Suffering is Real. AI Consciousness is Not. | TechPolicy.Press
Suffering is Real. AI Consciousness is Not. | TechPolicy.Press
Probabilistic generalizations based on internet content are not steps toward algorithmic moral personhood, write David McNeill and Emily Tucker.
Could just say:
If you accept either privacy of consciousness or phenomenal transparency then philosophical zombies must be conceivable and therefore physicalism is wrong and you can’t engineer consciousness by mimicking brain states.
Edit:
I guess I should've expected this, but I'm glad to see multiple people wanted to dive deep into this!
I don't have the expertise or time to truly do it justice myself, so if you want to go deep on this topic I'm going to recommend my favorite communicator on non-materialist (but also non-religious) takes on consciousness, Emerson Green:
- The Hidden Mind: For Privacy and Against Illusionism
- Zombies
- Phenomenal Transparency
- Interview with Philip Goff, talking about all of the above
But I'm trying to give a tl;dl to the first few replies at least.
Why does
privacy of consciousnessmean one can't engineer consciousness by mimicking states of the organ that probably has something to do with it?What does phenomenal transparency mean?
Plus, privacy of consciousness may just be a technological hurdle.
I added some episodes of Walden Pod to my comment, so check those out if you wanna go deeper, but I'll still give a tl;dl here.
Privacy of consciousness is simply that there's a permanent asymmetry of how well you can know your own mind vs. the minds of others, no matter how sophisticated you get with physical tools. You will always have a different level of doubt about the sentience of others, compared to your own sentience.
Phenomenal transparency is the idea that your internal experiences (like what pain feels like) are "transparent", where transparency means you can fully understand something's nature through cognition alone and not needing to measure anything in the physical world to complete your understanding. For example, the concept of a triangle or that 2+2=4 are transparent. Water is opaque, because you have to inspect it with material tools to understand the nature of what you're referring to.
You probably immediately have some questions or objections, and that's where I'll encourage you to check out those episodes. There's a good reason they're longer than 5 sentences.
Lol. This comment sent me down a rabbit hole. I still don't know if it's logically correct from a non-physicalist POV, but I did come to the conclusion that I lean toward eliminative materialism and illusionism. Now I don't have to think about consciousness anymore because it's just a trick our brains play on us (consciousness always seemed poorly defined to me anyways).
I guess when AI appears to be sufficiently human or animal-like in its cognitive abilities and emotions, I'll start worrying about its suffering.
If you wanna continue down the rabbit hole, I added some good stuff to my original comment. But if you're leaning towards epiphenomenalism, might I recommend this one: https://audioboom.com/posts/8389860-71-against-epiphenomenalism
Edit: I thought of another couple of things for this comment.
You mentioned consciousness not being well-defined. It actually is, and the go-to definition is from 1974. Nagel’s “What Is It Like to Be a Bat?”
It’s a pretty easy read, as are all of the essays in his book Mortal Questions, so if you have a mild interest in this stuff you might enjoy that book.
Very Bad Wizards has at least one episode on it, too. (Link tbd)
Speaking of Very Bad Wizards, they have an episode about sex robots (link tbd) where (IIRC) they talk about the moral problems with having a convincing human replica that can’t actually consent, and that doesn’t even require bringing consciousness into the argument.
These grifters would rather cry into their computers than oppose any actual suffering.