AI Watch

[Long slender neck]

How exactly is AI going to make us extinct?

[CEB]

How exactly is AI going to make us extinct?

We don’t fully know how AI’s work. An AI with decision making capacity and connection to the internet could do a lot.

[Long slender neck]

What AI are you talking about?

[CEB]

What AI are you talking about?

That’s part of the reason that AIs are a threat; there isn’t a particularly high barrier to creating an AI, people know how to do it. Regulation might restrict commercial use of AI, but it’ll be near impossible to police.
If you’re not concerned about what it could I advertantly or deliberately be used for (or decide to do itself) then you’re not reading the concerns properly.

[Long slender neck]

I think you're getting a bit carried away here, it all seems like hype to me since ChatGPT was in the news. Now people are slapping the AI label on anything computery or automated.

What exactly are you imagining an AI is and will do to us?

If it looks like an AI might make us extinct (how exactly could this happen?) I suspect we'll just unplug it.

[Max Fowler]

If it looks like an AI might make us extinct (how exactly could this happen?) I suspect we'll just unplug it.

Exactly. God-like super-human intelligence won't be able to cope with the old Ctrl Alt Del combo.

[Long slender neck]

Here's a comment from a tech site I read that you might find useful

Gross misunderstanding of the tool

There seems to be a collective and very gross misunderstanding of how these tools function.

Repeat after me. ChatGPT IS NOT AI. ChatGPT IS NOT AI.

It is a language statistical model that strings sentences together in ways it has been trained to do. It doesn't understand context. It doesn't understand truth.

A computer will do exactly what you *tell* it to do, NOT what you *want* it to do. And ChatGPT is no different. You give it a prompt, it will provide some kind of response to that prompt.

Any bearing on reality is entirely coincidental, yet for some god-forsaken reason, people are acting as if this thing is some kind of bloody oracle or something. ChatGPT is an automated bullshit generator, nothing more, nothing less. If you ask a very specific, closed question, it will (usually) produce a reasonably accurate answer to that question. This makes it very useful is software development when you can save some time by having it whip up some boilerplate code. But absolutely _nothing_ it outputs can be trusted.

[CEB]

That paragraph doesn’t actually deal with the issue.

Nobody talking about the threat of AI is saying “hey…. This is a little guy in my pc!!!!”

Part of the inherent threat of AI is exactly that it does what you tell it to - we don’t know how machine learning tools might extrapolate and have unforeseen circumstances.
As Trump says above, the idea that if it goes wrong we can turn it off is naive

[CEB]

This isn’t to doom monger - my biggest actual fear around AI is threat to jobs in creative industries, which is already happening.
But it’s naive in the extreme to think that tech bros are responsible enough that there’s nothing to fear from where this tech goes, and how quickly it might get there.

[Long slender neck]

That paragraph doesn’t actually deal with the issue.

Nobody talking about the threat of AI is saying “hey…. This is a little guy in my pc!!!!”

Part of the inherent threat of AI is exactly that it does what you tell it to - we don’t know how machine learning tools might extrapolate and have unforeseen circumstances.
As Trump says above, the idea that if it goes wrong we can turn it off is naive

Thats what computers do, nothing new is it?

I dont want to sound patronising but I think you dont really know what you're talking about.

[Long slender neck]

This isn’t to doom monger - my biggest actual fear around AI is threat to jobs in creative industries, which is already happening.
But it’s naive in the extreme to think that tech bros are responsible enough that there’s nothing to fear from where this tech goes, and how quickly it might get there.

Dont worry I suspect your job is safe.

[CEB]

That paragraph doesn’t actually deal with the issue.

Nobody talking about the threat of AI is saying “hey…. This is a little guy in my pc!!!!”

Part of the inherent threat of AI is exactly that it does what you tell it to - we don’t know how machine learning tools might extrapolate and have unforeseen circumstances.
As Trump says above, the idea that if it goes wrong we can turn it off is naive

Thats what computers do, nothing new is it?

I dont want to sound patronising but I think you dont really know what you're talking about.

The issue here is that you are unable to summarise the best version of the counterpoint to yours.

You’ve fundamentally misunderstood the point: yes, “dumb” computers do exactly what they’re told to do. The potential issue is the parameters of that. You and I are used to computers that do *exactly* what they’re told to do.
The whole point of AI is that you’re creating a machine that can eventually, ideally, make *better* decisions than humans might, towards a desirable outcome. But without careful restrictions on what AI can and can’t do/control - and without full exploration of how to safeguard against unwanted outcomes, there is an obvious and inherent risk to delegating big decision making to AI. Also big potential rewards.

[Long slender neck]

The issue is you're talking rubbish.

[CEB]

Good one. I guess you’re right though, I’ll trust the random who admins a forum and posts an attributed comment from “a tech site I just read” above many of the leading scientists and architects of AI.

[Long slender neck]

Again, which AI is it that you're referring to?

[CEB]

Sainsbury’s AI

[CEB]

The question you’re asking is equivalent to saying “which car are you worried might kill you?” during a conversation about whether speed limits are a good thing

[Long slender neck]

Sainsbury’s AI

A self aware self service checkout?

[Dunners]

Sainsbury’s AI

A self aware self service checkout?

I can just imagine now the tuts of disapproval from the self service checkout, as I scan through the week's supply of wine and beer.

[JimbO]

I Take all my Knowledge on AI from the Terminator films

[Max Fowler]

Again, which AI is it that you're referring to?

HAL 9001

[RedDwarf 1881]

I Take all my Knowledge on AI from the Terminator films

Exactly . When people start talking about A I and possible cause of the end of humanity that's the image I get in my head .

[tuffers#1]

Again, which AI is it that you're referring to?

The Terminator 1 I Think .

[Long slender neck]

How exactly is AI going to make us extinct?

We don’t fully know how AI’s work. An AI with decision making capacity and connection to the internet could do a lot.

This doesn't exist, does it?

[CEB]

You don’t think decision making is delegated to AI?