Comment by bogzz
1 month ago
> It's the pinnacle of human innovation. It should be revered as our greatest achievement. People should know about how its going to revolutionize scientific research.
How so? Colloquially, AI currently means LLMs. Why would we revere LLMs as our greatest achievement?
Because we've built something that's (functionally) intelligent, comparable to humans in terms of its ability to exhibit (functional) understanding of complex topics, and produce novel correct output. There's nothing even remotely close to this in human history. This was all science fiction 10 years ago.
Maybe the comparison is imperfect, but: we also built a wonderful, novel and excellent bomb, the atom one, not remotely close to whatever humans did before, yet I'm not totally thrilled by it.
People that I speak with often conflate LLMs with AGI (or at least what they thing AGI is).
I'm thinking LLMs lead to an AIG scenario (circa 2008) as opposed to AGI.
Because you can already fill a datacenter with hyper-productive PhD level autodidactic polymaths and we're still on the ground floor of the technology? These frontier models are like alpha builds and they're already ridiculous. AGI is marketing slop and machine learning doesn't need that promise to be the most impressive achievement in human history in my opinion.