4 ms·
The potential miscalculation is thinking deep neural nets will scale to AGI. There are also a lot of misnomers in the area, even the term "AI", is claiming syst
by proc0 3y ago
The potential miscalculation is thinking deep neural nets will scale to AGI. There are also a lot of misnomers in the area, even the term "AI", is claiming systems are intelligent, but that word implies intelligibility or human level understanding, which it is nowhere near as evidence by the existence of prompt engineering (which would not be needed otherwise). AI is ripe with overloaded terminology that prematurely anthropomorphizes what are basically smart tools, which are smart thanks to the brute-forcing power of modern GPUs.
It is good to get ahead of the curve, but there is also a lot of hype and overloaded terminology that is fueling the fear.
- atleastoptimal 3y agoWhy couldn't deep neural nets scale to AGI. What is fundamentally impossible for neural nets + tooling to accomplish the suite of tasks we consider AGI? Also prompt engineering works for human too. It's called rhetoric, writing, persuasion, etc. Just because the intelligence of LLM's is different than humans doesn't mean it isn't a form of intelligence.
- KingMob 3y ago> Why couldn't deep neural nets scale to AGI Speaking as a former cognitive neuroscientist, our current NN models are large, but simpler in design relative to biological brains. I personally suspect that matters, and that AI researchers will need more heterogeneous designs to make that qualitative leap.
- ryanklee 3y agoSeems like that's happening with mixture of experts already. Not sure you are presenting an inherent LLM barrier.
- Vecr 3y agoHinton's done work on neural nets that are more similar to human brains and no far it's been a waste of compute. Multiplying matrices is more efficient than a physics simulation that stalls out all the pipelines.
- KingMob 3y agoDoing the wrong thing more efficiently doesn't make it into the right thing.
- proc0 3y agoFair point. I guess nobody knows yet, and it's also worth a shot. In the context of AI alignment, I don't see strong evidence to suggest deep neural nets, transformers, and LLMs have any of the fundamental features of intelligence that even small mammals like rats have. ChatGPT was trained on data that would take a human several lifetimes to learn, yet it still makes some rudimentary mistakes. I don't think more data would suddenly manifest all the nuance of human intelligence... that said we could just be one breakthrough away from discovering some principle or architecture, although I think that will be a theory first and not so much a large system that suddenly wakes up and has agency.
- arisAlexis 3y agocertainly there is no need whatsoever for AGI to exist in order for an autonomous agent with alien/inhuman intelligence or narrow capabilities to turn our world upside down
- proc0 3y agoThat's true, although it's much less substantiated because we have no idea what form that would take and how much effort is needed to get there. So far we have no evidence that LLMs, transformers or any other NN architecture have some of the fundamental features of human intelligence. From my own perspective, one of the first features we will need to see to be on the right track is self-supervised learning (with very minimal initial conditions). This seems inherent to all mammals yet the best AI so far requires huge amounts of data to even get coherent outputs.