Re:Gorilla: LLM Connected with APIs
by ethanbond on Thursday May 25, 2023 @07:00PM
(#36077545)
Is this meant to be convincing? All you’re doing is asserting that the positives outweigh the negatives and starting from there.
That is the point of disagreement. We do not currently know if the positives outweigh the negatives.
I am saying that market incentives are good at spurring more people to scale these systems and their deployments up, and those incentives are not good at mitigating negative externalities.
What I am pointing out is that the system of incentives surrounding these technologies has failed to produce anything resembling a safe development trajectory. No, these systems are not quite sky-is-falling-scary yet, but if AI development achieves its promise, they absolutely will be.
We have no reason to believe, as a commenter asserts below, that in just a few more inches closer to potential catastrophe we (who?) will decide (how?) to pull the reins (where?).
That is the point of disagreement. We do not currently know if the positives outweigh the negatives.
I am saying that market incentives are good at spurring more people to scale these systems and their deployments up, and those incentives are not good at mitigating negative externalities.
What I am pointing out is that the system of incentives surrounding these technologies has failed to produce anything resembling a safe development trajectory. No, these systems are not quite sky-is-falling-scary yet, but if AI development achieves its promise, they absolutely will be.
We have no reason to believe, as a commenter asserts below, that in just a few more inches closer to potential catastrophe we (who?) will decide (how?) to pull the reins (where?).