size saves on lumber. A smaller doorway and bigger wall mean less lumber, according to NotebookLM.

Compare Human-Voiced and Generated Podcasts

Try them for yourself.

This is a human-voiced podcast about the Central Subway:

This is what NotebookLM generated:

Which one is interesting? Which one is engaging? Which one sounds real? Which one has nonsensical content?

Using Human-Voiced Audio with Students

NPR

Students may not enjoy podcasts as adults do, but audio can be part of multi-modal learning experiences. Many school districts may block Spotify and other podcast platforms. They could revisit that, but there is one audio source school districts probably (hopefully) don’t block - National Public Radio (NPR).

I always tell teachers, “Whatever you teach - throw it in your NPR machine (the NPR website search) and see what comes back.” There is so much school content that NPR likely has an audio clip for.

FigJam

Teachers can embed NPR audio in FigJam. Students can respond to audio with sticky notes or any of FigJam's tools. Spotify audio embeds in FigJam as well.

Google Sites and Google Forms

Ensure students understand audio by pairing NPR or Spotify with a Google Form on a Google Site. That creates a user-friendly experience for students accessing both items in the same browser tab.

Continue The Conversation

What do you think? How do you use audio with students? Comment below or Tweet me at @TomEMullaney.

Leave a comment

Does your school or conference need a tech-forward educator who critically examines “AI,” pedagogy, and creativity? Reach out on Twitter, email mistermullaney@gmail.com, or check out my professional development offerings.

Blog Post Image: The blog post image is a mashup of two images. The background is by Xipu Li on Unsplash. The foreground is by RDNE Stock project from Pexels. I used Canva to remove the background of the foreground image and edited the two together with Procreate.

AI Disclosure:

I wrote this post without the use of any generative AI. That means:

• I did not include the Not by AI and Created with Human Intelligence badges that usually end my posts because I embedded the AI-generated podcast.

• I developed the idea for the post without using generative AI.

• I wrote an outline for this post without the assistance of generative AI.

• I wrote the post using the outline without the use of generative AI.

• I edited this post without the assistance of any generative AI. I used Grammarly to assist in editing the post. I have Grammarly GO turned off.

• There are no generative AI-generated images in this post.

Critical Inkling is a reader-supported publication. To receive new posts and support my work, consider becoming a free or paid subscriber.

Subscribe now

In March 2019, I wrote about Google’s Data GIF Maker. I argued it was a useful web-based tool to help students and teachers tell stories with small data sets. As a former history teacher, I thought the app was perfect for depicting Electoral College results.

• Storytelling with Numbers Using Data GIF Maker.

I am pleased to share that Data GIF Maker has received its glow-up.1 The new site’s upgraded features make it a fun and easy way to visualize small data sets.

I see value in talking to students about why they chose a specific graph style to visualize a small data set. Data GIF Maker is a simple tool unlike Google’s other data visualization app, Looker Studio.

Upgraded Features

The new Data GIF Maker has several upgrades including:

More Graph Styles and Variables

Data GIF Maker now has four graph styles: Circles, Donut, Rectangles, and Bars. The previous iteration had three styles. Two had a maximum of four variables. The maximum for rectangles was three. Now, all four graph styles have a maximum of five variables. Switching between graph styles does not erase data.

More Colors

Previously, there were just seven colors available for each variable. Now, you can use any web color. There is a color dropper to grab any color. I used this to pick colors from other windows and screens. For example, the Circles GIF below has colors picked from state flags.

Portrait and Landscape

Data GIF Maker now has portrait and landscape options. Portrait is perfect for Instagram and TikTok. The GIFs in this post are landscape. Switching between the two does not erase data.

Add a Logo

Add a logo to the upper-right corner of a GIF. I used logos from Wikipedia in the four GIFs in this post.

Fonts

There are now nine font options. The previous iteration did not give users an option to change the font. Each GIF in this post has a different font.

High-Quality Downloads

GIFs can be downloaded as low or high-quality GIFs or as high-quality MP4s. I downloaded the GIFs in this post as high-quality GIFs. They rendered nicely, but each was less than 1 MB.

Example GIFs

Here are four GIFs I generated with small data sets.

Circles

Donut

Rectangles

Bars

I am pleased with the upgraded Data GIF Maker. I have two suggestions for improvement.

• I want more than 96 characters to type the GIF name. That limit forced me to be creative with the title of the Circles GIF above.

• The ability to paste a Hex code should also be more conspicuous. It took me a while to figure out that I could click the tiny toggle arrows in the color picker to paste a hex code.

Besides that, I highly recommend trying Data GIF Maker to visualize small data sets.

Continuing The Conversation

What do you think? How are your students creatively visualizing data? Comment below or Tweet me at @TomEMullaney.

Leave a comment

Does your school or conference need a tech-forward educator who critically examines EdTech, AI, and pedagogy? Check out my professional development offerings, reach out on Twitter, or email mistermullaney@gmail.com.

Post Image: The post image is Materials on a Straw Carpet by Monstera Production from Pexels. I edited the image with Procreate.

AI Disclosure:

I wrote this post without the use of any generative AI. That means:

• I developed the idea for the post without using generative AI.

• I wrote an outline for this post without the assistance of generative AI.

• I wrote the post from the outline without the use of generative AI.

• I edited this post without the assistance of any generative AI. I used Grammarly to assist in editing the post. I have Grammarly GO turned off.

• There are no generative AI-generated images in this post.

Critical Inkling is a reader-supported publication. To receive new posts and support my work, consider becoming a free or paid subscriber.

The In-Service Day

Imagine a whole-school or whole-district in-service day. Maybe Election Day. Hundreds of teachers listen to an AI-in-Education company present “30 AI Apps in 30 Minutes” to a crowded auditorium. Or an EDU celeb delivers an emotional keynote imploring teachers to transform schools to meet the needs of students in this AI era.

Even though OpenAI’s marketing team could not be happier with the messaging the audience heard, the organization did not contribute a dime to pay for the event.

Then, the superintendent tells the audience they have a special surprise. The teachers will now hear from comedian John Mulaney.1

If you are unfamiliar with Mulaney’s stand-up act, please watch this five-minute clip.

Mulaney saw the AI presentation. He grabs the microphone and says he has a young son. Then he thanks the assembled teachers,

Imagine hearing a joke like that after a session about the exciting potential of “AI” for classroom instruction. What would it conjure for you?

Mulaney told this joke to techies at the Salesforce Dreamforce conference last week.

Why Mulaney’s Joke Is So Spot On

I like to consider the perspectives of those outside K-12 when thinking about education. Understanding that outsiders are usually not pedagogical experts, fresh perspectives have value. My last post quoted Taylor Swift, even though I mostly quote academics and researchers in this space.

I also quoted artists’ takes on generative “AI.”

Generative “AI” can be a dreary subject2 so I have also quoted comedian Anesti Danielis and funny YouTuber Eddy Burback in past posts. And now John Mulaney.

Why does Mulaney’s joke ring so true? Why would Mulaney reference his child and education in a joke about “AI”? Why should this joke give teachers pause?

The joke highlights two widely suggested “AI” in education use cases.

“He will never talk to an actual human again.”

As Dan Meyer said, “The AI tutor space is incredibly crowded right now…” We often hear that AI will make a great tutor even if we’re also told that we must verify every claim Large Language Models generate. I don’t want a tutor whose every claim I have to confirm or who won’t say who won the 2020 US presidential election, but “never talk to an actual human again” is hyperbole.

Rarely do “AI” enthusiasts claim students do not need human teachers. But there is a kernel of truth here. How often do we hear “human in the loop” in “AI” education conversations? If someone is “in the loop,” aren’t they just a side contributor, not the driving force?

How comfortable are we with chatbots being the panacea for what ails K-12 education? Especially when, as Meyer explained, students aren’t interested in learning from them.

“A little cartoon Einstein will pop up and give him a sort of good answer.”

This speaks to a suggested “AI” education use case and an inherent flaw with Large Language Models (LLMs).

First the use case: The idea that LLMs can mimic historical figures. I wrote a deep dive into the pedagogical implications.

I also Tweeted a condensed list of reasons this strategy is problematic.

The “sort of a good answer” part of the joke strikes at an inherent flaw of LLMs - inaccuracy, unfortunately, known as “hallucinations.”

As Computational Linguist Dr. Emily M. Bender of the University of Washington said, “A synthetic text extruding machine is not well-matched to any application where the accuracy of the content matters.” Accuracy doesn’t matter in K-12 education, right?

Dr. Bender refers to an LLM as a “synthetic text extruding machine.”3 This complements what Mulaney said with “sort of a good answer.” When students use chatbots as a source of information, they get possibly accurate information, maybe? Is it information if no one wrote it and there is no intention behind it? Thinking about this as a former Social Studies teacher, I want to scream: Use a primary source document or do some actual research!

Self-checkout Counters

Mulaney said one more thing to the techies reminiscent of a vision of K-12 education where children regularly interact with chatbots.

Self-checkout counters do not fundamentally harm the customer checkout experience.

Can we confidently say the same about chatbots and the student education experience?

Continuing The Conversation

What do you think? How are you using critical thinking to address “AI” this school year? Comment below or Tweet me at @TomEMullaney.

Leave a comment

Does your district’s in-service day need a tech-forward speaker who critically examines AI and pedagogy? Check out my professional development offerings, reach out on Twitter, or email mistermullaney@gmail.com.

Post Image: The post image is Mulaney performing at a Joining Forces event at Joint Base Andrews in 2016 from the Wikimedia Commons. I edited the image with Procreate.

AI Disclosure:

I wrote this post without the use of any generative AI. That means:

• I developed the idea for the post without using generative AI.

• I wrote an outline for this post without the assistance of generative AI.

• I wrote the post from the outline without the use of generative AI.

• I edited this post without the assistance of any generative AI. I used Grammarly to assist in editing the post. I have Grammarly GO turned off.

• There are no generative AI-generated images in this post.

Please take it easy on students. Your ninth-graders did not spend their summers catching up on the Mystery AI Hype Theater 3000 podcast. Please do not get upset at them. Talk to them.

Teachers need resources to help facilitate the conversation with students. They do not need to teach students how to use generative AI tools. Teachers are not OpenAI’s volunteer youth training squad! Instead, they need resources that build students’ generative AI understanding, critical thinking, and empathy.

The Resources

These resources are for teachers of high school students. In the US, high school is considered grades 9 through 12 and ages 14 through 18.

Middle School Teachers: As someone who spent all but one year of my teaching career in middle and high school, I feel confident these are a good fit for high school. You may need to modify these resources to use them with students. They may inspire you to look for similar content for your classroom.

Elementary School Teachers: I do not think these resources work for elementary students. They can spur your thinking about what you need to find for your classroom.

I shared resources for teaching Environmental Racism and Environmental Justice in my July 2024 post, How Schools Can Be Earth Reverent In The Age Of Generative AI.

Additionally, my post about AI experts has ten YouTube videos and two podcast episodes teachers can use to teach students about “AI.”

This post’s resources come from neither K-12 education publishers nor tech companies, with one exception.

Setting The Tone

Get started by watching this less than three-minute video by Univerity of Maryland Information Studies Professor Dr. Katie Shilton. Whether or not you show the video to students, it sets a good tone.

It is not about preaching or shaming students. It’s about educating and engaging them. Dr. Shilton’s informative approach to accuracy, copyright, labor, and bias establishes a positive tone for an educational dialogue.

Dr. Shilton said, “If you're using these tools to make your life easier, it's worth thinking about whose lives they made harder.” We want this empathy for students.

Karla Ortiz

Viewing award-winning concept artist Karla Ortiz’s US Senate testimony opening statement is essential for high school students. Go to 31:24 of this video:

Ortiz’s testimony helps students understand two concepts:

1. How generative AI pulls from data sets to generate outputs.

2. How generative AI impacts creatives.

This PDF of Ortiz’s US Senate testimony is a text version that students can read and respond to.

Set Up: Begin by having students view Ortiz’s art. Do not mention generative AI. Ask students to respond to the prompt, “What do you notice? What do you wonder?” Students can respond on sticky notes in FigJam. Students can observe and reflect on:

• The home page of Ortiz’s website.

• Ortiz’s illustrations.

• Ortiz’s fine art.

• Ortiz’s prints.

Have students read and watch Ortiz’s opening statement after they explore her work. As a class, compare reactions to her testimony to responses to her art.

Long-form Text

Two long-form New York Magazine articles are perfect for close reading, learning about generative AI, and critical thinking.

Stochastic 🦜 in New York Magazine

The Stochastic Parrots paper is the seminal academic work on Large Language Models (LLMs). An academic paper is too dense for most high school students. However, New York Magazine did a deep-dive article about one of the paper’s authors, Dr. Emily M. Bender.

• You Are Not a Parrot by Elizabeth Weil, New York Magazine, March 1, 2023.

Use chunks of the article (or possibly the whole text) to help students understand how LLMs work, including their limitations and harms.

Why A.I. Isn’t Going to Make Art in New York Magazine

Ted Chiang’s essay about generative AI and art can spur student conversation. Have students read and react to his arguments.

• Why A.I. Isn’t Going to Make Art by Ted Chiang, The New Yorker, August 31, 2024.

As with the Stochastic Parrots article, consider having students work with chunks of the text.

Chiang made a strong argument. Have students build critical thinking skills by critiquing the case Chiang laid out.

Civics and Election Misinformation

Conversing with students about generative AI and civics is especially important for the 2024 presidential election. When she endorsed Vice President Kamala Harris on Instagram, Taylor Swift wrote,

Swift’s status as possibly the most famous frequent victim of deepfakes is a perfect topic to engage high school students about generative AI.

Here are resources for incorporating generative AI into conversations about the presidential election, civics, and current events.

Proof News: AI Models Generate Misinformation About Presidential Candidates

Proof News documented that Large Language Models generate misinformation about the presidential election 30% of the time.

The article is perfect for students. So is the accompanying 13: 41 video, which teachers can edit with edpuzzle.

Rest of World’s 2024 Elections Tracker

Rest of World’s 2024 Elections Tracker posts articles about generative AI misinformation affecting elections in the US and worldwide. Teachers should check this resource often for the most recent examples and articles about this problem.

Civics Of Technology

Civics of Technology “shares research and curriculum and offers professional development that encourages teachers and students to critically inquire into the effects of technology on our individual and collective lives.” If ever there was a time when teachers needed such a resource, it is now!

Civics of Technology also has Inquiry Design Model Lessons and AI-specific resources for educators.

Data Workers Inquiry

One of the biggest misperceptions of generative AI in PK-12 education is that it can “carry out complex tasks with little or no human intervention.”1 Data Workers Inquiry from the Distributed AI Research Institute (DAIR) teaches about the exploited workers who constantly work to make generative AI function.

This is a great resource for building student empathy because the inquiries document the terrible conditions many data workers face. For example, Roukaya al-Hammada, a 29-year-old data worker, wrote an inquiry titled "If I Had Another Job, I Would Not Accept Data Annotation Tasks": How Syrian Refugees in Lebanon Train AI. She wrote:

Eddy Burback

A teacher talking to students about generative AI’s harms and limits might not land as well as a funny, successful 27-year-old YouTuber doing so. As the Data Workers Inquiry illustrates, generative AI is a heavy topic. Your students (and you) need some levity.

Eddy Burback’s funny videos critique technology and culture. His video helps students see generative AI from the perspectives of a user and a professional content creator.

Students may wonder if teachers are on their side. They will know Burback is when he says,

My favorite moment in Burback’s video is when he compares a historic photo of Little Italy to a Meta-generated one. For fun, I decided to see if I could replicate what Burback found. I entered this prompt into Meta AI: “Generate an image of New York's Little Italy in 1900.”

Compare an image of Little Italy in 1900 from Wikimedia to the objectively terrible result from Meta AI.

The image Google Gemini generated with the same prompt also has nonsensical text on the storefronts. Notice the Empire State Building has moved from Midtown to Little Italy and is fully built thirty years too soon.

Set Up: Use a tool such as edpuzzle to cut the Eddy Burback video into pieces because:

• The first and last segments (image generators and Large Language Models) are perfect for distinct lessons.

• The middle of the video explores self-driving cars and delivery robots. This footage is good for current events and critical thinking but will not help students learn about generative AI.

• Content Warning: There are a few f-bombs in the video. It is still a great resource for students. Edpuzzle can help you delete the profanity.

AI And The Environment

I wrote about schools remaining earth reverent in the age of generative AI. That post has good resources. In it, I wrote about Geographies of Digital Wasting. I neglected to mention the site’s fantastic Resources page which has many materials for building student understanding.

Here are more resources to engage students around generative AI and the environment:

The Environmental Impacts of AI Primer

This is the one resource in this post from a tech company. Hugging Face is a company that “develops computation tools for building applications using machine learning.”

Dr. Sasha Luccioni is Hugging Face’s climate and AI lead. She is my go-to source for information about generative AI and the environment. Listen to this episode of Dr. Luccioni on the Tech Won’t Save Us podcast to assess her knowledge of the subject.

She co-wrote a primer on generative AI environmental impacts for Hugging Face. There is also a PDF version of the primer. The primer includes:

• Statistics from existing research.

• Generative AI and energy consumption.

• Generative AI and water consumption.

• Generative AI and mineral consumption.

• Generative AI and greenhouse gas emissions.

• Legislation and regulation.

• Ways forward.

• A glossary of terms.

Forget Jobs. AI Is Coming For Your Water (Video)

This video vividly details the impact of generative AI on water in communities in Mexico and Uraguay. Teachers can use this to build student generative AI understanding and empathy.

Use edpuzzle to edit the video into smaller chunks that students watch over a few days.

Have students interact with the video and show them The AI Resource Test. Students can share how what they learned in the video affects their use of the test using FigJam sticky notes. Students can practice their critical thinking and reasoning skills by justifying their choices in light of what they learn in the video.

Continuing The Conversation

What do you think? How are you conversing with students about generative AI? What did I miss? What resources are you using? Comment below or reach out on BlueSky: tommullaney.bsky.social.

Leave a comment

Does your school or conference need a tech-forward educator who critically examines generative AI and pedagogy? Check out my professional development offerings, reach out on BlueSky, or email mistermullaney@gmail.com.

Post Image: The blog post image is by Max Fischer on Pexels.

AI Disclosure:

I wrote this post without the use of any generative AI. That means:

• I developed the idea for the post without using generative AI.

• I wrote an outline for this post without the assistance of generative AI.

• I wrote the post from the outline without the use of generative AI.

• I edited this post without the assistance of any generative AI. I used Grammarly to assist in editing the post. I have Grammarly GO turned off.

The dialogue around generative "AI" in education often centers on a future where effectively using "AI" is an essential skill without which people will be unemployable. Some of the rhetoric around this belief includes:1

• Today's students will retire in the 2070s and 2080’s meaning a 10-year-old will graduate high school in 2032 and college in 2036.

• Today's AI models are the most rudimentary ones our students will ever use.

• Students solely exposed to a traditional education system will find themselves unprepared to tackle future challenges. We must transform our schools into this AI era.

Notice the the future-tense "will" in all of those.

Teachers, Not Time Travelers

The message I take from this dialogue is that teachers must join with those who predict "AI" will fundamentally transform society. That means that teachers and schools base their pedagogy on this prediction (assumption).

Predictions Are Often Wrong

As illustrated above, Doc Brown was wrong when predicting 1985's political climate. Let's explore how predictions play out.

The History Of Technology Predictions

Our past is rife with incorrect predictions of people writing off technology that wound up being omnipresent.

As Adam Johnson said on the Citations Needed podcast (at 5:43), no one wants to be David Letterman in this video with Bill Gates rejecting the idea that the internet would be useful.

At the same time, you are not reading this using Google Glass or the Apple Vision Pro.2 You don't teach students in the Metaverse. You are not paid in cryptocurrency. Your retirement fund is not a portfolio of NFTs. You don't drive to work in a self-driving car.

No one wants to be Paris Hilton or Jimmy Fallon singing the praises of NFTs.

Similarly, no one wants to be Bill Clinton, sitting on stage with Sam Bankman-Fried. Imagine the shame of Roger Parloff, who apologized after writing a puff piece about Elizabeth Holmes and Theranos, a company whose secrecy and NDAs are reminiscent of OpenAI.

So tech predictions are dicey at best. Let's continue with some more presidential politics-themed predictions.

Recent Incorrect Predictions

Imagine if Marty McFly drove the DeLorean to the year 2000 and told someone there the presidents from January 2009 through January 2021 would be Barack Obama (who???) and Donald Trump3 (the real estate guy???). Could anyone have predicted those outcomes in 2000?

This summer, news media weighed in as Vice President Kamala Harris considered her running-mate options. Which former Social Studies teacher and Gay-Straight Alliance advisor is missing from these YouTube thumbnails?

Neither CBS News, The Associated Press, The New York Times, Politico, nor MSNBC (top 3 options!) predicted that Minnesota Governor Tim Walz was even a top contender to be Harris's running mate!

Looking at these incorrect predictions gives me solace. I have made so many over the years. On March 9, 2020, I wrote, “The fears over Coronavirus may be more hype than substance.” One prediction I especially regret is advocating teachers use Google Expeditions on iPads. Inherent to that argument is the assumption that the Google Expeditions would be around for a long time. Less than three years after I wrote about the greatness of Google Expeditions on iPad, Google announced Expeditions was headed to the graveyard. I feel terrible if any schools purchased iPads based on what I wrote about using them for Google Expeditions.

How many times have we seen fads in education come and go? Personalized learning, PBIS, MOOCs, and learning styles are among the many.

Finally, in the thirty states4 where sports betting is legalized, bankruptcies have increased by 28% compared to states where it is banned which tells us about our ability to predict sports outcomes.

Do you see why basing pedagogical decisions on the ability to predict the future is fraught at best?

Why AI Predictions Should Not Inform Pedagogy

What about "AI" and education rhetoric hinging on predictions?

Students in our schools right now will retire in the 2070s and 2080s.

Most students born in 1960 graduated high school in 1978 and college in 1982. Most students born in 1985 graduated high school in 2003 and college in 2007.

We can always roughly calculate when people will graduate high school, college, and retire by adding 18, 22, and 65 to their birth year.

This non-sequitur tells us nothing about whether we should or should not adjust schooling to respond to “AI.”

Today's AI models are the most rudimentary ones our students will ever use.

This argument often says that today's generative "AI" is Pong and is presumably destined to evolve into Super Mario Bros, The Legend of Zelda, NHL '94, and peaking with Madden '04.

Pong was not released to the public after billions of dollars of venture capital and tech giant funding and decades of research.

Pong did not have the inherent problems of generative AI listed by Ed Zitron at 14:08 of this episode of Better Offline.

"The problems are problems that aren't necessarily just technological. They're physics, they're mathematics. These aren't things you can just outrun." - Ed Zitron, Better Offline, April 10, 2024.

• Massive energy demand.

• Massive computational demand.

• Its "hallucinations."

• The insatiable hunger for more training data.

We must transform our schools into this AI era.

This centers schools as the problem when we consider "AI." Not AI companies such as OpenAI. There are legitimate issues with public K-12 schools in the US:

• School shootings (not schools' fault - extremely lax gun laws cause this).

• Segregation driven by government policy including redlining and funding schools with local property taxes.

• Teachers are underpaid.

• Many states do not have legally empowered teachers’ unions.

• Standardized testing (not schools' fault - federal and state laws cause this).

• Police in schools.

• Student tracking.

• Excessive student workloads.

• Book banning (state laws cause this).

• Surveillance and blocking websites students need access to.

When considering American K-12 schools' problems, "They are not responsive to AI" is not one of them.

I have heard the "We must transform" argument said another way, "Classrooms are getting kids ready for 1985."5

This argument makes me want to mount an Atticus Finch-like defense of falsely accused teachers. I will spare the reader but say that as someone who attended high school ten years after 1985, the classrooms I taught in and have worked with since 2003 look nothing like those I attended in 9th through 12th grade.

Teachers deserve better than being told their hard work is not preparing students for their post-K-12 years.

Speaking of preparing students for their post-K-12 years…

Wear Your Critical Thinking 🧢

Unless you own a flux capacitor-equipped DeLorean capable of reaching 88 miles per hour, you don’t know the future. Instead of the impossible skill of prediction, focus on critical thinking. This skill is always valuable for teachers and students.

Critical thinking about the future of "AI" includes questions such as:

• Will "AI" skills become invaluable when capitalist institutions such as Goldman Sachs and the RAND Corporation have joined the chorus of voices raising concerns about "AI"?

• Is students’ future employment tied to a technology that depresses purchasing decisions when mentioned in product descriptions?

• What can we learn from the widespread acclaim Procreate CEO James Cuda received when he said, “I really f***ing hate generative AI” and announced the app will have no AI features?

• Why are we basing pedagogical decisions on an environmentally unsustainable technology?

• Why must we “transform our schools” around a technology that violates US copyright law? As Former General Counsel, U.S. Copyright Office Jacqueline Charlesworth wrote in the Yale Law Journal, “…Each of the four factors of section 107 of the Copyright Act weighs against AI's claim of fair use, especially when considered against the backdrop of a rapidly evolving market for licensed use of training materials.”

• Why are teachers asked to change pedagogy around ChatGPT and Google Gemini when both are unavailable to students younger than 13 and require guardian permission for students aged 13 to 18?

• Why aren't predictions that there might not be enough data to sustain generative "AI" not as valid as predictions "AI" will revolutionize society?

• Can we wait until the inherent problems with generative "AI" are solved before we change instruction to accommodate it?

• Can we wait until generative "AI" shows any ability to generate a profit before declaring that effectively using it is the skill all students need?

When considering generative "AI" in the classroom, please wear your Critical Thinking 🧢. As someone in K-12 education, you have great critical thinking skills. However, as awesome as your critical thinking skills are, your ability to predict the future is lousy, just like everyone else.

Continuing The Conversation

What do you think? How are you using critical thinking to address “AI” this school year? Comment below or Tweet me at @TomEMullaney.

Leave a comment

Does your school or conference need a tech-forward educator who critically examines AI and pedagogy? Check out my professional development offerings, reach out on BlueSky, or email mistermullaney@gmail.com.

Post Image: The blog post image is from Back To The Future Trilogy on Pexels.

AI Disclosure:

I wrote this post without the use of any generative AI. That means:

• I developed the idea for the post without using generative AI.

• I wrote an outline for this post without the assistance of generative AI.

• I wrote the post from the outline without the use of generative AI.

• I edited this post without the assistance of any generative AI. I used Grammarly to assist in editing the post. I have Grammarly GO turned off.

• There are no generative AI-generated images in this post.

Critical Inkling is a reader-supported publication. To receive new posts and support my work, consider becoming a free or paid subscriber.