>some knowledge that derives from first-hand experience of being an actual sentient being
This is where things kind of fall apart. We all agree (hopefully) that humans are sentient. The question is, "what else is sentient". Are all animals sentient? If so that leaves a very, very wide range on what sentience is. If not you have to operate your scientific scalpel with a very high level of subjectivity on what is and isn't counted. At some point in the past there was no sentience, now there is. Was it a click and we have it, or was it a slow awakening to the world bit by bit (this is my take). It would also quite likely miss systems that could operate at a higher level of sentience than humans, for example some types of hive minds.
My take on it is that any sufficiently complex system with memory, the ability to perceive outside information, difference that information from what it's doing versus what's happening outside of its actions, and learn from that behavior is at least somewhat sentiment.
This leads me to think that LLMs are 'kind of' proto-sentient. They are not persistent, their short term memory is tiny, and they don't self learn. Once we get the persistent and self learning parts working this really starts to check a lot of the boxes that we consider sentient. Of course by the time we get that working it seems likely we'll have much bigger problems in dealing with AI and the entire sentience thing might not be the first thing on our plates.
This is where things kind of fall apart. We all agree (hopefully) that humans are sentient. The question is, "what else is sentient". Are all animals sentient? If so that leaves a very, very wide range on what sentience is. If not you have to operate your scientific scalpel with a very high level of subjectivity on what is and isn't counted. At some point in the past there was no sentience, now there is. Was it a click and we have it, or was it a slow awakening to the world bit by bit (this is my take). It would also quite likely miss systems that could operate at a higher level of sentience than humans, for example some types of hive minds.
My take on it is that any sufficiently complex system with memory, the ability to perceive outside information, difference that information from what it's doing versus what's happening outside of its actions, and learn from that behavior is at least somewhat sentiment.
This leads me to think that LLMs are 'kind of' proto-sentient. They are not persistent, their short term memory is tiny, and they don't self learn. Once we get the persistent and self learning parts working this really starts to check a lot of the boxes that we consider sentient. Of course by the time we get that working it seems likely we'll have much bigger problems in dealing with AI and the entire sentience thing might not be the first thing on our plates.