Why can't anyone agree on how dangerous AI will be?

https://www.vox.com/future-perfect/2024/3/13/24093357/ai-risk-extinction-forecasters-tetlock

by T-A • 2 years ago

6 6 2 years ago

bell-cot• 2 years ago

Reaction I: "It's tough to make predictions, especially about the future"

Reaction II: When you say "A.I.", which of the following do you mean?

- Actual autonomous "SciFi" A.I., doing what it wants to do, because of it's own, not-human-derived motivations, to the detriment of humanity

- Supposed A.I. that is merely a techno-fig leaf for nasty sh*t that humans would be doing to each other anyway

- Hyped-to-the-stars A.I. that is sold as the great solution to a bunch of huge human problems...but that was 99% hype, so we're screwed when it doesn't really work

- Other

ganzuul• 2 years ago

“The best way to predict the future is to create it.” is optimistic. It asserts that we have agency.

The pessimistic argument seems to say we don't have agency. It follows from this attitude that we can only stave off a collapse.

It is one self-fulfilling prophecy against another. Presumably these attitudes are based on more fundamental philosophies.

There may be other philosophies with prophetic power than these two.

lm28469• 2 years ago

We can't even agree on how useful it is... Or on what jobs it'll replace... Or even what does it (ai) means

hnenjoyer_93• 2 years ago

I briefly skimmed through the text, and it seems to cover only the anti-AI arguments and how some people agree and disagree. At the same time there seems to be no pro-AI arguments, like improving the efficiency of work, medical advancements and other useful things. Maybe that's what was meant with such a title but I would like the author to cover the whole picture next time

putna• 2 years ago

what if it is more useful than dangerous?

aiforecastthway• 2 years ago

The study [1] doesn't say who the experts were, but a related paper [2] by the same authors states that the sample drew heavily from the Effective Altruism community.

I would be pretty skeptical of generalizing these results to a sample of people who aren't EA or EA-adjacent. That's a very specific type of a community/person with a lot of priors and methods of thought that are both highly specific and relatively uncommon outside of EA circles.

[1] https://static1.squarespace.com/static/635693acf15a3e2a14a56...

[2] https://static1.squarespace.com/static/635693acf15a3e2a14a56...

Why can't anyone agree on how dangerous AI will be?

Related Stories