There can be long periods of minimal quantitative progress, even quantitative regression, before a massive qualitative leap that's the result of making a few small changes.
Planes either fly or fall.
The difference is that a super-intelligent AGI could provide immense value in secret. It's somewhat hard to keep a plane secret. It's value is largely in moving people, so people have to know about it and they fly right there in the sky for everyone to see. Also, massive rockets are hard to hide, and a lot of the value was publicity.
I expect that somebody somewhere is going to have AGI for a period of time when the rest of the world doesn't know they have it... and right now feels like that time. People are openly publishing LLMs that float awfully close to those motorized gliders that decidedly, definitely didn't fly... until suddenly they did. It's only a matter of time until someone figures out how to make these fly, and the question isn't when, but when we'll find out.
There can be long periods of minimal quantitative progress, even quantitative regression, before a massive qualitative leap that's the result of making a few small changes.
Planes either fly or fall.
The difference is that a super-intelligent AGI could provide immense value in secret. It's somewhat hard to keep a plane secret. It's value is largely in moving people, so people have to know about it and they fly right there in the sky for everyone to see. Also, massive rockets are hard to hide, and a lot of the value was publicity.
I expect that somebody somewhere is going to have AGI for a period of time when the rest of the world doesn't know they have it... and right now feels like that time. People are openly publishing LLMs that float awfully close to those motorized gliders that decidedly, definitely didn't fly... until suddenly they did. It's only a matter of time until someone figures out how to make these fly, and the question isn't when, but when we'll find out.