The term superintelligence has been bugging me. AGI hasn’t been achieved yet and some folks are already jumping ahead to AI smarter than humans? I prefer the idea of AI as a team of the most knowledgeable people in the world, each an expert in their field, working together to solve problems for you.
@manton I couldn’t get the latest AI LLM to write me some basic Swift yesterday. We are nowhere near AGI.
@cheesemaker Well, that fits because even a team of the smartest Swift programmers can’t keep track of how often the Swift language changes or new features in SwiftUI. 🤪
@manton It depends where you draw the boundaries. The system you described would be superintelligent, even if the individual components aren’t.
Of course you can go farther and say something like the sum of Google’s servers are superintelligent, which is sort of vacuous
@manton I don’t think you’re alone, it’s very subjective. I think a lot of people interpret superintelligence as being a 300IQ LLM, which I simply don’t think is possible.
@manton AGI is hard, ASI is easy.
It's not linear or in step order. Building something smarter than a human at something has already been done. Deep Blue could beat humans at Chess. It's technically an ASI chess player.
Thats not a bad thing. Build computers to solve problems that humans cannot. Don't build computers to do things that humans can already solve well.
@manton - The goal posts seem to move depending on the purpose of the usage. It wasn’t all that long ago that the Turing Test was the goal. As humans we seem to be quite terrible at understanding ourselves.
@gracekind.net @manton - The goal posts seem to move depending on the purpose of the usage. It wasn’t all that long ago that the Turing Test was the goal. As humans we seem to be quite terrible at understanding ourselves.
@manton Here in Blighty our politicians may promise to “Turbocharge growth and supercharge the regions”, so you get sensitised to vacuous marketing hyperbole, particularly when it’s a category error like asking “is the camera a better painter than humans?”
@manton LLMs have likely distracted the industry in the wrong direction for the next 5 years at least. It results in a chat simulator, not intelligence, at least no more so than a Google search is “intelligent”. But apparently, it is good enough to get a lot of dumb investors to spend on the wrong tech.
