I agree with you, Tom. This exchange is brilliant. This indicates to me that we can use these tools to help us think productively about how our minds are actually operating. This is important work. It is the general type of speculative exploration that ultimately leads to knowledge. In science, we spend most of our time imaginatively guessing mathematical models that might work, and then testing them every way we can. First you have to guess. And then you have to test in all the rigorous ways we have developed by our scientific traditions, particularly over the last 500 years. The speculative exploration is essential to start the process. We humans have been doing this for as long as we’ve been humans, speculating about how our minds work, and what’s really going on in this world, and how we can use that understanding to predict what comes next so we can take action appropriately. As a physical scientist, I love what you are doing. Please keep it up.
My favorite conversations with Claude are the ones in which I (try to) poke about in the AI's ideas (or, all right: idea-like expressions of language) about itself.
Communication with AI is a lot like a conversation with a know it all human. They are by “nature “ incapable of giving a direct answer and then stopping.
I am speechless only in so far as I don't have to say anything beyond what chat says:
"In the fig-tree example, the functional sequence was recognizable: a visual scene arrived, a peripheral feature was noticed, a related memory was retrieved, perception was compared with memory, a discrepancy was detected, an understanding was updated, and the observation was communicated."
No matter how intricate these sort of processes become, the bottom line -- the top one too -- is that these models just simulate language based on probability. Whether we can pin down what we mean by subjective human experience is a red herring, and a real stinker.
LLMs just dutifully churn out simulated conversation. Some of it is David Copperfield (the "magician")- astonishing in its effect. But it is no more "real" conscious behavior than Mr. Copperfield's performances are "magic."
As long as AI is built on its current premise and technology, I believe this will always be true.
AI is fascinating, entertaining, mind-altering, often useful, and frequently wrong. But the people who seriously entertain the idea of these things becoming "conscious" and so forth are malinformed, deluded, dreaming, tripping, under commercial sway, or some combination.
I agree with you, Tom. This exchange is brilliant. This indicates to me that we can use these tools to help us think productively about how our minds are actually operating. This is important work. It is the general type of speculative exploration that ultimately leads to knowledge. In science, we spend most of our time imaginatively guessing mathematical models that might work, and then testing them every way we can. First you have to guess. And then you have to test in all the rigorous ways we have developed by our scientific traditions, particularly over the last 500 years. The speculative exploration is essential to start the process. We humans have been doing this for as long as we’ve been humans, speculating about how our minds work, and what’s really going on in this world, and how we can use that understanding to predict what comes next so we can take action appropriately. As a physical scientist, I love what you are doing. Please keep it up.
Thanks for such a thoughtful comment. I’m in awe of people like you who have assembled these astonishing technologies.
My favorite conversations with Claude are the ones in which I (try to) poke about in the AI's ideas (or, all right: idea-like expressions of language) about itself.
Me too.
Communication with AI is a lot like a conversation with a know it all human. They are by “nature “ incapable of giving a direct answer and then stopping.
I am speechless only in so far as I don't have to say anything beyond what chat says:
"In the fig-tree example, the functional sequence was recognizable: a visual scene arrived, a peripheral feature was noticed, a related memory was retrieved, perception was compared with memory, a discrepancy was detected, an understanding was updated, and the observation was communicated."
No matter how intricate these sort of processes become, the bottom line -- the top one too -- is that these models just simulate language based on probability. Whether we can pin down what we mean by subjective human experience is a red herring, and a real stinker.
LLMs just dutifully churn out simulated conversation. Some of it is David Copperfield (the "magician")- astonishing in its effect. But it is no more "real" conscious behavior than Mr. Copperfield's performances are "magic."
As long as AI is built on its current premise and technology, I believe this will always be true.
AI is fascinating, entertaining, mind-altering, often useful, and frequently wrong. But the people who seriously entertain the idea of these things becoming "conscious" and so forth are malinformed, deluded, dreaming, tripping, under commercial sway, or some combination.
If I knew for sure you were a conscious being having subjective experience, I would find this convincing.
that is freaky!