In you're in favor of "AI safety" (broad definition), what's your most compelling cast-length argument?

At fist automanufacturers made similar arguments against requiring seatbelts in cars. 

Imagine what we’d be saying about our great/grandparents if that line of thinking had prevailed.

Building something new. Host: /venturecapital /adventures /theitinerant

It took 60 years from Model T to first seat belt laws? So the technology was well understood.

Yes, that's true: it took 60 years of people dying before the political will could be mustered. The issue was known as early as the 1900s; not too dissimilar a situation from today's AI discussions.

AI is not a new technology, its potential is well understood. Regulation isn't going to kill it at this point.

Yes, regulations do form from precedent. Not because that's good, but because that's the only way political will can be mustered to overcome commercial pressure. 

How many lives (i.e., "cracked eggs") is an omlette worth to you? 

Assume these lives are close friends and/or family members, not some unknown individual.

It's not about its -potential-, it's about what it does.

Good regulations form from precedent. And yeah, gotta crack a few eggs to make an omelette.

Trying to "divine" regs from intuition never works. We have centuries of examples to prove it.

Totally fair :)

GPTs are new but AI as a field of research goes back to the 1950s. Where we are today has been a long anticipated by scientists.

Hard, yes, but important given that our unknowns are the result of the creation itself. Not an outside factor. 

I don't believe this has ever before been the case.

> AI is not a new technology, its potential is well understood. Regulation isn't going to kill it at this point.

I disagree with all 3 of those statements? :)

Every day there's new stuff and we still don't understand it. Hard to regulate something you don't understand effectively.