3 ms·

Artificial intelligence warning over human extinction labelled ‘publicity stunt’

7 points by cpncrunch 3y ago

coldtea 3y ago
Not just labelled, might very well be.
- coldtea 3y ago
  Let me qualify that: (a) it attracts tons of free publicity (b) allows them to sing the praises of their AI services as not just a nice business, but also as "extremely smart" to the point of being (eventually) dangerous (c) they already know the "moratorium" is BS, and ain't gonna happen, so no districtions to their AI business by their "suggestion" about having one (d) they're poised to talk to top political and military leaders to discuss this, present their businesses as national assets, get nice deals
ggm 3y ago
I think this is an unfair criticism which goes to tone more than intent. It's about getting some regulation back into the room before its too late. Is it really about extinction? No. Technology mis-application is as old as technology. We saw it with expert systems deployments in the 80s and 90s which codifed systemic bias in medical school admissions. At that time, the "blame chain" was clearly to the experts who codified the admissions policies too accurately (!) The problem with AI is that it's increasingly being used as a hall pass out from human review. So, the layer of indirection in the applied decision is really strong here. Who do you "blame" when a systems does poor inductive modelling from a fact set, using bad algo and then you take policy decisions? Sure, you made the mistake believing it, but the problem is .. capitalism drives hard to save money, so its de-skilling. Australia had a pre-AI version of this with something called "robodebt" when Algo decisions on welfare payment fraud were let loose at large. People died (suicide) because of the debt burden, for something which was just plain wrong. When you wind up with insufficient people willing to take money to fly and then release weapons from a reaper UAV, The obvious solution to a cash-strapped military mind is to let the GPT do it for you. "extinction" is hyperbole, sure. Thats how you get attention. But the societal effect of misapplied generative knowledge, and models, with strong belief outside a cogniscenti to their correctness is a recipe for disaster. Keep humans in the loop. Do you think Sam Altman and Geoff Hinton releasing a manifesto which said "keep humans in the loop" was going to get mindshare? I don't. BTW I am an AGI skeptic. I think AI is a terrible label for statistics and modelling. None the less, I'm glad they made the statement and signed it, and I am glad regulation is coming. I am also capable of believing Altman said it, to secure and retain competitive advantage. Just because one or two signees have cynical upsides doesn't invalidate the core concern: This stuff needs regulation.

7 points by cpncrunch 3y ago

coldtea 3y ago

Not just labelled, might very well be.

coldtea 3y ago
Let me qualify that: (a) it attracts tons of free publicity (b) allows them to sing the praises of their AI services as not just a nice business, but also as "extremely smart" to the point of being (eventually) dangerous (c) they already know the "moratorium" is BS, and ain't gonna happen, so no districtions to their AI business by their "suggestion" about having one (d) they're poised to talk to top political and military leaders to discuss this, present their businesses as national assets, get nice deals

ggm 3y ago

I think this is an unfair criticism which goes to tone more than intent. It's about getting some regulation back into the room before its too late. Is it really about extinction? No. Technology mis-application is as old as technology. We saw it with expert systems deployments in the 80s and 90s which codifed systemic bias in medical school admissions. At that time, the "blame chain" was clearly to the experts who codified the admissions policies too accurately (!) The problem with AI is that it's increasingly being used as a hall pass out from human review. So, the layer of indirection in the applied decision is really strong here. Who do you "blame" when a systems does poor inductive modelling from a fact set, using bad algo and then you take policy decisions? Sure, you made the mistake believing it, but the problem is .. capitalism drives hard to save money, so its de-skilling. Australia had a pre-AI version of this with something called "robodebt" when Algo decisions on welfare payment fraud were let loose at large. People died (suicide) because of the debt burden, for something which was just plain wrong. When you wind up with insufficient people willing to take money to fly and then release weapons from a reaper UAV, The obvious solution to a cash-strapped military mind is to let the GPT do it for you. "extinction" is hyperbole, sure. Thats how you get attention. But the societal effect of misapplied generative knowledge, and models, with strong belief outside a cogniscenti to their correctness is a recipe for disaster. Keep humans in the loop. Do you think Sam Altman and Geoff Hinton releasing a manifesto which said "keep humans in the loop" was going to get mindshare? I don't. BTW I am an AGI skeptic. I think AI is a terrible label for statistics and modelling. None the less, I'm glad they made the statement and signed it, and I am glad regulation is coming. I am also capable of believing Altman said it, to secure and retain competitive advantage. Just because one or two signees have cynical upsides doesn't invalidate the core concern: This stuff needs regulation.