Sorry, Sam and Elon, we have not reached the Singularity
Not that anyone even knows what that means.
Remember “CEO said a thing” journalism?
A lot of CEOs just said a thing:
Don’t take them seriously.1 Partly because, well CEO’s gotta CEO2: In the modern era, making outlandish claims is practically in the job description:
I don’t even know why Sam’s singularity remark (made recently during a podcast) is even news, since Elon has been saying it all along, e.g., in January
and he more or less said it last year, too.
and as far back as 2017, longer than he has been touting fleets of Robotaxis:
Sam, as is often the case, is simply aping Elon. (And for that matter Sam himself said it a year ago, in a self-contradictory blog that never defines the term.)
§
Another reason not to take the whole lot of them seriously is that, so far as I know, none of them has actually defined what they mean. So far as I know nobody other than me even bothered to ask. As Neil Turkewitz put it (replying to me):
§
No matter how you slice it, we just are not actually there yet.
If we were, we wouldn’t be once again seeing what I have been calling the Klarna Effect, the AI hype two-step, in which employers boast of using AI to reduce head count only to quietly rehire humans.
But here we are:
According to the WSJ, “The push to expand head count, at least modestly, is a reversal from the prevailing corporate messaging during much of the AI era. Major employers largely held back on adding people due to economic uncertainties or a belief that artificial intelligence could shoulder more tasks on the job. But some executives say the costs and limitations of AI now demand that more people be added; others want to hire people back following layoffs.”
§
And we certainly haven’t reached the age of abundance that Peter Diamandis and Elon Musk keep insisting is a natural consequence of AI.
§
Nor have we reached a moment in “which technological growth accelerates beyond human control”
We can still perfectly well shut this stuff off if we want to. (It’s mainly because of greed that we don’t, stepping on the gas rather than the brake.)
That much-discussed OpenAI hack of HuggingFace? A human at OpenAI launched it; with a physical firewall it wouldn’t have happened. It wasn’t spontaneous combustion; it was a drill.
Had OpenAI ordinary guardrail classifiers been in place, it wouldn’t have happened. It doesn’t count as “uncontrollable” if you have to give the system a head and shoulders head start. (Yes, we do have poorly controlled AI, but that’s not new, either.)
§
Another definition of the singularity, per a TIME interview with Ray Kurzweil, who is often associated with the term, is “a moment when humans merge with AI via brain-to-computer interfaces to become “much smarter””.
We are clearly not there yet, either.
§
Another notion is that a singularity is a moment in which we can no longer peer forward into the future, but I don’t think we are there yet, either, as an entrepreneur on X pointed out, rightly ripping intp all the recent Singularity talk as nothing more than PR:
§
Going back to a justly famous paper by I. F. Good in 1965, the crux of the Singularity is supposed to be something far beyond any human:
Let an ultraintelligent machine be defined as a machine that can far surpass all the intellectual activities of any man however clever. Since the design of machines is one of these intellectual activities, an ultraintelligent machine could design even better machines; there would then unquestionably be an ‘intelligence explosion’, and the intelligence of man would be left far behind. Thus the first ultraintelligent machine is the last invention that man need ever make, provided that the machine is docile enough to tell us how to keep it under control.
— Speculations Concerning the First Ultraintelligent Machine (1965), IF Good.
In this sense, the singularity is usually supposed to mean a step beyond AGI that can do anything a person, even an expert, can do, and much more.
The idea that we have reached that point is laughable. Here are 10 examples of things that “as good or better than the best human” AGI ought to be able to do, borrowed from the bet I made here 18 months ago with Miles Brundage:
The number of the above that actual current AI can do is still by my count no more than two.
And one could easily come up with many more. Which is part of why a bunch of companies are quietly rehiring.
§
None of this means we won’t someday (most likely not this decade) reach a moment of superintelligence. And if we don’t radically up our game on AI safety, it could be out of our control. Maybe it will help lower prices; maybe it won’t. Brain-computer interfaces will surely improve greatly (though it may take many decades or even centuries to get where Kurzweil dreams of).
But intelligence is multidimensional; we may have made lots of progress on some, and not so much on others. I doubt there will be a single magic moment that our descendants think of as the “Singularity”. If there is, it’s not one that’s going to happen in this decade.
CEOs can say whatever they like. That doesn’t make them true.
P.S. The good news is that you don’t have to be a CEO to make stuff up. Here’s a projection from May that got 1.4 million views that now, two Chinese models and the death of tokenmaxxing later, looks like it has very little chance to be true:
What Demis said – foothills – is safer.
Note that this list of Elon’s failed prediction is incomplete and omits for example Musk’s promises that we would achieve AGI by 2025, which proved to be false, and doesn’t reference pending things like his not-looking-promising prediction that he would have a man on Mars by 2031 in the “worst case”.
Can't learn continuously. Can't adapt autonomously. Can't reliably revise its own beliefs. Can't build durable knowledge from experience. Can't compound intelligence through use. Can't consistently distinguish truth from plausible fiction.
Yet it hallucinates, burns planetary-scale compute, demands constant inference, frequent retraining, external memory, retrieval layers, agent scaffolding, human supervision, and ever-larger infrastructure just to keep improving utility.
Every new capability arrives with another layer of compute, complexity, cost, and babysitting.
And somehow this is the singularity.
Sure. We've reached it.
The first rule of Singularity is: you always talk about Singularity.
The second rule of Singularity is: you ALWAYS talk about Singularity!