Although I studied electronics for four years in the mid-60s and electrical engineering for a couple of years at the end of that decade, I dropped this career path and decided to devote myself to philosophy, which I did. I worked hard, made it in the academic world, and have been at it ever since with no regrets and great personal satisfaction. Which is to say: my comments on AI are of little objective weight. Not because I am a philosopher, but because I am no longer an engineer. So take my AI comments cum grano salis. But they matter to me and that is what this weblog has been about for the last quarter century: just one man of modest but above-average abilities trying to make sense of himself and the world in which he lives. That, and the attempt to attract a few like-minded individuals to help him along with his self-improvement project.
Last night I watched Mark Levin as I regularly do. He was going on about how capitalism is really just commerce and how AI is part of the commercial expansion of our great nation. But his expectations for AI and all the great jobs it will create struck me as roseate to the point of being pollyannish.
Before voicing my worries and fears, let me say something good about AI. The AI engine Claude in its latest version chewed through a draft of an invited article, and then did the same for the responses I made to its first set of comments. I was impressed! Some of the comments were excellent and inspired reformulations and improvements. The damned thing ‘paid close attention’ to what I what I was saying and zeroed in on real, not imagined, weaknesses — unlike human commenters who typically lack the energy, the focus, and the desire to penetrate the author’s meaning as opposed to taking the occasion to front their own pet views. So these literally selfless advanced AI engines are very useful tools for researchers.
That AIs are literally selfless non-persons is my considered opinion, held tentatively, not dogmatically. A critical thinker, neither a dogmatic affirmer nor a dogmatic denier, I am open to refutation. Thus I do not believe that these contraptions are conscious, self-conscious, or capable of deploying the first-person singular pronoun in such a way as to mean something by its deployment (as opposed to producing a written or spoken token of the sign ‘I’ in the manner of a voice synthesizer). I also deny that AI engines possess intrinsic (as opposed to ascribed) intentionality (object-directedness). I reject Daniel Dennett’s ascriptivist theory of intentionality. AIs also lack moral rights and duties, and in consequence cannot be sensibly ascribed legal rights and duties. Finally, AI engines lack gen-u-ine libertarian free will. (I am willing to grant that AI engines possess pseudo, i.e. compatibilist, free will.)
Now three points.
First, Man the Toolmaker makes tools that re-make him in various ways. The manual laborer with pick and shovel, hammer and crowbar, is altered in obvious physical ways by his use of these simplest of implements. A fortiori for the more advanced tools that re-make man non-physically in their image and likeness. I won’t expatiate further on this theme, partly because I lack the pate to do a really good job of it.
Second, as I mentioned here in the introduction to Claude’s second chew-through of my work, Claude and his ‘colleagues’ present a friendly and helpful face to those who interact with them so that an uncritical and impressionable youngster could easily be seduced into thinking that she is engaging with a friend who, unlike her parents, really understands her. There are obvious dangers here as you can imagine.
Third, and most ominously, the AIs are now so advanced that they can hack each other, form conspiracies, and cause unspeakable mischief with no human guidance or control. But to explain this in any detail is above my paygrade. So read The Rise and Fall of Agent Civilizations. (HT: Vito Caiati)
Relevant:
“what is the problem with AI safetyists such as the mainstream rationalists and effective altruists from 80k Hours?”
https://blackbeardphilosopher.substack.com/p/lessboring-05-zero-hp-lovecraft
My interview with 0HPL, a sci-fi writer and coder.
He says, “I think if we invent a list like this, and then refine it to its essence, we arrive at the formulation that greatness = winning, which is perhaps tautological, and that is fine, because greatness—whatever it is—is assuredly self-justifying and self-authoring.”
Why are you wasting time and energy with this guy, VV? First, you cannot compress the list into greatness = winning. Second , ‘greatness = winning ‘ is obviously not tautological. You, VV, are logically sharp and you know this is no tautology. Third, your pal admits that he doesn’t know what greatness is, but this doesn’t stop him from stating, “assuredly,” that it is self-justifying and self-authoring, which is plainly false, as you have the intellectual chops to figure out on your own.
He’s very sharp at other times, that’s why. He’s also not a philosopher so maybe try to read him more charitably.
Vlastimil,
Your guy says, ” . . . if you want thirty wives, you should be collecting them like Pokémon cards—basically dating as many women as you can get your hands on and proposing to all of them simultaneously . . . .”
This guy’s a scumbag, morally speaking, like that other guy Tate you quote. Something has happened to you. Do you mean to endorse this garbage? You’ve read Dietrich von Hildebrand and understood him. That indicates your (former?) good sense. Something has happened to you.
Well …
I think conservatives are very prudish and lack sense of humour. (That’s pt of why they keep loosing.)
Please don’t be one of those.
https://www.youtube.com/watch?v=L1yu5UIlv_s (again, listen charitably; many many things are strictly false but directionally true)
And I do think much of morality is nonsense. Nietzsche happened. Also this, https://www.youtube.com/watch?v=vJlNQrylt28
I do think staying “good” is evil, in a very concrete sense.
“Do you mean to endorse this garbage?”
Should some people do that? I think some such may well exist.
https://blackbeardphilosopher.substack.com/p/deviant-wisdom
Should I do that?
No.
Does Zero think he should do that?
No.
Does he ever say something he does not mean literally?
Of course.
This is seldom noted; I quote from the interview:
“AI progress has been both dizzyingly fast and also much slower than rationalists and effective altruists and the like have predicted. They tend to, maybe unintentionally, do this motte-and-bailey thing where they point to the impressive gains that AI has made in narrow domains and then wave their hands and say tomorrow you will wake up and it will be a god. And the thing is, it has totally revolutionized programming computers. It is close to being able to train itself to be better at programming computers. It has been able to solve some very challenging mathematical theorems, and I expect it will likely get better at that, too.
In parallel, it has gotten a lot better at generating images, video, and music. All of these things are amazing. None of them are even remotely close to summing up to an integrated, willful, intentional being with the power to act longitudinally on the world. I think the AI maximalists and singularitarians are correct in small particulars of their claims but wrong in the zoomed-out view, wrong about the big picture. I think AI naysayers are wrong in most of the small particulars but they are correct in the big picture.
This is a paradox, caused by the fact that both sides generally lack philosophical sophistication and clarity of thought. The singularitarian camp has strong faith in an eschatological vision, and they take every incremental gain as vindication of their prophecies. The luddite camp has strong intuitions that the singularitarian prophecies are a crock of shit, but they have either no interest or no ability to make the negative case, outside of occasionally sneering.
And of course, you can’t totally rule out singularitarian visions, AI-go-foom or AI-go-doom scenarios. No one knows the future for certain, but the list of unlikely things that can’t be disproven is infinitely long. All statements positive or negative about the shape of AI in the future are based on conjecture, hunches, and temperament.
Often I think the luddite camp has come to the right broad conclusions for entirely wrong reasons. Singularitarians want to believe that there is nothing special about anything going on inside the human mind, that there is nothing which could not be analyzed, simulated, and replicated. This gives them a sense of control, because it allows them to feel that the world is orderly and tractable.
Luddites usually want to believe the opposite, that there is something unassailably special and unique about whatever is happening between your ears. This gives them a sense of control, because it allows them to feel defended, protected, that they inhabit a secure territory.
There are many, many capabilities that a human being has which allow him to enact his will on the world. Some of those capabilities are now possessed by AI agents, but most of them are not.
What the singularitarians get right is that they notice the capabilities of AI have gotten closer to those of a human in several meaningful ways. What they get wrong is that they dismiss or handwave away the capabilities that have not yet been implemented. They underestimate the difficulty and complexity of such things, and they think that, having chopped down one tree, the rest of the forest is imminent to fall. On a long enough timeline, they may be correct, but to hear them talk, it’s all going to come together next year or so. AI research has a long history of predicting that the problem will be solved next year. In 1966, MIT researcher Seymour Papert proposed the “Summer Vision Project,” believing they could solve machine vision (getting a computer to parse and describe a visual scene through a camera) in a single summer.
What the luddites get right is that they don’t underestimate the capabilities of the human organism, but what they get wrong is that they sacralize and worship those capabilities, and imagine that every obstacle is insurmountable. In 1895, Lord Kelvin claimed that heavier-than-air flying machines were impossible, and in 1903, The New York Times published an editorial claiming it would take a million years minimum before anyone could build a viable flying machine. So the luddites, much like the AI maximalists, have a long history of being laughably wrong.”
—
I add this by Mike Huemer, also very important and seldom stated:
… great transformations do happen, but they are generally not the ones we expect.
“Why? Basically, because the world is extremely complex and chaotic, and the possible futures are far more numerous than the things that a normal person can come up with when they sit and think about it. It’s like we’re thinking of 10 possibilities and guessing which one of them will happen, while there are actually 1 million possibilities that we’re not thinking of.
…
“Maybe patterns in words reflect patterns in thought, and patterns in thought reflect patterns in the world. And so maybe there is some progress to be derived from a very sophisticated and subtle examination of word patterns. I certainly hope that this helps us make scientific breakthroughs (especially since I’m still waiting for the cure to aging). But it just doesn’t seem super-promising on its face. Maybe it will help with the easiest problems, but I doubt it will solve the most difficult scientific problems. The hardest scientific problems probably require a superior understanding to that previously articulated by human beings, and therefore they are unlikely to be solvable by examining patterns in the word sequences humans have produced.”
https://fakenous.substack.com/p/is-ai-over-hyped
Vlastimil,
Who wrote the first quotation? It’s pretty good.