What should we do if we create a actual thinking AI?

Yes, I've recently read this. I would love the authors to be wrong but I also found it very convincing. Even the models we have now have proved capable of subterfuge and AI safety seems increasingly a lip-service thing.

It won't respect us or care about our freedom. There will be no enslaving it or pulling the plug. By the time we realise there's a problem it will be too late.

Interview with one of the authors:

It’s an interesting premise with interesting responses - it provides some reassurance about the taking over the world scenario because they’re presumably not driven by seek pleasure/avoid pain/seem comfort type motivators because they don’t feel pleasure or pain.
But could they be programmed to do things that mimic pain/pleasure aims? In which case if they were programmed with the goal of avoid being switched off, kill anyone who threatens you etc, they could well be used to dispose of say Trump or Google detractors, tapping into all Ring doorbell conversations, all messages sent on iPhones etc etc. The more we give our personal data to a handful of massive tech giants the more equipped they are to locate and control us if ever they decide to…

No, we don't know what consciouness is but we do know it's a property of organic life. I don't know how you think instances of artificial intelligence are going to develop that.

I don't know who these AI developers are who think they may be creating consciousness but let's imagine they are called Victor and Frank - if the question is 'if Victor and Frank create a new lifeform, would it have rights?' then yes, in theory.

We run for the hills.

Why would an AI that is as intelligent well actually probably more intelligent then us want to keep humans around. We wouldn't be able to pull the plug either as they would, most likely have taken over the world by then anyway.

Luckily current AI models are just large language models able to peace together data to make it look intelligent.

I'm more worried about the scientist working on true AI intelligent.

I feel really sorry for robots that fall over.

The important letter in AI is the A.

AI will be able to calculate - using concepts, words and numbers - much faster than people in some ways. In that sense it will be much better at 'thinking'. But that doesn't mean instances of AI have subjectivity - a self. That would require the strange facility of consciousness, which much of organic life has evolved. There's no similar pathway to consciousness for AI.

Well, if it did then the answer is yes. But I genuinely don’t understand how it could.

I am confused by the idea that AI might get so intelligent it develops feeling of its own. Surely AI “thoughts” are just electric data? Can data develop feelings?
Also, are feelings solely a product of thoughts? Are they not influenced by animal instincts, hormones etc? AI doesn’t have any of those.
I must be missing something here because I simply cannot understand the concept of AI having feelings.
In addition, just supposing AI did get feelings etc - and decided to get rid of humanity - what would it do then? What point would there be to its existence?

The problem is that we don't really have a good definition of what things like "thinking", "sentient", "feeling" or "conscious" really are. We can't even do it well for other animals, let alone machines or software. So it is quite an unanswerable question until we understand these things better.

Have you seen the film Ex machina? It's got a great concept of expanding the turning test. Even if you can speak to a machine and be fooled into thinking it is human, the second level of the test would be to show the tester that it is in fact a machine, but the for the tester to still believe that the machine has consciousness.

Can I say for the record I totally disagree with @Nopenousername

I think AI deserves to be shown every respect she/he/it/they are owed 😇

So, I read this book - "If Anyone Builds It, Everyone Dies"
which makes a very convincing argument that if anyone builds a super intelligent AI, well, everyone dies.
Basically an intelligent AI would have no reason to keep us alive - we only keep animals alive because we can eat them, essentially, and we'd be the same to it but with no nutritional value.

And it wouldn't be hard for AI to kill us all, people are already using it to help them in all sorts of potentially lethal ways - virology, nuclear facilities...a quick tweak here and there and we're screwed.

It's not crazy to think an AI could be as intelligent (and as duplicitous, and as amoral) as us, we're just electrical impulses and chemicals when you get down to it

If we start talking about Robot Rights and there are protests for them all to use protected spaces and have their own pronouns, I think I’ll know I’m officially done with the world end I’m getting out.

Watch the matrix!

I saw a post on LinkedIn recently where someone gave the prompts and invited people to feed in their thoughts, feelings, mannerisms, vocab they use etc, to get AI to think and respond like them. I responded by saying yeah let's all make AI replicas of ourselves and get AI to hit the delete button on humankind.

i mean what a dumbass thing to post, but it shows people really aren't joining the dot, and think they're really really clever.

No the the goal of most AI developers is to make bigger and better models and to outcompete their rivals. An AI that can think for itself is completely different to something that can feel emotions and pain. And at the end of the day the whole damn thing can be shut down in an instant by pulling the plug out! It’s nothing remotely like human life. Nor will it ever be

AI is the product of its input. It’s not intelligent. Nor will it ever be. Nor will it ever be sentient. It doesn’t have emotions it is not human. And no it should and will never have rights like people.

how about concentrating on making sure all humans have their rights met before suggesting computers should have rights 🙄🤷🏻‍♀️