5 ms·
The fatal mistake of this AI cycle was calling LLMs AIs, and publishing impressive chatbots before they were wired up with more useful stuff than passing the Tu
by namaria 2y ago
The fatal mistake of this AI cycle was calling LLMs AIs, and publishing impressive chatbots before they were wired up with more useful stuff than passing the Turing test.
It made some billionaires who will argue it was a tremendous idea. But in the long term I think it will cause another AI winter that will dry up funding for useful research that would take longer to mature. Or maybe it's just like fusion... promising on paper but so incredibly expensive to handle as to render it useless in practice.
- vacuity 2y agoI don't know that an AI winter is bad. Frankly, I don't think society is ready for "actual" AI. It'll probably be used to entrench corporations and increase exploitation if we don't fix our current problems.
- namaria 2y agoThat's assuming 'real AI' is even possible. On a different note, there's an argument dating back to Hobbes' Leviathan that the State and corporations are in fact a form of AI.
- HKH2 2y agoIf you can assume you're talking to a human when talking to an AI then it is real enough.
- namaria 2y agoI strongly disagree. Chat gpt seems to be a great argument against that.
- HKH2 2y agoSo you get tricked into thinking it's a real person, but it's not real enough?
- wkat4242 2y agoI think it is. I'm surprised also we need such extreme resources to do it. After all, nature can do it with 4kg of brain mass and less than 100 Watts. There must be a way we can do this in tech too. But yeah I doubt it would be a benefit for society too.
- namaria 2y agoYeah the thing is, nature can't do it with 4 kg 100 watt human brains. Nature can do it with planetary ecosystems running on supernova detritus bathed in just the right amount of star light. It took hundreds of millions of years just on brain development to get here. The same way that openai did what they did using the whole of common crawl, god knows how many phds, decades of development of basic concepts and several years of datacenters worth of compute. And they got a chatbot.
- nomel 2y agoThe AI is the head of that causal chain. It required all of that, including human brains, to come into existence.
- tech_ken 2y ago> On a different note, there's an argument dating back to Hobbes' Leviathan that the State and corporations are in fact a form of AI Woah neat I've been wondering about this myself, any reading you'd recommend?
- namaria 2y agoThe Cybernetic Hypothesis https://mitpress.mit.edu/9781635900927/the-cybernetic-hypothesis/ https://mitpress.mit.edu/9781635900927/the-cybernetic-hypoth... The Stack: On Software and Sovereignty https://direct.mit.edu/books/monograph/3504/The-StackOn-Software-and-Sovereignty https://direct.mit.edu/books/monograph/3504/The-StackOn-Soft... The Age of Surveillance Capitalism: The Fight for a Human Future at the New Frontier of Power https://www.hbs.edu/faculty/Pages/item.aspx?num=56791 https://www.hbs.edu/faculty/Pages/item.aspx?num=56791 Platform Capitalism https://www.wiley.com/en-in/Platform+Capitalism-p-9781509504909 https://www.wiley.com/en-in/Platform+Capitalism-p-9781509504...
- NemoNobody 2y agoFusion has made leaps recently and is no longer looking as unviable as you suggest.
- namaria 2y agoI'm aware, I follow it closely. It's still not viable tho. I'll be thrilled if it ever becomes viable.
- insane_dreamer 2y ago> The fatal mistake of this AI cycle was calling LLMs AIs, and publishing impressive chatbots before they were wired up with more useful stuff than passing the Turing test LLMs are extremely useful but it's important to recognize the limited use cases: - summarizing information - faster search (summarizes search results for you) - writing in fluent English - translating between common languages - realistic sounding chatbots, customer support - some lower level coding tasks - technical answers from published documentation - RTFM for you None of these are any more "AI" than the previous cycle of "AI" technology