Since when exactly has artificial intelligence ever implicated the existence of a mind at all? It's always been about knowledge, intuition and the application of such, so of course large language models at least technically count.
If you specifically want a mind, then perhaps you're interested in something more specific than mere artificial intelligence.
Sometimes I see arguments of the form "what does it matter whether or not there is a real mind or mind-like construct, if the output is nonetheless just as comparable?". My answer, I think, is that a more rigorous process gives me greater confidence in the output. I don't consider current LLMs very rigorous by that definition. I end up trying to add most of the rigor myself, and even then I'm not as confident.
This seems to be a very typical fallacy. "AI is just a collection of stuff people made." No. The Internet was that. AI is something that was TRAINED on stuff people made. So are you.
The logical outcome of this fallacy is believing that AI will collapse if it has to train on data it generated. (It won't: a lot of famous AIs were trained this way, including AlphaGo many years ago.)
That’s not at all what the article says and isn’t even what it’s mainly about. It starts by framing AI as a different way of processing data we already have (which is true), and uses that framing to introduce its main point.
The article literally says "the new programs mash up work done by human minds." Implying that they are not doing any real work of their own, just "mashing up" work done by humans.
AI, like automation, means different things and these are often conflated.
AI can mean roughly 'advanced software functionality' but it can also convey the intention of somehow getting output with higher value than the input available. One of them is used in a misleading manner for various reasons.
Automation may mean restricted behavior as in "this contraption will always do Y if X happens" but it can also mean "this contraption will always do the right thing regardless of what you throw at it". Again, one of them is a way to mislead people.
11 comments
[ 0.26 ms ] story [ 8.2 ms ] threadIf you specifically want a mind, then perhaps you're interested in something more specific than mere artificial intelligence.
The logical outcome of this fallacy is believing that AI will collapse if it has to train on data it generated. (It won't: a lot of famous AIs were trained this way, including AlphaGo many years ago.)
AI can mean roughly 'advanced software functionality' but it can also convey the intention of somehow getting output with higher value than the input available. One of them is used in a misleading manner for various reasons.
Automation may mean restricted behavior as in "this contraption will always do Y if X happens" but it can also mean "this contraption will always do the right thing regardless of what you throw at it". Again, one of them is a way to mislead people.