“Computers enable fantasies” – the continued relevance of Weizenbaum’s warnings

Re:“Computers enable fantasies” – the continued relevance of... by modzu on Friday January 27, 2023 @10:33AM (#34547130)

there are a lot of interesting and deep ideas here, about technology in general. but the critics of ai really need to do better. according to this argument, once you understand how your own brain works, you see human intelligence is also just an "illusion". fundamentally though, there is no reason a computer cannot think. any argument against this requires an understanding of the nature of thought and intelligence, in humans, which is quite outside the wheelhouse of most classical computer scientists. for a more refreshing perspective, read anything by hinton

[ Parent | on HN ]

Re:“Computers enable fantasies” – the continued relevance of... by qsort on Friday January 27, 2023 @10:42AM (#34547272)

The more simple counterargument, very well within the wheelhouse of any computer scientist, is that the question is completely irrelevant as it's entirely predicated on one's definition of "intelligence".

Successful software solves a problem correctly and efficiently. It doesn't matter that GCC doesn't compile code like I would if I had to do that by hand, you still end up with compiled code, which was the original problem to solve. It doesn't matter that the sort() routine doesn't sort numbers like I would, it still does.

Playing the imitation game for its own sake is kind of silly.

The "illusion" lies in the attribution to a machine of capabilities it does not have.

[ Parent | on HN ]

Re:“Computers enable fantasies” – the continued relevance of... by modzu on Friday January 27, 2023 @01:34PM (#34549853)

yes, but we havent pulled off this engineering end run yet

[ Parent | on HN ]

Re:“Computers enable fantasies” – the continued relevance of... by AlexandrB on Friday January 27, 2023 @10:52AM (#34547380)

> fundamentally though, there is no reason a computer cannot think.

This is true, but I think there is also a fundamental reason computers cannot understand us. If I tell a future AGI "I am sick" or "I am happy", the AGI may have a conceptual understanding of what I mean, but cannot empathize with my condition. Many human experiences are intimately tied to our bodies and our hormonal systems. How do you explain a concept like "pain" to an intelligence that has no nerves?

[ Parent | on HN ]

Re:“Computers enable fantasies” – the continued relevance of... by noorkersz on Friday January 27, 2023 @10:54AM (#34547404)

by using the cost function somehow??

[ Parent | on HN ]

Re:“Computers enable fantasies” – the continued relevance of... by modzu on Friday January 27, 2023 @12:21PM (#34548776)

its a great point. but then we needn't constrain this problem to computers, there are plenty of human experiences other humans cannot possibly empathize with (what do you really know about ptsd, autism, even different tastes in food or music). i think there must be ways to build something like understanding through analogy and other modalities, accepting that it is different than empathy through shared experience

[ Parent | on HN ]

Re:“Computers enable fantasies” – the continued relevance of... by pixl97 on Friday January 27, 2023 @02:11PM (#34550329)

But that is a different question entirely that is typically called the "AI alignment issue". If something understands us and or is aligned with our needs is completely and totally different if it is intelligent/conscious.

[ Parent | on HN ]