An AI safety researcher wrote 30 pages on how to escape the simulation. He never argues we’re in one.

Through the looking glass…

An AI safety researcher wrote 30 pages on how to escape the simulation. He never argues we're in one.

He asks an engineering question instead — and walks into a trap.

AI safety spends millions on "boxing": sealing a superintelligence in a sandbox it can't escape. But if our universe is simulated, it's just a very large box with us inside it. Same problem, opposite ends.

So one of these is false:

• We can permanently contain a superintelligent AI
• We can escape the simulation

Both can't be true. Pick.