This is a really good question but far above Lemmy’s pay-grade man. Neuroscientists are very actively researching and doing debates on this topic (consciousness in AI)
The main problem is that we don’t even have a consensus, fool-proof theory of consciousness… We can barely tell whether a lot of living creatures are conscious or not (think octopus with their distributed intelligence, plants with their chemical signaling, …), and if so, when in their development stage does consciousness arise. Machines are one step removed from even that; even if we give AI models the benefit of the doubt and say they can develop consciousness, how in the world are we supposed to measure it
I found this highly-cited Nature Reviews Neuroscience paper (Seth, A.K., Bayne, T. Theories of consciousness. Nat Rev Neurosci 23, 439–452 (2022). https://doi.org/10.1038/s41583-022-00587-4) (try Internet Archive if you can’t access it) that might be of interest… If you follow the citation trails, there are a ton of new articles on debating consciousness in AI. Like this arxiv article on consciousness in AI
I guess so, if it is true that would mean current AI companies did some serious ethics fuck-ups…
I’m giving current-generation AI the benefit of the doubt though. I personally don’t think the current transformer-based LLM designs can achieve consciousness (although I think machines in general can, just not the current model architectures)
This is a really good question but far above Lemmy’s pay-grade man. Neuroscientists are very actively researching and doing debates on this topic (consciousness in AI)
The main problem is that we don’t even have a consensus, fool-proof theory of consciousness… We can barely tell whether a lot of living creatures are conscious or not (think octopus with their distributed intelligence, plants with their chemical signaling, …), and if so, when in their development stage does consciousness arise. Machines are one step removed from even that; even if we give AI models the benefit of the doubt and say they can develop consciousness, how in the world are we supposed to measure it
I found this highly-cited Nature Reviews Neuroscience paper (Seth, A.K., Bayne, T. Theories of consciousness. Nat Rev Neurosci 23, 439–452 (2022). https://doi.org/10.1038/s41583-022-00587-4) (try Internet Archive if you can’t access it) that might be of interest… If you follow the citation trails, there are a ton of new articles on debating consciousness in AI. Like this arxiv article on consciousness in AI
If AI (LLM) have a consciousness, would that not mean that they only think when they receive input, because they are not computing otherwise?
I guess so, if it is true that would mean current AI companies did some serious ethics fuck-ups…
I’m giving current-generation AI the benefit of the doubt though. I personally don’t think the current transformer-based LLM designs can achieve consciousness (although I think machines in general can, just not the current model architectures)
There is also the question if they can perceive time, or have some kind of emotion (if we assume they are conscious). An interesting topic to ponder.
But I agree, I don’t think that current AI will be conscious, but I don’t see it as impossible just very complex/difficult.