4 ms·
I would argue that ChatGPT has reached a certain level of understanding about what it's saying. That's because you can ask it questions about what it says, and
by jasfi 4y ago
I would argue that ChatGPT has reached a certain level of understanding about what it's saying. That's because you can ask it questions about what it says, and it can continue to reason along the line of what it's previously said. It does sometimes make mistakes in this, but this is improving. It's just that people want understanding to look a certain way, that seems more familiar to us.
- jmfldn 4y agoI think you are right in a sense. Whether ChatGPT 'understands' might be moot if it produces responses to that effect. Understanding to me implies perceiving the intended meaning, which to me implies sentience. Does it matter though if it doesn't have this property? As the aphorism goes (paraphrasing) "asking whether machines think is like asking whether submarines swim", ie. it's kind of beside the point. If you get the result you want, in this case understanding-like responses, then that's enough right?