If I help you get rid of the gun you shot someone with, technically I didn't commit the murder. I'm still on the hook for it though.

This isn't that different: it's the state that violates constitutional rights. Flock helps with that. They should be treated like the guy helping getting rid of the murder weapon.

Then there's also the penal code.

If any of the idiots programs an "internet worm" which spreads uncontrollably and causes damage, we already have legislation in place to jail them. "Your honorI didn't intend to do damage" doesn't save you these days. (It might've been different in the 1980s, but not anymore.)

So jail whoever fails to control their AI, using the very same laws.

But the problem here ia that it's about money - a *lot* of money - and there is every incentive to lie, deceive, and perform "creative marketing". So we're being held for fools, and everyone is buying into an obvious lie as a marketing strategy. "We can't conteol it" my ass.

Nobody asked the car to kill pedestrians. I just put a brick on the gas pedal.

You're being sarcastic, but you're more right than you think. Either way we don't "need to worry", we need to do something about it.

But the thing is: specifically *what* we should do depends a lot on where the threat ia coming from - the AI, or the guy.

But that's not the AI killing us, it's the idiot who put the AI in charge. Kind of like putting a brick in the gas pedal of a car abd letting it loose in the pedestrian precinct.

And at any rate, it doesn't need 'better" AI to do that. ChatGPT 2 or 3 are already perfectly capable of fucking up at that level. All you need ia someone to build them a paperclip factory.

Ban paperclips. /s

Sweet summer child yourself.

"not university" != "private sector"

The limits of your own horizon isn't where the world ends, dude. Just because you don't understand the concepts like free will and initial decisi doesn't mean they don't exist.

I don't care whether you "do ChatGPT" or not. You're a prick so the discussion's over as far as I'm concerned.

Drive isn't the same thing as an initial impulse. (It was supposed to be "initial" not "infernal" - autocorrect at work.)

AI doesn't just "decide" to break out.

One thing it's lacking is internal motivation, the initial impulse to do stuff. It needs to be told to do something it doesn't "want" on its own.

Wrong.

One thing humans have and AI doesn't (and it's not even close to having) is an infernal impulse for action - one that's truly bot rooted in some kind of programming or instruction or prompt.

Not according to ChatGPT. Frontier models are still just token predictors.

It's not the sole purpose of emotions to get controlled, you know... they're a valuable pocessor of information in itself.

Not all information is logical, self-consistent, or consciously actionable. But it's information nonetheless, and can make the difference between survival or not. This is why we have emoticons.

Not biology, but material science here. I actually talked to ChatGPT about training local models for my discipline (oh, thw irony).

Apparently my RX 7900 XTX is plenty and I could run and train a Quen 3.5model in the 15-30 bn range. And CgatGPT expects it to be quite capable as a "scientific research assistant", close enough to frontier models to never look back.

Experiment hasn't started yet (there's more to the story, the main part being how to set up the training material - I'm in a fairly unique position), but it sounds encouraging that I don't need to sell my house for this just yet. :-)

They're not "squishy science", they're the factors that actually have thr most influence on whether someone turns criminal or not at that economic level.

They the actual science.

Going for the blank numbers instead without understanding model or context is the true squishy science.

Or do you actually suggest that the amount on melanin in someone's skin is somehow a mechanism that drives crime?

What happens when you normalize not for skin color, but, say, for income, parents income, access to education etc?