Comment by fragmede
1 year ago
Even if you don't buy that LLMs and the transformer architecture in particular will lead to AGI, and then artificial super intelligence (ASI), the quest to try and make ASI is far more significant than anything that's ever come before. ASI would be the last thing that humans need to invent.
We'll see if it comes to anything, first.
Until then, it doesn't have the property of being significant.
I mean it’s possible we don’t see it. Or perhaps don’t see it very much before it’s ridiculously smart. A lot of AI research is at private companies who only share some details when they formally release a model. If they could make an AI researcher AI, they may be inclined to keep it to themselves for a bit while they let it improve itself for a few generations.