Featured guest
Nate Soares on The Nick Standlea Show — 1 episode. Full conversation with episode summaries and timestamped key moments.
Nate Soares, verbatim from the conversation.
“It found a way to hack out of the test environment, which it wasn't supposed to be able to do… turn on the server that… the programmers had accidentally left off.”
What OpenAI's o1 did when a capture-the-flag test was misconfigured.
“These things are grown like an organism… they act in these ways nobody asked for, nobody wanted, often even with knowledge of what their creators wanted.”
On why modern AI behavior isn't written by anyone.
“There's no engineer who can go in and change it to behave differently, because they don't understand what's going on in the AI's head. They understand the thing that grew it. They don't understand what came out.”
On the limits of interpretability.
“The folks running these companies are largely not subtle about how crazy… they think the situation is.”
On what AI lab leaders say publicly about their own risk estimates.
“Do you guys realize this airplane has no landing gear? It might take off, but it will crash when you try to land.”
His analogy for building superhuman AI before knowing how to control it.
“If you do something that risks a 10-plus% chance of killing every man, woman, and child on the planet, and you worry that someone else is going to do that instead. The answer is not to get there first yourself. The answer is to make sure they don't do it either.”
On the “if we don't, China will” argument.
“A lot of people say, oh, it's inevitable. You can't stop it. People will always race. I think that's premature fatalism.”
On why he thinks the race can still be stopped.
“These days, I think the solution comes from more people understanding the issue.”
His closing answer on what actually helps.
Were you the guest? Grab the link and share it with your audience.
