If you were going to predict the emergence of AGI by looking at progress towards it over the past 40 years and extrapolate into the future, then what parameter(s) would you measure and extrapolate?
Kurzweil et al measure raw compute power in flops/$, but as has been much discussed on LessWrong there is more to AI than raw compute power. Another popular approach is to chart progress in terms of the animal kingdom, saying things like “X years ago computers were as smart as jellyfish, now they’re as smart as a mouse, soon we’ll be at human level”, but it’s hard to say whether a computer is “as smart” as some organism, and even harder to extrapolate that sensibly into the future.
What other approaches?
Dislaimer: I’m not saying this is actually a good way to predict when AGI will emerge!
If you were going to predict the emergence of AGI by looking at progress towards it over the past 40 years and extrapolate into the future, then what parameter(s) would you measure and extrapolate?
Kurzweil et al measure raw compute power in flops/$, but as has been much discussed on LessWrong there is more to AI than raw compute power. Another popular approach is to chart progress in terms of the animal kingdom, saying things like “X years ago computers were as smart as jellyfish, now they’re as smart as a mouse, soon we’ll be at human level”, but it’s hard to say whether a computer is “as smart” as some organism, and even harder to extrapolate that sensibly into the future.
What other approaches?
Dislaimer: I’m not saying this is actually a good way to predict when AGI will emerge!