AI Safety Meme of the Week

Would You Take a Job with 40% Odds of Killing Us? AI Engineers Would

Courtesy Of

While we appreciate @AISafetyMemes’ dark humor, the more you look into the joke, the harder it is to laugh. You’ll see public voices in AI like Yann LeCun, chief AI scientist at Meta, saying AI has no real safety issues and people who think it does don’t know what they’re talking about. (We’ve already commented in a previous Safety Meme on why open-sourcing powerful AI is insanely dangerous. LeCun is the guy who did that.)

Here’s a bar graph from a 2023 survey of 841 AI Engineers.

(For those of you unfamiliar with the terminology, P means probability, and doom is exactly what you think.)

Looked at in a different way, only 10% of AI engineers think there is no chance AI will go wrong. 30% think there is between a 1/4 and 1/2 a chance it kills humanity, 15% think that chance is between 1/2 and 3/4, 10% think it is between 3/4s and 99%, and 2% of these AI engineers believe THERE IS A 100% CHANCE IT WIPES US OUT. This is a field where reckless speed is the new normal. They want to keep it that way.

This isn’t sci-fi. That’s what Yann LeCun and Marc Andreesen want you to think.

The professionals in LeCun’s field think otherwise.