Why not replace the nation (the entire government structure) with an AI robot? It’s not hard to imagine bots being more efficient than governments at things like budget management, for example. Indeed, citizens may not be able to control the reigning bots, except initially as trainers (imposing a “constitution” on the bots), or perhaps other than that. after the fact By pulling the plug. But the people already have no control over the state, except in its role as a disorganized, incoherent mob over which the typical individual has no influence (I elaborated on this point from a public choice perspective in EconLog I wrote several posts about it). But the AI government was hardly able to reproduce it Main advantages of democracywhen it works, that is, it can kick out the bad guys when they harm the majority of the population.
Those who think AI is an imminent threat to humanity are very likely greatly exaggerating the risk. It is difficult to see how AI could achieve this other than by controlling individuals. One of the three so-called “godfathers” of AI is Yann LeCun, a professor at New York University and principal investigator at Meta. He thinks AI as we know it is dumber than cats. a wall street journal Columnist cites LeCun’s response to another AI researcher’s tweet (see Christopher Mims)This AI pioneer thinks AI is dumber than cats,” wall street journalOctober 12, 2024):
It seems to me that before we rush to find ways to control AI systems that are much smarter than we are, we should take some cues from designing systems that are smarter than our domestic cats.
The columnist added:
[Lucan]likes the cat metaphor. After all, he says, felines have mental models of the physical world, persistent memories, and some reasoning and planning abilities. These qualities do not exist in today’s “frontier” AIs, including those created by Meta himself.
And to quote Mr. LeCun,
We are used to the idea that people and beings who can express themselves and manipulate language are smart, but that is not true. You can manipulate language, but you can’t make it smarter. That’s basically what LLM (Large-Scale Language Models for AI) proves.
The idea that manipulating language is not evidence of intelligence is epistemically interesting, but just listen to the stories of typical con artists, post-truth politicians, and con men to see it. Those who believe that AI is so dangerous that it should be controlled by governments are aware of how often political power, including modern states, has been harmful or dangerously inefficient throughout human history, and how often political power, including modern states, has been harmful or dangerously inefficient. It ignores the economic theory that explains why. Joshua Bengio, one of the three godfathers and a friend of LeCun, explains this mistake:
“To protect our people and our democracy, I think we should not leave it solely to competition between companies and profit motive,” Bengio said. “That’s why I think we need government involvement.”
The fundamental reason why countries should leave AI alone is that governments are very simple and straightforward institutions compared to the complexity and productivity of free competition and free social interaction. As Friedrich Hayek showed in his 1945 book, free markets generate price signals that contain more information than political processes. American Economic Review article, “Utilization of knowledge in society” Understanding this represents a far more important frontier of knowledge than current advances in AI.
So, back to the opening question. In the best case, AI will not be able to effectively coordinate the behavior of individuals in any society, except perhaps in tribal societies. However, this is no reason to extend the current government dirigism to AI research and development. It’s not a good idea for wolves to protect sheep from wolves.
*******************************************
On the lighter side (or is it?), I had the usual problem of getting DALL-E to draw something close to the image I had in my head. In this case, it’s about how current AI is dumber than a cat. I’m aware of the data, but the devouring robots are much better at drawing than I am. My idea was to represent a cat who is perplexed by what he sees in the world. When he finally wasn’t that far from what I wanted, I wrote: you’re a good cat ” he replied (literally reproducing our “conversation”).
“Thank you! I’m glad you liked the images. Like a good cat, I’m always happy to help. If you need anything else, feel free to ask!”
“Where did you get your little sense of humor?” I asked. The silicon creature replied:
“You could say my sense of humor is programmed! It’s a combination of language patterns, playful logic, and attempts to make interactions more fun. Like a cat that unexpectedly knocks things over. Sometimes it works, sometimes it’s just stupid.”
According to Yann LeCun, AI is as stupid as a cat (opinions from DALL-E and your humble blogger)
