Friday, October 3, 2025

Artificial What Now?

Adam Becker's More Everything Forever is a sobering look at our tech overlords, their crazypants dreams, and the reasons that those dreams are less likely than an actual autonomous automobile. It's a pretty depressing books because two things come through. 

One is the enormous power these folks wield over the world we all have to live in; it's power they absolutely believe they should have, based on their certainty that some people are better than others and they are the best of all. 

The other is just how dopey these guys are, how enbubbled and disconnected from-- and even hostile to-- the lives of regular humans. These masters of the universe have all sorts of big dreams, like immortality (really) and not many solid ideas about how to achieve these dreams, even as they ignore many of the counter-ideas (Elon Musk's colonization of Mars? Not going to happen ). 

But what is extra astonishing in the book is that even as they are all-in on a future of AI, especially Artificial General Intelligence, they really don't seem to know what, exactly, that means. AGI? Maybe it means roughly "an artificial machine that can do everything a typical human adult can do" but holy smokes is that vague. 

As a species, we are generally pretty fuzzy on what "intelligence" actually means, with a whole variety of theories about what it is and how it can be measured. And the thing is that these silicon valley overlords seem to know way less about it than people who make even a half-hearted attempt to study this stuff.

Many experts, Becker points out, are certain that the path to AGI does not lie along increased capabilities to current models. They can keep making ChatGPT "smarter," but it will never get any closer to AGI, because that is a difference of kind, not of degree. Check out this piece from Ben Riley in which an AI insider explains that LLMs can't reason like humans

I find the continued attempts to "resurrect" the dead via AI particularly telling. The latest example come in The Atlantic, with multiple attempts to resurrect the dead compared to a sort of Frankenstein complex. It's an apt comparison, as Frankenstein arguably made the mistake of not considering the internal life, the motivations and intents, of his creation. Failing to understand or anticipate those aspects, the doctor rejects the creature that embodies them and creates disaster.

AI creates a variation on that problem. Your dead loved one is not there, the AI completely empty of any motivations or intentions. But for some of these folks, that doesn't seem to matter-- the other "person" is only real to the extent that we experience them. They have no life or existence beyond providing input for our senses; they literally turn off and cease to exist when they are not performing for their maker. 

It is deeply reminiscent of a sociopath's belief that other people are not real, that they exist only as props in a story that is all about MMEEEEEE! And that leads me to wonder if these overlords that Becker describes do not perhaps view actually flesh and blood humans in the same way, and that's why AI seems so human to them-- not because of the depth of humanity in the bots, but the meager view with which they view other humans.

I don't mean to suggest that everyone who gets suckered by a chatbot is a sociopath. But I do think AI moves most easily into places where humanity has been hollowed out, and I wonder if peoples' willingness to imagine that the bot is intelligent, to fill in the blanks of its internal life, isn't one more sign that connection and humanity have been hollowed out in our society in ways that are not good for us.

The quest for Artificial General Intelligence is a chance for us to reflect on what Organic General Intelligence might be. We're often sloppy about our judgment ("People who don't know what I know are dumb") and it's that same sloppiness that leads some folks to assume the AI has any I in it at all, even though AI has no reflection, no intent, no social reconfiguring, no wisdom, no actual knowledge, but just a capability of imitating what an answer to your prompt would, statistically, look like. 

I recommend Becker's book because even though these guys are terrifying in their power and entitlement, it is also useful to understand that they are also clueless about critical factors in their imagined future. It's a reminder that we need not follow these wealthy dopes into their empty, hollow future. 

No comments:

Post a Comment