Over the last few months, researchers and CEOs have not only shrunk their estimates on when humanity will build super-powerful AI systems, but also seem to have more confidence that these systems will get built.
Anthropic CEO Dario Amodei has said that AI systems that are better than humans at nearly all tasks are just 2-3 years away. He also said that his confidence that such systems will get created has increased over the last six months, and said that AI progress was not hitting a wall.
”There have been maybe five or six times in (the last) 10 years where we thought that we were hitting a wall, where we thought that something didn’t work. And within a few months, something slightly different was invented — either we didn’t actually hit a block or something slightly different was invented that let us get around it. The scaling of AI, it feels like this, this river. Every once in a while it runs into a stone, but it always (finds a way around it),” he told CNBC in an interview.
“Does look like it’s hitting a stone to you right now?” the interviewer asked him. “It absolutely does not,” Amodei replied. “In fact, right now I’m more confident than I have ever been at any previous time that we are very close to powerful capabilities (in) the 10 years that I’ve been working in this field,” he said.
Amodei though said that there could be unforeseen obstacles which could slow down or stop the growth of AI systems. “I’ve always said I don’t know for sure. This is a direction it seems like it’s going in. We could be getting to very powerful human level systems, (but) I still think there’s uncertainty. I think it’s important to be humble, but over the last six months, I would say that uncertainty for me has, has decreased,” he said.
“At some point, we’re going to get to AI systems that are better than almost all humans at almost all tasks. The term I’ve used for it, an essay I recently wrote is a country of geniuses in the data center. It’s a sort of evocative phrase for all the power and all the positive things. You know, all of the potential negative things. That’s the thing that I think we are quite likely to get in the next two or three years,” Amodei said.
“I don’t know exactly when, but the things that I’ve seen inside Anthropic and out of it, over the last few months, lead me to believe that we are on that trajectory,” he added.
Amodei isn’t the only person in the AI space who believes AI systems are set to become incredibly smart. Meta’s Chief AI Scientist Yann LeCun has said that AGI is possible in 5-10 years, while Elon Musk has said that AI will exceed the collective intelligence of humanity by 2030. OpenAI’s Sam Altman, on the other hand, says his company knows what it has to do to create AGI, but it might be hard for humanity to tell when it’s been achieved. And with experts now seemingly more confident about their predictions than before, the world might be set to change in ways that are more dramatic than what most people currently expect.