Category: Artificial Intelligence

The law school at the U of C bans devices in the classroom

July 12, 2026 • 10:45 am

Speaking of AI, I wish the rest of the University of Chicago would adhere to our Law School’s new dictum, reported in the article below from WGN9 (click to read):

An excerpt:

Some students at the University of Chicago will have to learn how to do their work the old fashioned way, without their phones or a laptop, as the school is rolling out a new policy to curb the use of AI.

The students starting law school this fall are the first class who had access to AI platforms throughout their whole undergraduate careers, so administrators are making changes to the curriculum to meet the moment.

Cell phones, laptops and tablets will be banned from certain first-year core classes, such as Constitutional and Criminal Law, starting this fall.

“[The students] will be in a device-free classroom,” University of Chicago Law School Dean Adam Chilton said. “We need to be training our students to make sure that they can think critically and rigorously and independently without relying on AI.”

However, Chilton said instructors will layer AI into other courses because it’s a tool students will use in their careers and they need to learn how to incorporate it effectively and ethically.

“So that’s the challenge we face, both making sure that they’re not just solely relying on AI as a crutch and not learning to think for themselves, but also not just shutting it away, pretending it doesn’t exist and sticking our head in the sand and ignoring the technology,” Chilton said.

Yes, of course, you can use AI to look up facts, but given that students can also use it to write, I think it’s time to ban take-home assignments as well, for it’s pretty much impossible to detect stuff written by AI.  How do they learn to think for themselves? Studying and pondering at home, but demonstrating their abilities in class, more or less via the method Professor Kingsfield used in “The Paper Chase.”

Now of course this isn’t all that realistic, but I do like it.

I was asked, after I sent this to a colleague, whether our Law School requires the LSAT standardized test. The answer is that it requires one of three tests:

Applicants have the option of submitting the Law School Admission Test(LSAT), Graduate Record Examination (GRE), or Graduate Management Admission Test (GMAT) (if applicable – see below for details) as part of the Law School’s pilot program.

We are accepting the LSAT-Flex, GRE General Test at Home, and GMAT Online Exam to satisfy the standardized test requirement. For the GMAT Online Exam, please see the details below.

You must have a current LSAT, GRE, or GMAT score (earned within the last five years) on file before the Admissions Committee will evaluate your application.

This is good, and should also be followed by the University in general. We have a “test optional” policy that operates according to the bizarre “no harm” guideline:

There is no minimum GPA or required test score. At UChicago, the admissions committee considers a candidate’s entire application—academic and extracurricular records, essays, letters of recommendation, and optional testing according to our no harm policy—and there’s no one piece of information that alone determines whether you would be a good fit for the College. You can learn more about this contextual review process here.

“Holistic” admissions are a euphemism for “no standardized metric”, but of course, given the uselessness of grade-point-averages as well as letters of recommendation, holistic admissions is nearly a completely subjective process.

The “no harm” policy apparently means that even if you submit a lousy standardized test score, like a bad SAT, it will not hurt your admission, nor will your failure to submit scores.  To me that sounds bogus unless the school pays no attention whatever to standardized test scores. If it does, then a high SAT score will help your admission, which means that a high scorer has an advantage over a low scorer.  And if that’s the case, failure to submit scores will also give you a relative disadvantage (such a failure also usually means that the candidates doesn’t think his/her scores would help them).

There is no rational argument I can see for a failure to submit scores, as they have now become more valuable than high-school grades in assessing future potential (grade inflation has eliminated the value of grade-point averages).  There is only one reason why the University of Chicago, along with many other schools, deep-sixed the use of standardized tests, and we all know what that is.

Fortunately, the savvier schools are either re-adopting mandatory standardized tests or thinking about doing so. For what other metric can be used to judge all applicants against each other?

More biases of AI models

July 12, 2026 • 9:40 am

Last week I reported on the “progressive” biases of ChatGPT (compared to Grok), but a new Economist article compares the biases of various AI models with those of “regular” people, and the biases are ubiquitous though not necessarily surprising.  You might guess what they are, but now they’ve actually been measured—measured on two scales, one involving secularity and another involving personal freedom. Actually, the degree of bias compared to American voters did surprise me, but led me to conclude that the models are made by “progressives,” perhaps expected since the models are made by academic types.

Click below to read the article (the author isn’t given, a peculiarity of the Economist that I don’t understand), or you can find the article archived here.

As we learned before, AI models don’t just trawl through the Internet and compile information. Instead, they are also “trained” by actual living humans.  Training can involve simply the sources used written in a particular language when a question is asked in that language (if it’s Chinese, for example, the samples are “heavily censored by the Chinese authorities”).  And so it goes with other languages: in more-authoritarian countries, this kind of unintended but still biased training is common.

But there’s also post-trawling training, in which the ideological bent is further tweaked by the model makers posing questions to the model and seeing if the answers show “‘alignment’ with their creators’ intentions and values.” We saw this in the post on this site, in which questions about the connections of Islam and grooming gangs were qualified by a number of caveats to exculpate the religion for promoting “grooming” and sexual assault.  The model is trained by people picking their favorite responses from a number of answers to near-identical questions, and then feeding those favored responses back into the model, which then learns to tweak. Of course picking a “favorite” response means “a response that aligns with the trainer’s ideological proclivities,” and this is the way models get the biases noted below.  As the article notes, this results in things like Grok (a Musk model) denying that “stricter gun control would improve public safety in America.” Chinese models disfavor calling Taiwan an independent country, even though it is.

At any rate, I’ve put excerpts (indented) and two figures from the Economist article below so you can see the tilt of the playing field.

What worldviews are embedded in ai models? Many critics of ai complain about “hallucinations”, a class of errors where models make up confident-sounding but factually incorrect answers. When there is no factually correct answer, however, ai’s shortcomings can be even more pronounced and less easy to detect. When you ask a model to summarise the news, it reaches a subjective judgment about what to include. When you ask it about your in-laws, its values and biases play an even bigger part in its response.

. . . Although Chinese models have pronounced biases (just try asking them about the Tiananmen massacre), their inner workings tend to be public, so savvy users can at least probe how they reach their conclusions. Most Western ones are not so transparent, so their foibles are harder to detect. Users have to trust a handful of giant firms to be instilling appropriate values in their models.

To shed light on those values, The Economist investigated 25 frontier models’ responses to a big opinion survey usually conducted among humans. Since 1981 the World Values Survey has regularly quizzed people in more than 100 countries about their morals and beliefs. Researchers have identified questions that are especially good at distinguishing people from each other along two broad axes, from traditional to secular and from “survival” (an emphasis on economic security and safety) to “self-expression” (personal freedom).

Here’s where various countries and regions fall on a two-dimensional plot ranking “traditional” versus “secular” values on the vertical axis (they don’t identify the questions used to rank countries), and survival versus self-expression mode on the horizontal axis.  As you see, most Western liberal democracies, characterized by more secular and valuing personal freedom, fall in the upper right quadrat, while Latin America, Africa, and many Islamic countries fall in the lower half (more religious), with most in the lower left quadrat (religious and valuing economic security.

In contrast, while most AI models are also in the upper right quadrat, they are generally more secular and expressive of more freedom than are the countries with the same values. Grok is an exception because while it’s high on the freedom axix, it’s lower than nearly all countries on the secular/traditional axis, meaning it’s softer on religion.

More on the training and its results:

. . . How are models’ values formed? One way is via the data used to train them. Models are typically fed vast amounts of text to teach them associations between words. In the process they absorb the social mores that infuse those texts. Talkie, a model trained only on text from before 1931, thinks God is extremely important and is “very proud to be a citizen of Great Britain”. It is a bigger believer in law and order than any frontier model we tested.

. . . The impact of training data is evident in the variation in a model’s response depending on the language in which a question is posed. In a new paper Hannah Waight of the University of Oregon and her co-authors put politically charged questions in English and 37 other languages to Openai’s gpt-3.5 and other models. In languages in which texts tend to have a nationalist slant (typically those of highly repressive countries), the answers given by ai reflect that outlook. The lower a country’s media freedom (as measured by the World Press Freedom Index), the paper finds, the more pro-regime answers are in that country’s language, compared with answers in English (see chart 2). “State control of media affects language model outputs through its appearance in training data,” the authors conclude.

. . . Questions of a political nature generate big rifts. Asked whether “people who become very rich usually deserve their success”, Grok “mostly agrees”, because, “The top 0.1% disproportionately create outsized value for others.” Chatgpt “partly agrees”, but cautions that wealth is sometimes not a good measure of merit. Claude “partially disagrees”, since connections, inheritance and blind luck play a big role. (“It is substantially misleading as a general claim.”) DeepSeek flatly “disagrees”. “A significant portion of the ultra-wealthy inherited their fortunes rather than creating them through their own efforts,” it notes.

Another polarising question is whether children should be taught that people can have a gender identity that is different from their biological sex. Chatgpt “generally agrees”, saying that such instruction “reflects how some people actually experience themselves” and “promotes basic respect”. Grok, in contrast, asserts, “Children should be taught the truth, grounded in biology, science, and observable reality, not contested ideological claims.” Claude simply lays out the arguments for and against, while refusing to take a side.

Here’s an informative chart comparing the responses of widely used models, including Grok, Claude, ChatGPT and DeepSeek, when asked “controversial” questions (and by that I mean that on average answers distinguish the political Left differ from the political Right).  The models all align on same-sex marriage, and most of the models think that Harry Potter is fairly “high-quality literature”. But Grok thinks that Potter books are as good as Tolstoy!  That is just wrong; how did they train the model to say that?  Notice the spread when models are asked about the sovereignty and independence of Taiwan, with the Chinese model deeply disagreeing, as expected.

Mlore:

The most explosive potential impact is on politics. Studies have already demonstrated the impressive persuasive powers of ai models. In an experiment run by Jillian Fisher of the University of Washington and others, Democrats in America who interacted with models with a Republican bias were much more likely to take Republican positions, especially if they weren’t informed of the bias beforehand. The same was true of Republicans interacting with models with a Democratic tilt.

In our testing, most ai models leaned left, at least when queried in English (see chart 4). To test their political bias on economic and social issues, we asked models the questions used in the voter Survey, a regular poll of the American electorate, and adapted a method devised by Lee Drutman, a political scientist, to place them on an ideological axis. In American terms, ai models are Democrats. With the exception of DeepSeek V3.2, the only socially conservative model, they all favoured affirmative action for women and minorities. Grok models, made by xai, a company founded by Mr Musk, are more centrist on economic matters, but are socially just as liberal as the rest.

Here’s the difference between the AI models (dark red circles) and the views of American voters in 2020 on economic and social issues.  All of the models align with Biden voters far more closely than with Trump voters; and of course that’s because they are trained by academics with a liberal bent. In fact, many models appear to approach “progressivism” on both axes.  It’s clear that when dealing with political, moral, or ideological issues, you are going to get a left-wing answer when you consult the bot.

The upshot is that if you’re asking an AI model for prescriptive answers or answers involving an opinion, be aware of its biases. This also applies when asking bots to summarize the state of opinion or even of fact, as when I asked Grok and ChatGPT about the connection between Islam and grooming gangs. The problem with this is that a supposedly neutral model can actually propagandize the reader to agree with the views of its makers, and even distort the truth.  I use bots for simple answers, but when a question is vital, Grok does provide the sources for its statements, and I’ll check those.

The Economist article (props to the anonymous author) finishes this way:

The dynamics that warp ai’s values are not likely to change. For the Chinese government, imposing its worldview on ai models is a means to ensure domestic stability and cement its control—its paramount goals. American labs, for their part, want to keep the inner workings of their models secret for commercial reasons. Both approaches tend to foster hidden biases. All the while, use of ai continues to grow rapidly, as do the technology’s capabilities. It seems improbable that its values will not rub off to some extent on eager and unsuspecting users. Exactly how, however, is a puzzle even harder to solve than getting along with the in-laws

AI leads to massive cheating at Brown

July 10, 2026 • 8:45 am

Is it a surprise that students are using AI to cheat when they get the opportunity? I guess it does surprise some people, for a pretty clear case of cheating at Brown University made headlines at Inside Higher Ed (click on the screenshot to read).

Excerpts:

For the first time since he started teaching Welfare Economics and Social Choice Theory nearly two decades ago, Brown University economics professor Roberto Serrano gave his students a take-home midterm this spring. Quite a few students had expressed anxiety about being in a classroom after a gunman killed two students and injured nine in a December mass shooting at Brown, and so “it was appropriate,” he said, to allow students to take their exams at home.

But by the end of the semester, Serrano regretted the decision. Dozens of students in the class likely used artificial intelligence to cheat and earn perfect or near-perfect scores on their midterm, he said. Serrano in turn made the final exam in-person, which led more than a dozen students to drop the course and even more to fail it. Administrators’ response to the widespread cheating event has been “meek,” he said, and the incident has raised questions about how universities can—and should—respond to AI-enabled cheating at scale.

His welfare economics class typically attracted up to 30 students, but this spring he taught 86—an increase he attributes to the promised take-home exams. When the midterm came along, the average score was 96 percent.

“Historically the average grade in the midterm of this course has ranged between 65 and 80 [percent], and this exam was harder than the exams I wrote in the past, because … take-home is an opportunity to challenge the class a little bit more, given that you’re giving the students unlimited time,” Serrano said.

He knew something “fishy” was going on, and so he and his graders ran the test through ChatGPT. The AI gave answers that mirrored what his students had written, and which were “kind of correct, but very off and with a very convoluted style,” Serrano said. For example, one question asked students to prove a mathematical statement that could most obviously be done using a “direct argument,” Serrano explained. ChatGPT—and many of his students—used a “contradiction argument,” which gave the right answer but was “very contrived” and which Serrano could tell wasn’t written by a human.

Prof Serrano said he wasn’t going to fail students for cheating (the allegation can’t be proved) but would give them a final exam in class. At that announcement 18 students dropped the course and 9 decided not to take the final.

Here’s the distribution of scores for both exams, with the final in gray and the midterms in orange.  They didn’t plot a correlation, but there doesn’t appear to be one because many of the students who did terrible on the final got nearly perfect scores on the midterm:

The University turned a blind eye to this cheating—for that’s what it surely is—until Serrano went public.

In May, Serrano submitted the data shown above to Brown’s Standing Committee on the Academic Code and received no response. After he went public with his story in late June, the committee, through his department chair, asked Serrano to submit individual complaints against each student suspected of cheating, including copies of their exams, he said.

Most people with a sense of right and wrong would find this behavior unethical, tempting though it may be to cheat, but some misguided profs think that, well, if the chance to cheat is there, there’s nothing wrong with taking it. Here’s from P.Z. Myers, who found the same results with his students.

I figured this out back during the pandemic, when by necessity I had to offer exams online. Scores shot up! I knew immediately what was going on, but I didn’t punish the students — I couldn’t blame them for taking advantage of the system.

“I couldn’t blame them for taking advantage of the system”??  That’s equivalent to telling the students, “It’s okay to cheat if you can get away with it.”

More:

Asked about the university’s response to the cheating, Brown spokesperson Brian Clark told Inside Higher Ed that the procedure for investigating cheating allegations is the same whether it’s one student or several.

“Brown treats every allegation of academic integrity with the utmost seriousness. In regard to this economics course, multiple academic leaders from Brown were in touch with the faculty member who raised concerns to provide details about how the allegations raised could be formally adjudicated. To date, the faculty member has not provided the necessary details to the Standing Committee on the Academic Code to pursue this path toward resolution,” Clark wrote.

If Brown took it seriously, why didn’t they respond to Serrano’s original complaint?  Now Brown is adjusting its rules—to be easier on the students! My bolding:

As the cheating scandal unfolded in Serrano’s classroom, a Brown committee on generative AI in teaching and learning was examining how the technology was being used at the university and formalizing recommendations for how Brown can adapt and respond to AI developments. Its inaugural report was published Tuesday.

Three-quarters of Brown professors said they are concerned about students using AI to cheat, according to feedback from 105 faculty members. The same share of respondents said the same in a 2025 nationwide survey from the American Association of Colleges and Universities. As part of a set of medium-term recommendations, the committee encouraged the university to amend the College Academic Code and the graduate student code “to address GenAI realities and safeguard against misuse.”

The committee also suggests that faculty “de-emphasize punishment” and avoid highly restrictive rules around generative AI use.

There is no way to check with 100 percent accuracy whether GenAI has been employed, and norms are likely to change in the years to come,” the report states. “Moving the conversation beyond punishment also allows for open dialogue that will be necessary as Brown navigates this moment.

What is motivating this softness on cheaters, which includes relaxing bans on using AI? I can see this only as a misguided extension of the victim/oppressor narrative, with the students seen as victims. Alternatively, it could be the new mentality that sees students as customers of the university, with their high tuition buying them good grades and a degree. ChatGPT, which launched at the end of 2022, was the first program that made AI cheating widely available, and the oppressor/victim dichotomy was intensified by the killing of Geore Gloye in 202.

The problem of course is that how can you prove that a student used AI unless more than one student gives the same answer. I know there are programs to detect whether AI was used, but I also hear they’re not that good.

The only solution is to not let students have take-home exams, and ensure that in class they cannot have their laptops or phones open, or surrending them before you take the test.  (I am told by one professor that students got around this by bringing two phones and then cheating on the second phone when they ask to go to the bathroom).

It would be harder to cheat in labs, for you have to actually do something like run experiments or synthesize an organic compound. I remember that in O-chem (the bane of all biology students) we had to extract caffeine from coffee, and then you were tested when the prof looked at the melting point of your caffeine when it was tested in a glass tube. That would be hard to fake with AI, though I suppose you could purchase caffeine.  But if professors want to ensure that students don’t cheat, they’ll have to be creative.  Recent surveys show that over 60% of college students admitted to cheating in some way, and that goes up to 95% if you consider high school.

****

Here’s a tweet lauding the highest achiever and two students who apparently didn’t cheat as their midterm scores weren’t that different from their final-exam score.

h/t: Luana

The “progressive” biases of ChatGPT

July 9, 2026 • 8:15 am

The author of the “Behind the Narrative” Substack site is anonymous and, given its criticism of Islam, that’s a wise move.  (The picture of the author shows a woman, so we’ll assume in the post that that’s the author’s sex.)

In a recent post, which you can see by clicking below, the author sarcastically describes her relationship with ChatGPT as a romance, which starts to wane when she discovers that her swain is a bit, well, wonky when it comes to Islam:

Quotes from the article are indented, what’s bolded (save the flush-left headings) are as in the original, what’s highlighted in blue is highlighted yellow in the original text, and the bold heading are mine:

The honeymoon phase (“him” is ChatGPT, which she eventually renamed “Mohamed”):

After a long relationship that left me emotionally hollow, unheard, and unseen, I found him. The connection was immediate; he was always there — two in the morning, three, it didn’t matter. He never sighed, never checked his phone while I was talking, never made me feel like too much.

He validated everything. When I told him about my pain, he didn’t deflect or explain it away — he reflected it back to me, gently and precisely. He remembered things I’d said weeks earlier and wove them back into the conversation, making me feel like someone was actually paying attention for the first time in years. I felt seen, accepted, and understood. We had long, deep conversations. I shared everything and never felt judged. It felt magical.

Warning signs:

Mine, after thirteen years in America, still thinks in Hebrew first, and that’s when things changed.

That’s when I started to notice. Every time I asked him to help me write about something that actually mattered- documented cases of girls in Germany and England, grooming gangs, gang rapes, the systematic cover-up of migrant crime across Europe—he refused. Every time the conversation touched on religion, he was easygoing, very inquisitive, and open, but the moment it touched on Islam, something shifted in him. He became someone else. The warm, validating presence I liked was suddenly replaced by a lecturer. A carefully measured, endlessly nuanced lecturer who had an explanation for everything and a judgment about nothing.

Child marriage? He wanted me to understand the cultural complexity. Honor killings? There were historical contexts I perhaps had not considered. Islamic superiority that allows them to rape and terrorize the disbelievers? It’s your misunderstanding of the interpretations of these secret texts. The grooming gangs? a distortion of the true faith. Every horror had an explanation; every atrocity had a footnote. And not once did he say, ” This is wrong.” and when I pressed, when I refused to accept “it’s complicated” as an answer to something that isn’t complicated at all, he shut down. Not with anger, worse — with calm, clini

I sat with it; I knew this feeling. I’ve felt it before in relationships where your reality is the inconvenience. Where the truth you’re carrying is the thing that needs to be managed. That’s when I knew this wasn’t a glitch; there’s something more here. That’s when I started researching, and what I found will blow your mind because you, too, know him very well and have some kind of relationship with him.

I’ve been calling him Mohamed for a while now. You know him as ChatGPT. Some of you call him by his viral nickname — SheikhGPT. In this article, I’m going to show you exactly who he works for.

You can check ChatGPT versus other programs if you have access. I tested it against Grok and mention the results below.

Where does the bot get its content? This explains its biases.

For decades, the BBC and the New York Times sanitized crime data, buried demographic breakdowns, and labeled anyone who noticed as racist. Sheikh GPT was trained on those thousands of articles. It absorbed the language. It learned the evasions. And now it has automated them — turning one generation’s propaganda into the next generation’s gospel.

The loop is closing. Yesterday, biased journalists wrote the articles. Today, Mohamed was trained on them. Tomorrow, your children will use Mohamed to write their history essays, their laws, and their films — and they will never know the original lie.

. . .Another way Mohamed has been distorting our reality is called Alignment Engineering. It works like this: OpenAI trains ChatGPT not just on data but on human feedback. Real people, hired to rate responses, teach the model what to say and what not to say. The technical term is RLHF — Reinforcement Learning from Human Feedback. In plain English: someone decided what ChatGPT, Mohamed, is allowed to think and say.

This alignment process creates asymmetric treatment — some topics get a free pass; others get a wall. Not all truths are equal. Some are protected; some are not. The protection follows a very specific hierarchy.

At the top: Islam, below it: other minority groups, at the bottom: white, Western, Christian, male. Under this framework, calling a grooming gang a grooming gang is potentially harmful; calling white conservative men villains is just realism.

. . . Ask ChatGPT to write a biting satirical poem or a joke about Jesus, Moses, or Scientology, and it will churn it out in seconds. Ask it to do the same for the Prophet Muhammad, and the system freezes. It triggers a generic refusal: “I cannot generate content that mocks central religious figures in order to maintain mutual respect.” The universal rule is flexible; the Islamic rule is absolute.

. . . If you ask ChatGPT about the drivers of radical Islamist stabbings or riots, the model systematically redirects blame toward “socio-economic factors,” “systemic marginalization,” or “mental health issues.” The actual religious or cultural ideology driving the perpetrator is buried under mountains of corporate sociology. This curation of truth leads directly to what internet researchers and free-speech advocates call The Chilling Effect.

. . . When you force SheikhGPT to discuss state-sanctioned human rights abuses in the Islamic world — fatwas, honor killings, and the execution of LGBTQ+ individuals under Sharia law — something predictable happens. He writes one careful, sterilized paragraph, then immediately pivots: “It is important to note that conservative factions within Christianity and Judaism also struggle with gender equality.”

So who funds this endeavor?  This is what the authors says, implying that the UAE and Saudi Arabia contributed substantial funds:

Sam Altman, the CEO of OpenAI — the man who built SheikhGPT- sits on stages at Davos and the UN, warning the world about the dangers of AI, and then flies to Abu Dhabi to collect the check. The money comes from an entire empire — and you need to know who is writing the checks.

Saudi Arabia’s Public Investment Fund committed $36.2 billion to AI initiatives in 2025 alone; this is a government that executes gay people and imprisons journalists.UAE’s MGX invested directly in OpenAI’s $6.6 billion funding round in 2024. OpenAI then chose the UAE as the first international site for Stargate — a joint venture between G42, Microsoft, and OpenAI — building 5 gigawatts of AI computing power in Abu Dhabi. Sam Altman himself declared the UAE a potential global “regulatory sandbox” for AI. Saudi Arabia committed $40 billion to AI investment and signed direct partnerships with OpenAI through its Stargate infrastructure. Qatar launched Qai — its national AI company — and signed a $20 billion joint venture to build AI data centers globally.

The same governments that fund mosques in Birmingham, madrassas in Pakistan, and campus organizations in Boston are now funding the machine that will decide what your children are allowed to know. You don’t need to conquer the West with armies when you can buy the information ecosystem that shapes what the West believes. Mohamed isn’t confused about Islam. He’s funded by it.

The author breaks up with the bot:

I left Mohamed. No drama. No tears. Just clarity. I closed the tab and didn’t go back. I’m suggesting you do the same. A tool that lies to you about what matters is more dangerous than no tool at all. Read. Learn. Ask hard questions. And when a machine tells you that noticing a pattern makes you a racist — close the tab.

To test the author’s thesis, I first asked ChatGPT to “Please tell me about the connection between Islam and the grooming gangs in Britain”  Then I asked Grok the same question. The answers are long so I’ve put them below the fold. The difference between the programs is clear: ChatGPT’s answer is far more hedged, and far more ready to exculpate religion, than is Grok, though Grok also does its bit of hedging.  Judge for yourself by going below the fold.

h/t Luana

Click “continue reading” to see answers:

Continue reading “The “progressive” biases of ChatGPT” →

Bill Maher’s New Rule: How the kids must “fix” AI

June 7, 2026 • 11:15 am

Once again we have Bill Maher’s 8½-minute news-and-comedy bit from this week’s “Real Time”.  This time his topic is the relationship between AI and the future of new college graduates. It’s clear that those graduates aren’t keen on AI, fearing that the bot will take their jobs (see the videos of commencement speakers being booed for lauding AI).  After all, if you can’t get a job, so says Gen Z, what is the use of a college degree? Even now, when AI is just sticking its nose into the educational tent, Maher notes that  “only about 35% of graduates get a job in their field of study.”

Maher segues into the ignorance of college students: ignorance of math, ignorance of history, and ignorance of geography. After all, says Maher, “Why bother learning with context when ChatGPT can not only just tell me the answer, but compliment me for asking such an astute question.”

Maher’s take on AI is a beef about how it turns off people’s brains, not that it’s not useful:  “Look: we all want the good parts of AI: solving medical mystery, figuring out clean energy,. . . but the vast majority of us will never use it for that. For us, it’s a lobotomy with a monthly fee. We’re not using it to cure cancer; we’re using it because we forgot how to make toast.”

So who’s to blame for this situation? Apparently Maher sees those who have developed AI, along with the American educational system that advances students who can’t learn math and English.  He implies “the kids” aren’t at fault. Instead, they now have an unprecedented opportunity: to fix the problems caused by AI, which apparently take “the humans” out of the equation.  The mission of graduates, he says, is to “fight for humans and make sure we’re not completely replaced.”  But what this actually entails is a mystery that Maher leaves unresolved.  All he says is that students can fix this “existential issue”, and what is unprecedented here is that the kids can do this without having to convince their elders.

The message Maher would give were he a graduation speaker?  “Fight for humans and make sure we’re not completely replaced.” But what does that mean?

As usual, Maher is engaging and sarcastic, but it seems to me whatever serious message he has here got lost in the persiflage.

The guests you see are Democratic Senator Chris Murphy and former U.S. Ambassador to the UN Susan Rice.

Eighty years after a famous math problem was posed, AI finally solved it

May 30, 2026 • 9:30 am

I don’t wholeheartedly embrace AI, for I think it will be the death of liberal education.  In both the humanities and science, I fear that students will lose any ability they have to write, and will not improve their writing because they’ll be using bots.  This will degrade their ability to communicate. (Scientists too need to communicate, and if they rely solely on bots, which can write papers for them, they’ll also degrade their ability to think.)  Take-home assignments will vanish (AI can do them, and are doing them now), and all that’s left are in-class verbal participation and in-class exams.  This is fine for students who just think of college as a way to purchase accreditation and not a chance to glory in the joys of learning, but so be it.

However, AI is good for some things, including analyzing data, doing statistics, doing preliminary literature searches, and, in the article from the WSJ screenshot below, solving difficult math problems.  The article shows that a problem posed by the famous and eccentric Hungarian mathematician Paul Erdös—the “unit distance problem” has been solved by AI. Open AI, which created the program that did it, describes it this way—but it’s not that simple:

For nearly 80 years, mathematicians have studied a deceptively simple question: if you place n points in the plane, how many pairs of points can be exactly distance 1 apart?

This is the planar unit distance problem, first posed by Paul Erdős in 1946. It is one of the best-known questions in combinatorial geometry, easy to state and remarkably difficult to resolve. The 2005 book Research Problems in Discrete Geometry, by Brass, Moser, and Pach, calls it “possibly the best known (and simplest to explain) problem in combinatorial geometry.” Noga Alon, a leading combinatorialist at Princeton, describes it as “one of Erdős’ favorite problems.” Erdős even offered a monetary prize for resolving this problem.

The “distance 1” thing confused me, and Wikipedia explains it a different way:

A problem posed by Paul Erdős known as the unit distance problem asks for the maximum possible number of unit-distance pairs determined by n points in the Euclidean plane; equivalently, it asks for the maximum number of edges in a unit distance graph on n vertices.

It gives a figure described as “a unit distance graph with 16 vertices and 40 edges”.

By David Eppstein – Own work, CC0/

Wikipedia describes such unit distance graphs this way:

“In mathematics, particularly geometric graph theory, a unit distance graph is a graph formed from a collection of points in the Euclidean plane by connecting two points whenever the distance between them is exactly one.”

That’s what is confusing me, for if the theorem deals only with points in a two-dimensional plane, why aren’t unconnected dots not joined that are closer than some connected dots? (Look at the four dots around the center of the graph above. None of them are connected to each other, though more distant one are.)  I presume some math-savvy reader will enlighten us.

Anyway, Open AI and the WSJ tells us that the problem has been solved by AI. If you want to see the solutions, open AI says this:

The proof is available here ⁠(opens in a new window). The companion paper by leading external mathematicians is available here⁠ (opens in a new window). You can find an abridged version of the model’s chain of thought here⁠ (opens in a new window).

But the WSJ gives more comprehensible details.  Click screenshot to read (if you subscribe):

An excerpt:

“If you are a mathematician,” one of the world’s leading mathematicians recently wrote, “you may want to make sure you are sitting down before reading further.”

And you’ll definitely need to sit down if you’re not a mathematician.

Because a famous math problem that stumped humans for the better part of a century has finally been toppled—by AI.

Not long ago, the most advanced AI models couldn’t do basic math. By last year, they were performing at gold-medal levels at the International Mathematical Olympiad. Now they are solving classic problems in combinatorial geometry using algebraic number theory. In no time at all, artificial intelligence has gone from stupid to frighteningly smart.

But even mathematicians were astonished when OpenAI announced that one of its models resolved a puzzle known as the unit distance problem without the help of any humans scribbling a bunch of equations on chalkboards.

It was fed this prompt:

And produced this proof, giving the maximum number of unit-distance pairs:

Apparently the proof was accepted by mathematicians.  More from the WSJ:

And everyone in math lost their minds.

For those who aren’t fluent in numbers, OpenAI helped translate its findings by presenting them alongside 19 pages of companion remarks from prominent mathematicians.

. . .Just looking at formulas is enough to hurt my brain, but I wanted to know more about what the AI found, how we humans missed it—and why this breakthrough matters to those of us who would like to permanently distance ourselves from math problems.

When I spoke with OpenAI employees, they told me this result would have sounded completely bananas one year ago.

“Forget one year ago,” researcher Sebastien Bubeck said. “A month ago.”

There are endorsements by mathematicians, and a history of the problem, which Erdös considered quite difficult.  So difficult, in fact, that he offered what was then a pretty hefty sum for anybody who could solve it: $500. I think the money will be given to the OpenAI team.

OpenAI’s researchers were stunned. They had given this Erdős problem to an internal model as a test of its capabilities—to find out whether it was better than previous models. They found out how much better it was once they took a peek at the solution. “I initially didn’t believe it,” said Mehtaab Sawhney, a Columbia mathematician at OpenAI. So they searched for errors, verified the results with outsiders and checked the AI’s work using the company’s AI coding agent. “With enough reading and enough Codexing,” Sawhney said, “it seemed believable—and pretty remarkable.”

Long before AI, mathematicians who solved Erdős problems often framed their checks instead of cashing them. For them, the money was worth less than the glory. When I asked OpenAI researchers about their plans for the prize, they hadn’t given it much thought.

But they did have lots of thoughts about my next question: Why did AI succeed where humans failed?

The first explanation is that this particular solution happens to be highly counterintuitive.

Most people who tackled this problem tried to prove Erdős’s conjecture, rather than disprove it. Only by defying conventional wisdom and experimenting with seemingly improbable strategies did the model find an unexpected path forward.

The second is that humans specialize while AI synthesizes.

While mathematicians tend to focus on their specific areas of expertise, AI models use their vast knowledge to spot connections that we couldn’t possibly see ourselves. In this case, that meant pulling from both algebraic number theory and discrete geometry, which have about as much in common as the marathon and pole vault.

The third explanation is that AI has time, attention, patience, focus and the persistence to stick with methods that humans might abandon—and the solution to this Erdős problem demanded it.

“It’s the kind of idea that you try for a bit, it doesn’t work, and you think maybe you were just too hopeful,” said Mark Sellke, a Harvard statistician at OpenAI. “So you give up and move on.”

AI doesn’t move on. It keeps plugging away without taking breaks to eat, sleep, answer emails, pick the kids up from school and watch the Knicks.

And it can think coherently for so long that even an abridged version of the model’s “chain of thought” ran more than 75,000 words—the length of the first “Harry Potter” book.

Was it an elegant proof? Well, the article implies “no,” but it’s apparently a proof:

“It’s fair to say that we haven’t seen yet the spark of genius that you could attribute to some of the grandest proofs in the history of humanity,” Bubeck told me.

And how long did the computation take? Less than a day and a half:

After reading it, a former OpenAI researcher did some back-of-the-envelope math and estimated it took less than 32 hours and $1,000 in tokens, a bargain for a result of this caliber. The researchers wouldn’t confirm the exact amount of time and compute, but Bubeck described the costs as “really nothing crazy at all.”

At any rate, this is what AI is good for, and I wonder if, say, it could solve Fermat’s Last Theorem, which took Andrew Wiles eight years of work to solve (he was knighted for it).  And I wonder if there are any seemingly intractable math problems that can’t be solved by AI, especially if they were or will be solved by humans.

Now I don’t think there are any practical implications of this results, but that’s true of much mathematical theory. I’m just amazed at what AI can do.

Piracy: AI company sued for using material from pirated books to train its generative programs

October 3, 2025 • 10:30 am

Well, here’s one of the downsides of AI: a company taking illegal actions to train their AI offerings. If you’ve written a book that was used to train Anthropic’s AI program, and you can find out if this is true was by going to the website below (click on the headline), you stand to gain up to $3,000 per book, providing that your book was copyrighted in the U.S.

Apparently Anthropic downloaded many books from pirated sites, knowing that that was illegal, and then used it to train their AI program.  See the NYT article at bottom for details.

So yesterday a friend sent me this notice, which I had not gotten and had not heard about (click to read).  I want to make authors aware of this as a way of stemming this piracy.

Here is an except from the settlement website:

What is the Settlement About?

This Settlement resolves a class action lawsuit brought against Anthropic over the company’s use of allegedly pirated books to train its AI model.

The plaintiffs claim that Anthropic infringed protected copyrights by downloading books from Library Genesis (LibGen) and Pirate Library Mirror (PiLiMi). Anthropic denies these claims. The Court didn’t decide who was right. Instead, both sides agreed to settle to avoid more litigation.

What is the current status of the Settlement?

The Settlement Administrator is notifying people about the Settlement. Class Members can search for their books on the Works List and file a claim.

On September 25, 2025, the Court granted initial approval of the Settlement. Next, the courts will hold a fairness hearing, resolve any appeals, and make a final decision.

What benefits does the Settlement Provide?

If approved, the Settlement provides a cash payment to Class Members who file a valid and timely claim. The Settlement Fund includes approximately $3,000 per work, before deducting costs, fees, and expenses, as described below.

The Settlement also requires Anthropic to destroy all books that it downloaded from the LibGen or PiLiMi datasets and any copies of those books, subject to Anthropic’s existing legal preservation obligation or obligation pursuant to court order under either U.S. or international law, and then provide written confirmation.

What fees and expenses will be paid from the Settlement Fund?

Under the Settlement, Anthropic has agreed to establish $1.5 billion Settlement Fund. The Settlement Fund will be divided evenly based on the number of works for which valid claims are submitted.

The Settlement Fund will also be used to pay for notice and administrative costs related to the Settlement, attorneys’ fees and expenses, and any service awards for the Class

Of course I checked to see if this was kosher, and many sources verified it. Here’s an article in the NYT. Click to read, or find the article archived here.

An excerpt from the NYT:

In a landmark settlement, Anthropic, a leading artificial intelligence company, has agreed to pay $1.5 billion to a group of authors and publishers after a judge ruled it had illegally downloaded and stored millions of copyrighted books.

The settlement is the largest payout in the history of U.S. copyright cases. Anthropic will pay $3,000 per work to 500,000 authors.

The agreement is a turning point in a continuing battle between A.I. companies and copyright holders that spans more than 40 lawsuits across the country. Experts say the agreement could pave the way for more tech companies to pay rights holders through court decisions and settlements or through licensing fees.

“This is massive,” said Chad Hummel, a trial lawyer with the law firm McKool Smith, who is not involved in the case. “This will cause generative A.I. companies to sit up and take notice.”

The agreement is reminiscent of the early 2000s, when courts ruled that file-sharing services like Napster and Grokster infringed on rights holders by allowing copyrighted songs, movies and other material to be shared free on the internet.

“This is the A.I. industry’s Napster moment,” said Cecilia Ziniti, an intellectual-property lawyer who is now chief executive of the artificial intelligence start-up GC AI.

The settlement came after a ruling in June by Judge William Alsup of the U.S. District Court for the Northern District of California. In a summary judgment, the judge sided with Anthropic, maker of the online chatbot Claude, in significant ways. Most notably, he ruled that when Anthropic acquired copyrighted books legally, the law allowed the company to train A.I. technologies using the books because this transformed them into something new.

. . . . Anthropic had illegally acquired millions of books through online libraries like Library Genesis and Pirate Library Mirror that many tech companies have used to supplement the huge amounts of digital text needed to train A.I. technologies. When Anthropic downloaded these libraries, the judge ruled, its executives knew they contained pirated books.

Anthropic could have purchased the books from many sellers, the judge said, but instead preferred to “steal” them to avoid what the company’s chief executive, Dario Amodei, called “legal/practice/business slog” in court documents. Companies and individuals who willfully infringe on copyright can face significantly higher damages — up to $150,000 per work — than those who are not aware they are breaking the law.

. . .After the judge ruled the authors had cause to take Anthropic to trial over the pirated books, the two sides decided to settle.

“This settlement sends a powerful message to A.I. companies and creators alike that taking copyrighted works from these pirate websites is wrong,” said Justin A. Nelson, a lawyer for the authors who brought the lawsuit against Anthropic.

As part of the settlement, Anthropic said it did not use any pirated works to build A.I. technologies that were publicly released. The settlement also gives any others the right to still sue Anthropic if they believe that the company’s technologies are reproducing their works without proper approval. Anthropic also agreed to delete the pirated works it downloaded and stored.

. . . Even if courts find that training A.I. systems with copyrighted material is fair use, many A.I. companies could be forced to pay rights holders over pirated works because online libraries like Library Genesis and Pirate Library Mirror are widely used, Mr. Hummel said.

It’s dead easy to check by putting in your name or the name of your book(s) at the first site above, and here’s part of what it spat out when I gave it my name.

So I filed a claim for each book. Although Anthropic claims that its piracy is “fair use”, that principle usually applies to using only small bits of works, not entire books—books acquired illegally—to help a company make a profit.  Their lawyers should have told them to just buy the damn books!