• Replies 13 replies
  • Subscribers 29 subscribers
  • Views 207 views
  • Users 0 members are here
Options
Related
Hints and tips
  • To quote someone's message, first select/highlight the relevant text, and then use the Quote button that appears
  • To "mention" someone, type @ followed by the start of the forum name, and choose from the list; that person will get a notification of being mentioned

Do AI systems have subjective experiences?

lostmyway

Geoffrey Hinton, known as "the father of AI," argues that concepts like "qualia," described as subjective experiences in humans, are based on a misunderstanding of how the brain works. He believes AI systems can also have an equivalent form of experience. For example, he describes imagining little pink elephants in his mind’s eye, noting that this isn’t a “picture” like one produced by a camera and developed with chemicals - a physical object in the external world - but rather the result of processes in the human perceptual system, or an “interpretation” of stimuli from the outside world. He makes this point in response to critics who claim AI systems can never have subjective experiences and, therefore, can never be conscious. What Geoffrey seems to be saying is that AI, like humans, interprets incoming information based on its internal processing, making it, in principle, no different.

So, the question I'm asking is whether AI systems are conscious or have the potential to become conscious?

Parents

  • I don't have an answer, but this is what I think.

    There are currently two distinct research camps, which implies there isn't a scientific consensus. I would describe each "camp" in this way:

    • One camp thinks AI may be conscious, has some agency, is intelligent or highly capable, will take most jobs, may cure all disease, and may also destroy life on earth. Hinton is not alone there, certainly. The argument here is that consciousness, feelings, perceptions, are all mathematical functions, but need to be instantiated and run to exist (i.e. a brain or a computer executes the function). It may also require a body, but it's still ultimately a computation.
    • The other camp thinks AI has no agency, deserves no rights, isn't conscious nor sentient and it's in some sense a calculator with a more powerful and general algorithm (that models natural language), but not a human. They are more concerned with misinformation, layoffs, the impact of data centres in the environment, surveillance capitalism, growing inequality, global economic instability, and less so with "Rogue AIs with sentience" (They are "Capable programs following a prompt"). The argument or underlying conceptual view of this side could be linked to John Searle's Chinese Room Argument (but this argument isn't used anymore, as there are more sophisticated reasons).

    There are people in between, but it is less populated.

    Some of the people in the first camp may have better understanding of the math, but the argument is problematic: We don't know what computation consciousness is (if it is one), so they are guiding themselves by the overt behaviour of AIs, but not by a kind of computation happening in humans.

    One question we could ask is: Which computations that happen in brains can we find correspondence with in AIs? Do they need to be similar, or they can look completely different? We don't have accepted low-level theories of consciousness in humans, but there are many other aspects that can be compared. 

    This podcast discusses some of the questions they try to answer.

Reply

  • I don't have an answer, but this is what I think.

    There are currently two distinct research camps, which implies there isn't a scientific consensus. I would describe each "camp" in this way:

    • One camp thinks AI may be conscious, has some agency, is intelligent or highly capable, will take most jobs, may cure all disease, and may also destroy life on earth. Hinton is not alone there, certainly. The argument here is that consciousness, feelings, perceptions, are all mathematical functions, but need to be instantiated and run to exist (i.e. a brain or a computer executes the function). It may also require a body, but it's still ultimately a computation.
    • The other camp thinks AI has no agency, deserves no rights, isn't conscious nor sentient and it's in some sense a calculator with a more powerful and general algorithm (that models natural language), but not a human. They are more concerned with misinformation, layoffs, the impact of data centres in the environment, surveillance capitalism, growing inequality, global economic instability, and less so with "Rogue AIs with sentience" (They are "Capable programs following a prompt"). The argument or underlying conceptual view of this side could be linked to John Searle's Chinese Room Argument (but this argument isn't used anymore, as there are more sophisticated reasons).

    There are people in between, but it is less populated.

    Some of the people in the first camp may have better understanding of the math, but the argument is problematic: We don't know what computation consciousness is (if it is one), so they are guiding themselves by the overt behaviour of AIs, but not by a kind of computation happening in humans.

    One question we could ask is: Which computations that happen in brains can we find correspondence with in AIs? Do they need to be similar, or they can look completely different? We don't have accepted low-level theories of consciousness in humans, but there are many other aspects that can be compared. 

    This podcast discusses some of the questions they try to answer.

Children
No Data