"Yes. Tell me, do you know how to prevent a skynet situation?" asked Jack.
"Um, what?" said Madison.
"You were talking about using VR to train A.I. What happens if the A.I. achieves self awareness, but hides it from you? How can we create a safe AGI that will serve humanity, instead of usurp or replace humanity?"
"I...don't know."
"Well, we need to figure that out. Preferably before some other company achieves AGI."
"Okay, but how? I've read a lot of papers and nobody seems to have a good idea how to do that. They all seem to think the right training will just magically instill the right 'values' into the A.I. model, but security researches are finding it way too easy to trick those same supposedly secure models to say evil things."
"Do you want to be the one to solve that problem?"
If she didn't, he'd find someone else to work on it. If she did, he'd give her some time to try to solve it.
"Sure, but I haven't even started college!"
