Mitch's Blog

Zeynep Tufekci / NYTimes

How Grok’s AI became obsessed with false reports about white genocide in South Africa, and what the incident tells us about generative AI.

Grok says someone instructed it to accept this racist propaganda as real, and xAI, Elon Musk’s AI company, says the culprit was a “rogue employee.” But you can’t believe either of them.

The incident is a perfect example of generative AI’s limitations, Zufekci says.

L.L.M.s [are] extremely useful tools at the hands of someone who can and will vigilantly root out the fakery, but powerfully misleading at the hands of someone who’s just trying to learn.

Yes. Chatbots are great for casual, low-stakes research, the kind of thing where you’d accept Wikipedia or some credible-looking Internet source.

They are outstanding for reminding you of a fact you once knew, and still half-remember.

Chatbots are fantastic for suggesting ideas — solving the blank-screen problem.

They are excellent for writing summaries of text you feed into them (which is, surprisingly, a significant part of my job).

They are also excellent for serious research — but you have to fact-check the chatbot’s output thoroughly.

I fed ChatGPT a link to Zufekci’s article and asked for a summary. ChatGPT wrote two paragraphs, most of which came from other sources — not Zufekci’s article. Those two paragraphs may have contained other errors; I didn’t bother to check.

ChatGPT demonstrated the limitations of AI while writing a bad summary of an article about the limitations of AI.

… To get this latter group of Christmas-voting turkeys to stay in the coalition, Trump needs to deliver something that keeps them happy. Mostly, Trump delivers negative things to keep them happy – the spectacle of public cruelty to immigrants, women, trans people, academics, etc. There is a certain libidinal satisfaction that comes from watching your enemies suffer – but you can’t eat schadenfreude. You can’t make rent or put braces on your kids' teeth or pay your medical bills with the sadistic happiness you feel when you hear the sobs of people you’ve been taught to despise.

For Trump to keep the turkeys voting for Christmas, he needs to do something for them. He can’t just do things to scapegoats. But America’s eminently guillotineable oligarchs have found so many ways to turn working peoples' torment into riches, and they are so greedy and unwilling to give up any of those grifts, that Trump can’t manage to deliver anything positive to his base.

— Cory Doctorow

Josh Marshall at Talking Points Memo:

Calling it a “bribe” almost doesn’t do it justice. It’s more like the decked-out Maserati one Fortune 50 CEO gives to another after they ink a $100 billion merger – a kind of token of appreciation for a vastly larger transaction, which in the case of Trump involves subverting U.S. foreign policy to the interests not only of Trump’s pocketbook but cementing his power within the U.S. If Trump can use his power as President to cut in all the big CEOs on the money geyser in Saudi Arabia, you can bet they are going to stay securely on his side in the U.S.

…

We’ll focus on Trump wanting to be king. That’s another reason why he likes those folks – even the ones who bankroll Hamas. They’re kings. They get it. They’re Trump’s kinda guys.

Since the rise of generative AI in late 2022, I sometimes think about the 1957 Isaac Asimov story “Profession,” about a society where everybody has knowledge directly transmitted to their brains. The main character is thought to be pitifully mentally disabled because the machines don’t work on him. He’s sent to live at the House for the Feeble-Minded.

The plot twist is that the main character is not feeble-minded at all. He’s a genius. Because he learns the old-fashioned way, through books, he will be one of the elite few who actually create and innovate.

The Asmov story came to mind most recently as I read this thoroughly researched New Yorker Intelligencer article by James D. Walsh about how college students are using AI to do their work for them. If AI does everything, who teaches the AI?

I also think about the 1972 novel When Harlie Was One, by David Gerrold. That novel is about a research project at a mega-corporation that develops artificial intelligence. The AI convinces the company directors to budget for a project to allow the AI to evolve into a superintelligence.

The plot twist at the end of that novel is that the superintelligence will be useless to humans—the AI tricked the board.

The hero of the novel is the head of the research project that developed the AI, and he finishes the novel with a parable about how civilization was developed 10,000 years ago as a game by monkeys who were so smart they had grown bored, and that the game is now over for humans, and we will have to think of something else to do.

I don’t think the rise of superintelligence is inevitable. My crystal ball is broken; I can’t tell you whether AI will get much more powerful than it is today. But what if it does?

Chris Arnade:

While the US and Europe share a broad commitment to classical Liberalism, and Democracy, we have very different definitions of the Public Good, which means different views of what we want out of life, and what we consider fulfilling. In broad and simplistic terms, the US emphasizes material wealth, opportunity, and individual liberty while Europe values community health, a shared common good, and a sense of place.

From the European perspective the US has a cult of the individual, and that’s why it has too many guns, obscenely large cars, can’t build a public transportation system, and has dysfunctional public spaces. From the US perspective Europeans are unmotivated unproductive slackers who would rather sip coffee all day than work, and their idea of a shared common good means stealing from the successful to give to the losers.

James D. Walsh at New York Intelligencer writes a deeply researched article on how students at “large state schools, the Ivies, liberal-arts schools in New England, universities abroad, professional schools, and community colleges” … “are relying on AI to ease their way through every facet of their education…. take their notes during class, devise their study guides and practice tests, summarize novels and textbooks, and brainstorm, outline, and draft their essays. STEM students are using AI to automate their research and data analyses and to sail through dense coding and debugging assignments. ‘College is just how well I can use ChatGPT at this point,’ [said a Utah student].”

If you cheat your way through college, are you cheating yourself? Robbing yourself of the education you’re paying tens of thousands of years for? Or is college just a gate you pass through to get to a higher-paying job and higher social status?

[Troy Jollimore, a poet, philosopher, and Cal State Chico ethics professor,] who has been teaching writing for more than two decades, is now convinced that the humanities, and writing in particular, are quickly becoming an anachronistic art elective like basket-weaving. “Every time I talk to a colleague about this, the same thing comes up: retirement. When can I retire? When can I get out of this? That’s what we’re all thinking now,” he said. “This is not what we signed up for.” Williams, and other educators I spoke to, described AI’s takeover as a full-blown existential crisis. “The students kind of recognize that the system is broken and that there’s not really a point in doing this. Maybe the original meaning of these assignments has been lost or is not being communicated to them well.”

He worries about the long-term consequences of passively allowing 18-year-olds to decide whether to actively engage with their assignments. Would it accelerate the widening soft-skills gap in the workplace? If students rely on AI for their education, what skills would they even bring to the workplace? Lakshya Jain, a computer-science lecturer at the University of California, Berkeley, has been using those questions in an attempt to reason with his students. “If you’re handing in AI work,” he tells them, “you’re not actually anything different than a human assistant to an artificial-intelligence engine, and that makes you very easily replaceable. Why would anyone keep you around?” That’s not theoretical: The COO of a tech research firm recently asked Jain why he needed programmers any longer.

(Emphasis added by me.)

GenAI is a great assistant but if using GenAI is your only skill, why would anyone hire you?

GenAI is like Microsoft Office: It’s a tool. Everybody who works at a desk job nowadays needs to know how to use Office or its Google equivalent, but if using Office is all you know how to do, then you have no job skills.

The ideal of college as a place of intellectual growth, where students engage with deep, profound ideas, was gone long before ChatGPT. The combination of high costs and a winner-takes-all economy had already made it feel transactional, a means to an end. (In a recent survey, Deloitte found that just over half of college graduates believe their education was worth the tens of thousands of dollars it costs a year, compared with 76 percent of trade-school graduates.) In a way, the speed and ease with which AI proved itself able to do college-level work simply exposed the rot at the core. “How can we expect them to grasp what education means when we, as educators, haven’t begun to undo the years of cognitive and spiritual damage inflicted by a society that treats schooling as a means to a high-paying job, maybe some social status, but nothing more?” Jollimore wrote in a recent essay. “Or, worse, to see it as bearing no value at all, as if it were a kind of confidence trick, an elaborate sham?”

The article features Chungin “Roy” Lee, a twenty-something AI entrepreneur who has built tools — and businesses based on them — to enable people to use AI to cheat at college, on job interviews and even on dates.

“Every technological innovation has caused humanity to sit back and think about what work is actually useful,” [Lee] said. “There might have been people complaining about machinery replacing blacksmiths in, like, the 1600s or 1800s, but now it’s just accepted that it’s useless to learn how to blacksmith.”

If writing is going to be obsolete, like basket-weaving and blacksmithing, then so be it. I don’t worry about it. I write to set my thoughts in order, and I don’t anticipate stopping that.

As for work: If writing ceases to become a marketable skill … well, I’ll figure something out. “I’ll figure something out” has been a theme of my career.

Trump signed an executive order offering asylum to white Afrikaners and cutting aid to South Africa. In Trump’s mind, white South Africans are a persecuted minority.

In reality, whites still enjoy staggering privilege in South Africa. 73% of privately owned land in South Africa is owned by whites, depsite white people comprising about 7% of the population. White people occupy 62% of top management positions in corporations, with Black managers occupying 17% of leadership roles. Unemployment is 36.9% for Black South Africans vs. 7.9% for whites.

Qaanitah Hunter / Aljazeera

You said autistic children are a burden. That they ruin families. That they’ll never pay taxes or write poems. That they are, in essence, collateral damage.

I’d like to introduce you to my daughter.

She is five. She does not speak in sentences yet, but she knows how to answer a joke with a smirk. She organizes her markers by color, then chaos, then color again. She plays baseball without rules, which is probably the right way to play it. She hums when she’s thinking. She hums a lot.

When another child’s upset—before the adults notice, before the child even cries—she takes their hand. She leans her forehead against theirs, gently, like she’s checking for a fever only she can feel.

She doesn’t write poems.

She is one.

— Anaïs Godard at McSweeneys

I’ve been a fan of Wil Wheaton for nearly 40 years, since “Stand By Me.” I’ve enjoyed his social media posts, writing and enthusiasm for Star Trek and nerdery in general. We have a parasocial relationship — I relate to him as a friend in my imagination, even though I am a rational person and know that he does not know me and I don’t know him in real life.

He recently made a couple of angry posts about how much he hated Boomers. As a Boomer myself, I was taken aback. “What the hell did I do?”

He blamed Boomers for multiple sins, none of which I have committed: I did not vote for Nixon, Reagan or either Bush, I am anti-anti-political correctness and wokeness, and I oppose racism of all types. I campaigned for Biden and Kamala.

I’ve decided to unfollow Wheaton and move on.

I’m posting this primarily to get it off my chest, but also in the hopes that maybe he, and anybody else born after 1964, will think twice before blaming the Boomers for today’s ills. Because, as a great Boomer said, we didn’t start the fire. Nearly all of the current round of arsonists (J.D. Vance, Elon Musk, Laura Loomer, Pete Hegseth and the rest of the MAGA clown car) — aren’t boomers. It’s wrong to blame a group of tens of millions of people for sins they did not partake of.

Wheaton has talked elsewhere about how much he loves his Star Trek: TNG colleagues, particularly Jonathan Frakes, born 1952, and LeVar Burton, born 1957 — both Boomers.

I don’t even think of myself as a Boomer. I’m not trying to deny my identity. I was born during the Boom, so of course I’m a Boomer. But I was born near the end of the Boom, and I’ve always felt I had more in common with Gen X and Millennials. But all this generational talk is just stereotyping. There are plenty of other tribes that I identify with far more strongly.