Ask Lemmy

Would you let artificial intelligence take care of your children?

I mean, there are presumably going to be AI systems of a sort that I'd consider capable of reasonably acting in a life-critical role.

If we're talking something like the current 2026 crop of LLMs, like Grok or ChatGPT, then no. Self-driving cars have systems developed using machine learning, and those can operate in limited, specialized life-critical roles in 2026. If you call those AIs, then I'd trust them to drive a kid from Point A to Point B.

Could you have a romantic, platonic or other relationship that imitates interpersonal relationships with artificial intelligence?

I don't see any theoretical reason why it'd be impossible, given the right environment. But I'd be inclined to think that any sort of human-AI relationship will look different from a human-human relationship, just because the kinds of human-human relationships that we invest time into are because we are, well, humans, with everything that comes with that. We can't easily be modified, we invest in building long-term relationships, we are expensive to create, we can't be "rolled back".

A sexy chatbot or something like that, whether-or-not it can be fun, isn't really the same sort of situation that one has with humans. You can erase or tweak the thing at will, just walk away. You can "undo" things that you've said. I doubt that, by-and-large, those properties, which exist now, will be removed from AIs. I'd say that those characteristics necessarily create a different environment, no matter how sophisticated or human-like you could make something, no matter what kind of robotic systems you hook up to them, no matter what visual representation they have, no matter what their speech can sound like.

Further, I'd say that it'd similarly change what human-human relationships looked like if human-human relationships had those properties. What if you could "undo" things that you said to a human partner? Fork them, modify them, and try interacting with them with those modifications? Discard forks freely? I'd think that that'd likely change how we interact with human partners a lot.

With a human partner, you are, to steal a video gaming term, playing on ironman mode, with permadeath, and with a fair bit at stake. That is going to affect how you act. I think that it is unlikely that human-AI relationships will, by-and-large, wind up in that situation, whatever technological developments happen.

Do you believe that artificial intelligence will ever gain consciousness?

To steal a famous quote from Edsger Dijkstra:

The question of whether machines can think is about as relevant as the question of whether submarines can swim.

That is, it's really a matter of definition. Lots of words in the English language haven't been fully-defined. How many grains of sand are there in a "heap"? If one grain isn't a heap, but a lot are, there must be some sort of point where you transition to a heap. But...we can usefully use the word "heap" without ever bothering to specifically nail down any precise definition, because our use doesn't really rely on there being a precise, agreed-upon split between "heap" and "non-heap".

In general, in philosophy, questions about definitions aren't terribly interesting. That is, if I have already defined what "green" means precisely, asking whether-or-not something is green is interesting. But if I haven't, well...asking whether or not a given hue is blue or green (if you even have a language that has the blue-green distinction, which not all do) isn't really all that interesting. You could define it whichever way you want, get people to adopt that convention, and the world would still go on, using that now-nailed-down definition.

I think that we all recognize that there is some level of complexity and sophistication where something is self-aware, and that humans

absent suffering serious brain damage

normally qualify as being self-aware. And that we also think that a four-function calculator is not self-aware. If we were less-and-less complicated, our degree of self-awareness would probably become more-and-more primitive, and at some point, we wouldn't have it. But...as to the question of what level of complexity, what set of features exactly "counts" as self-awareness...shrugs It's like asking what qualifies as a heap. We've never bothered to define it, since it hasn't really been something that we've had to deal with. You could define it wherever you wanted. It's possible that someone could define self-awareness in some sort of way that it'd be possible for it to be impossible for an AI to do. I personally feel like you'd have to go out of your way to do that, but my real point is that I just don't think that the question is all that interesting, because it's really one of definition, of defining something that isn't today defined.

I think that we will, sooner-or-later, achieve computer systems that can increasingly reason and act in ways that today, only humans can do. It is possible that we will never build computer systems that work in exactly the same way that humans do. For example, we could probably build a camera that works more like a human eyeball, but we have digital cameras in 2026 that work on different principles that are preferable for our uses. Aside from pure amusement or academic curiosity, we don't have much reason to make a "human-eye-style camera". But I think that our machines will increasingly be able to do what humans do. I don't think that there is some sort of obvious and intrinsic "unreachable" property of self-awareness that humans have that human-built machines cannot have. As long as it is desirable to build a machine that can think about itself as a thing, which I think it probably is, then I imagine that we will build such a thing sooner-or-later.