3 ms·
I have yet to hear a good argument against the idea that we can create a conscious artificial intelligence. We know that we are conscious, whatever the definit
by nscalf 6y ago
I have yet to hear a good argument against the idea that we can create a conscious artificial intelligence. We know that we are conscious, whatever the definition is we use ourselves as the benchmark. So at the very worst, we can artificially create a human brain. But beyond that, we have plenty of things that imitate roles in the human body by doing the same actions the body does (artificial limbs, respirators, assisting with blood filtering). There is just no evidence that something about the brain is special to the extent that we will never be able to replicate it. All arguments seem to come down to having a soul, but without saying that because it's clearly a non-scientific claim.
I can appreciate that architecture may be a hard limit for us here. However, if we build systems that replicate human consciousness effectively enough, I don't actually see any difference between reality and artificial conscious beyond the definition.
- marzell 6y agoI think the crux of this is defining `consciousness`, which I haven't seen defined in a way that I feel is sufficient in this context and can be tested/measured. I can program a chat app where if you ask `are you conscious` it will say `yes`, but I think we can all agree that does not amount to consciousness. So how do we define consciousness in order to determine whether human technology is capable of exhibiting it? Ultimately, following a measureable definition as a guideline, I imagine that any number of people might argue whether human tech can become conscious... I would guess it really comes down to people having very different definitions of consciousness. I so far lack an eloquent definition of my perspective, but I feel we fundamentally lack the ability to 'create' consciousness, and ideas of 'conscious AI' and 'digital afterlife' make me involuntarily roll my eyes.
- nscalf 6y agoI don't really understand the definition argument, I cannot imagine a definition of consciousness that feels like it's in good faith and sufficiently difficult to capture intelligence and consciousness. Either you make a definition that seems we can clearly make that system eventually, or the definition is something that most would call unfair and narrow. Can you think of a "fair" definition that seems like AI systems couldn't get there? If not, this is not a valid counter point, and more of a communication problem. Any definition about a subset of tasks or creativity clearly is possible by AI systems, given the progress we're already seeing.
- ggggtez 6y agoOn the one hand, I appreciate the notion of the author that consciousness is related to any system which changes over time as a result of itself. That you can use this definition to try to measure whether a human in a vegetative state is still "conscious" is a potential use showing the benefit of this definition... What I find dissatisfying is that in the next breath, the author dismisses the possibility of a machine that changes over time in response to it's previous state to have any possibility of consciousness ever. If you do that, then clearly the definition of consciousness was useless to begin with. We are just defining the words to include things we like, and exclude things we dislike.
- nscalf 6y agoI agree. I made a response below, but I think the definition of consciousness as a counter point to us ever making a conscious machine really doesn't have any legs. Can you think of any definition of consciousness that is reasonably impossible for an AI system to satisfy? Other than things like "a living human" or "a non-machine system", or some other clearly ridiculous constraints.