Sutton’s bitter lesson is that figuring out how to leverage computation using general methods that scale with more computation beats trying to perform a task by encoding human-learned specific knowledge about the task domain. You still have to come up with the general methods.
I’ve had some discussion with Tsvi about why I think AGI comes soon. I think the points made in this post are mostly correct.
My prediction is that neuroscience and ml research are close to discovering a new general method, more similar to the one the brain uses. I expect that once this new general method is discovered, true AGI will be unlocked. Perhaps this new method will work alone, or perhaps it will work in combination with existing methods.
Meanwhile, while current LLMs aren’t true AGI, they are becoming increasingly helpful in speeding up coding and analyzing large swathes of academic papers. I think these contribute to hastening the coming of a new general method.
I’ve had some discussion with Tsvi about why I think AGI comes soon. I think the points made in this post are mostly correct.
My prediction is that neuroscience and ml research are close to discovering a new general method, more similar to the one the brain uses. I expect that once this new general method is discovered, true AGI will be unlocked. Perhaps this new method will work alone, or perhaps it will work in combination with existing methods.
Meanwhile, while current LLMs aren’t true AGI, they are becoming increasingly helpful in speeding up coding and analyzing large swathes of academic papers. I think these contribute to hastening the coming of a new general method.