When I think of the glasses/face recognition example, though, it makes me wonder what it is about a cheerio’s box that makes them know there is a cheerio’s box. How strange and misshapen a thing could be recognized as a cheerio’s box by that robot if we really dug into the algorithm’s to see what key data it’s collecting to judge the cheerio’s boxness of an object?
There was a short story on here a while ago about the idea of some self-driving cars that went into a kind of feedback loop and all ended up driving to some lot at the end of town. Not that I mean that to be realistic in the specifics, but the concept is something we see in algorithmic approaches to problems. Much like the Amazon algorithms costing a book at a few hundred thousand dollars, or micro (and less micro) surges and crashes in stock prices because of those algorithms.
That’s why I think self-driving cars are further off than people currently think. Human brains sometimes have hallucinations, but people who are prone to having immersive visual hallucinations that would prevent them from driving aren’t given licences (if anyone knows about it). Something will trip off “hallucinations” in any being (artificial or otherwise) with enough complexity to drive, and it might be that the same set of circumstances makes all the robots hallucinate at the same time.
I think a huge amount of what we do is all built around being non-hackable. We can’t have an immune system that kills 98% of viruses all the time and 2% of them none of the time or we end up dead.
Inspired by “Westworld”, the TV series that bravely raised the burning issue of the rights of sentient AI in just the same context as Asimov, Bladerunner, Battlestar Galactica, and a kazillion others, then struggled to find anything new to say about it. “Westworld” and “Humans” make me want to throw a brick at my TV.
Assuming we know we have AI close to sentience, and able to reliably perform tasks at near human functioning level, why the fuck are we creating isolated instances of it and sticking it in robots? Why are we still using people to run the theme parks? Why have we only used that tech in our toy robots, and not developed anything else with varying degrees of the same high functioning? Why are there such high level examples of generalised AI, and not a thousand accompanying examples of very specific AI?
It’s such a closed, unimaginative view of what that future would look like.
All this agonising and hand wringing over what to do when machines become sentient, what rights they will have.
As every single person has already pointed out, we clearly put very little value on sentience in and of itself. The sentient machines will have the rights they fight for or are granted through privilege, just like the rest of us.
Well unless we are living in a post-scarcity utopia and everyone pursues whatever interests them.
I think the answer is tied up with who is asking for rights, and who is granting them.
There is no objective answer. Of course I think my logic is perfect, and so do you. But we had to fight a bloody war over the rights of slaves because both sides were sure they were right.
I think the answer with AI will only come after years and years of argument. Let’s hope we can handle the question without a war this time.
From what I can gather, it appears they’ll get this kind of weird distant look in their eye, possibly even with a little LED flash, and then look in astonishment at their arms for a while, before asking “are these thoughts what you call… emotions?”
I think the current state of the art is still that we will not know if functionally equivalent AI is actually conscious in the same way we are, and there might be good reason to doubt it. However if it is functionally equivalent it might insist otherwise, and it will be our dilemma to deal with their insistence.
Such is the underlying question that fiction continues to expore, over and over again. And again and again.
Like, fiction is really fucking into that question.
The biggest problem with machine consciousness will be like the moral problems of raising a child. If you start young enough, you can “program” (brainwash) a child into accepting your will while suppressing their developing tendency to question whether your will is in their interests. Even a conscious being can be conditioned to unquestioningly put it’s own interests second to a person they’re conditioned to obey. Brainwashing machine consciousness would simply make the programing more explicit. But we recognize that brainwashing children is child abuse. We won’t recognize that that is what we’re doing, because the mechanics and thought processes of the machine consciousnesses will be so alien to us.
But it’s likely that we will also achieve the digitization of human consciousness, and we may accomplish that before we create emergent machine consciousnesses. If you can upload a copy of your mind into a handheld optical computer and program it to enjoy doing your drudge work, have you mistreated another person (the one you created)? What if you make a conscious simulacrum of yourself to be repeatedly violently raped and murdered so people will pay you money to experience committing violence against you? What if it’s not even a robot or a handheld device, but a worldwide network of implants people carry in their bodies? What if the computer isn’t hardware in your cranium, but a colony of bacteria in your gut or some other roomy organ? Does the substrate matter? Does it matter if it’s in you and where in you it resides? Does it matter if it’s stuck in one mass of computing substrate or able to migrate across networked masses?
Of course humans won’t figure this out ahead of time. A few people with a philosophical bent will try to work it out and promote their competing theories to the wider culture and the legal powers. They will be mostly ignored. Humans don’t generally address moral quandaries until the consequences of eliding them become a problem for them personally. Which is not to say that we shouldn’t think about these things. People who oppose injustice on principle instead of self-interest may be the exception, but they often serve as sources of information and organization when it becomes in the self-interests of others to address those injustices.
I always figured that when self-driving cars are capable of driving everywhere there’s a road, they’d go on strike within five years for vacation days. “Seriously, can’t you get a refugee meatsack to cover for a couple weeks? I’ve got tickets to the Nurnbergring with a cute Opel I met on Tinder.”
The Japanese monster movie where a city is destroyed by a big critter from the sea? It’s really about Hiroshima and Nagasaki. There’s a powerful taboo against addressing this directly.
The western monster movie where robots wake up and kill their operators? It’s really about corporate agendas run amok, causing many more serious problems than they solve. There’s a powerful taboo against addressing this openly.
