Is AI Fear this Century’s Overpopulation Scare?
Is AI Fear this Century’s Overpopulation Scare?
Is Yudkowsky the False Prophet Ehrlich was?
“Sometime in the next 15 years, the end will come.”
Reading this in 2023, you could be forgiven for thinking these words came from an artificial intelligence (AI) alarmist like Eliezer Yudkowsky. Yudkowsky is a self-taught researcher whose predictions about intelligent machines veer into the apocalyptic. “The most likely result of building a superhumanly smart AI… is that literally everyone on Earth will die”, Yudkowsky wrote earlier this week in an article for Time. “We are not prepared. We are not on course to be prepared in any reasonable time window. There is no plan.”
But the words “sometime in the next 15 years, the end will come” are not Yudkowsky’s. They are from a similarly clueless, similarly obstinate prognosticator over fifty years ago: Paul Ehrlich. Ehrlich, a biologist at Stanford, authored The Population Bomb, which argued that Earth’s booming population would spell doom for humanity. (The idea that there are too many people, proven embarrassingly wrong decades ago, is sadly still fashionable in some circles.) In 1968, when Ehrlich wrote The Population Bomb, there were three-and-a-half billion humans. Now, there are over eight billion of us. The famines that Ehrlich predicted never happened – while world population doubled, agricultural efficiency tripled. People are not burdens; we can innovate out of resource scarcity.
“Most of the people who are going to die in the greatest cataclysm in the history of man have already been born.” The words are Ehrlich’s but they could easily have come from Yudkowsky, who once declared that children conceived in 2022 have only “a fair chance” of “living to see kindergarten”. Ehrlich and Yudkowsky are both fatalists. Their prophecies are those of panic and helplessness – the forces of modernity, or capitalism, or natural selection have already spiraled out of control. There’s nothing you can do but wait for death. (Notoriously, the first sentence of The Population Bomb was “the battle to feed all of humanity is over.”) Yudkowsky reckons that for humanity to survive, he would have to be jaw-droppingly wrong. “I basically don’t see on-model hopeful outcomes at this point.”
Luckily for humanity, history will prove Yudkowsky just as wrong as Ehrlich. Nevertheless, his misplaced faith that we’re all going to die has deep roots. Millions of years of evolution causes humans to see impending doom, even when it’s not there. That’s true for religious ends-of-days, or illusory famines, or runaway AI. Yudkowsky’s anxieties aren’t even original: evil machines have been a plot point in sci-fi for over a century.
Judgement Day isn’t coming; Yudkowsky is systematically wrong about the technology. Take his tried-and-true metaphor for superintelligence, chess computers. Computers, famously, have been better at chess than humans for a quarter-century. So, Yudkowsky claims, humans fighting a rogue AI system would be like “a 10-year-old trying to play chess against Stockfish 15 [a powerful chess engine]”. Chess, however, is a bounded, asynchronous, symmetric, and easily simulated game, which an AI system can play against itself millions of times to get better at. The real world is, to put it simply, not like that. (Not to mention Yudkowsky’s assumption that there are no diminishing returns on intelligence as a system gets smarter. Or that such a system is anywhere close to being built.)
AI fearmongering has consequences. Already, luminaries like Elon Musk, Steve Wozniak, and Andrew Yang have signed an open letter which pressures researchers to halt the training of powerful AI systems for six months. (“Should we risk loss of control of our civilization?”, it asks, melodramatically.) This letter is ethically ruinous and those who signed should be ashamed. Suffocating AI won’t save humanity from doom, but it will kill people whose deaths could have been prevented by the AI-assisted discovery of new drugs. Yudkowsky, however, thinks the open letter doesn’t go far enough – he wants governments to “track all GPUs sold” and “destroy… rogue datacenter[s] by airstrike”.
There are grim echoes here. An agonizing number of well-intentioned, logical people take Yudkowsky’s ideas seriously, at least enough to sign the open letter. Fifty years ago, a comparably agonizing number adopted Paul Ehrlich’s ideas about overpopulation, including the governments of the two largest countries on Earth. The result was forced sterilization in India and the one-child policy in China, with backing from international institutions like the World Bank.
Yudkowsky and Ehrlich are people you want to take seriously. They associate with academics, philosophers, and research scientists. They claim to be rational, scientific, unsuperstitious people. They are not. They are fearmongers, devoid of self-awareness and too psychologically and reputationally invested to question their own positions. And they are both wrong. Utterly, utterly wrong.
Paul Ehrlich, now 90, continues to believe his only error was assigning the wrong dates to his predictions. Yudkowsky will never stop catastrophizing, even after he has been proven wrong for decades. Like Ehrlich, he will keep shouting ‘the end is nigh’ long after the crowd has moved on.
“How many years do you have to not have the world end” to realize that “maybe it didn’t end because that reason was wrong?” asks Stewart Brand, a former supporter of Ehrlich, quoted in the New York Times. Anyone predicting catastrophe or collapse should have to answer this question. What is the latest time that, if doom hasn’t happened, Yudkowsky will admit that he was wrong? He won’t ever. For a doomsday cultist, apocalypse lurks around every corner.
Other interesting reads:
 | A guest post by
|
> Chess, however, is a bounded, asynchronous, symmetric, and easily simulated game, which an AI system can play against itself millions of times to get better at. The real world is, to put it simply, not like that.
True. Which is why superhuman chess AI was built decades ago, and a general superintelligence hasn't been built yet. Do you have evidence that AI that deals with messy uncertain reality as opposed to chess is impossible, as opposed to just taking more R&D? Already AI can deal with lots of tasks that are much less of a perfect toy problem than chess.
> (Not to mention Yudkowsky’s assumption that there are no diminishing returns on intelligence as a system gets smarter. Or that such a system is anywhere close to being built.)
Yudkowsky analyses the evolution of hominids and says there doesn't seem to be steep diminishing returns around human level. There are likely to be diminishing returns and physical limits somewhere. The question is whether those kick in before or after AI is smart enough to destroy the world?
The human brain has a bunch of limits imposed by evolution that wouldn't apply to AI (like being small enough to fit through the birth canal, only having 20 watts of power etc.) Not that human brains look like they are near the physical limits of space and power efficiency. I think it is very hard to win against an enemy smarter than you. So if you think humans win, either you are expecting some very sharp diminishing returns just at the level of top humans, or you are expecting humans to hold a massive advantage of some kind. A billion copies of an Einstein smart AI all working flawlessly together towards a common goal is probably already enough to destroy the world.
No. It's the new *nuclear* scare, and Yudkowski is Heinlein, not Elrich. The Population Bomb was always transparently wrong for those who understood Economics and human behavior. But nukes, when they first arrived, looked very plausibly as if they would end the world. It turned out that MAD works, but we got lucky.
There's no equivalent for MAD for AI.