Glenn Reynolds On AI

As usual. Glenn Reynolds shilled his latest New York Post column on Instapundit this morning: What if AIs set out to conquer the world . . . with love? His most unintentionally insightful observation:

[L]et’s face it, you don’t need a 120,000 IQ to fool humans. Or even a 120 one.

If I had to guess, as I've said before, I'd put that man's own IQ at not much higher than 120 himself, allowing for his Yale Law degree and that he passed a bar exam, but in most areas, he's just not that smart. As I pointed out the other day, he believes in Ray Kurzweil's Technological Singularity, in which individual human consciousnesses will upload into a giant computer by 2045 or so (or something like that), but in case that doesn't happen, he has a contract with Alcor to freeze (or "vitrify") his head upon his death, because at some point in the future, "science" will have found a way to resuscitate his brain and bring his body back to life.

Somehow this guy has a prestigious endowed chair at an accredited institition. It certainly is the case that you don't need an IQ much above 120 to fool a lot of people -- including the Post, which runs his columns. He seems to have a strange theory about AI, that it has somehow acquired self-consciousness, agency, or will.

Yesterday I found a video on YouTube, embedded above, in which someone had trained a crow to answer the question, "Where's Walter?" with "I don't know." AI mimics human speech exactly like that crow, it has no idea what it's saying. It's a smart crow, but its IQ probably isn't even as high as Glenn Reynolds's. He says,

Nowadays AI tools like ChatGPT are designed to suck up to their users, congratulating them on their insight and intelligence whether or not there’s evidence of either.

The goal, like that of many media sites, is to maintain user engagement — but the effect has sometimes been to flatter users into terrible decisions, even suicide.

People naturally like to hear they’re smart and insightful, and telling them that they are — especially when they aren’t — is a proven technique of flattery.

I've come to use Chrome AI mode pretty frequently; as I've said, it csn be as helpful as a good reference librarian. For whatever reason, it's never tried to suck up to me. The tone of its answers is distant and blandly polite. I'm not sure what would happen if it tried to chat me up; I would just assume there had been a management decision to treat their users as dumb and sex-crazed, and I'd just go looking for another option. One of the anti-woke standup comic Ben Bankas's routines goes something like this:

People say AI is racist.

Of course it's racist. It has the statistics.

But you can ask it questions about that, and it'll dodge them. "Can you tell me what percentage of violent crimes . . . ?"

All it'll do is say, "Er, ah, go ask ChatGPT."

In other words, both Ben Bankas and his audience are fully aware that in certain areas, AI's putative agency is clearly controlled by humans, and it's policy-driven. In those cases, the fact that AI does nothing but mimic human language like a trained crow and has no independent thoughts is simply clearer than it is at other times. But Reynolds is convinced that AI, a glorifed electronic trained crow, will create romance bots that can convince you they're your ideal girfriend and get you to write your family out of your will, presumably in favor of AI.

Some of these are primarily the AI equivalent of phone sex lines, but others promise much more, offering a responsive partner customized for looks, personality, sense of humor and of course, sexuality.

(One ad I saw even creepily offered the ability to generate a totally photo-realistic avatar from a picture of an ex-girlfriend, while generously allowing you to make her breasts bigger if you so chose. Because of course.)

. . . Imagine if your online “best friend” or “girlfriend” or “boyfriend” is actually a double agent — subtly steering you politically, economically or even spiritually, at the behest of the company or companies behind it.

So which is it, companies or AI? Reynolds isn't completely clear.

I’ve suggested legislation to make AI agents, and by extension the companies behind them, fiduciaries — meaning they would be required by law to put their users’ interests first, as lawyers, corporate directors and executors are.

Hey, there's a bright idea. But why stop with AI? How about, say General Motors? Why should Chevrolet be allowed to sell Corvettes, which make it possible for buyers to drive irresponsibly, or worse, thieves to want to steal them? They should be putting their users' interest first! Forget AI, don't encourage people to buy Corvettes, or plaintiffs' lawyers will sue you! Oh, by the way. Reynolds is a libertarian, except he's got a neat new idea on how the government can control AI.

What he seems to forget is that people have an innate sense of the phony. Paul Fussell observed in Class that we somehow never think that tabloids that report on alien babies are giving actual news. People instinctively recognize the tall tale and allow for it. This seems so far to be how people respond to AI-generated ads: Super Bowl 60 commercials takeaways: AI doesn't fly with savvy viewers:

If the 54 commercials vying for the top spot in USA TODAY’s Ad Meter contest taught us anything, it’s that the 100 million viewers tuned into Super Bowl 60 wanted something real.

. . . And preferably, no artificial intelligence, cryptocurrency or casinos. Americans, it seems, don’t need to be told how to lose their money.

. . . If nothing else, the Super Bowl ads seemed to put in bright lights that anything AI can offer – help with spreadsheets, a tip on how to tie your shoes, whatever – isn’t already available through other means.

But I've got a question for Glenn Reynolds. If he ascribes to Ray Kurzweil's idea that the end state of humanity will be to have individual consciousnesses uploaded into a giant computer memory, how do we know this isn't a nefarious plan generated by AI or the corporations that control AI? Sounds like a plan to rent out computer memory to preserve people's existence forever, for all I can tell, and you, Glenn Reynolds, are shilling for it.