The New York Times is Not Too Stupid to Understand AI
We now have extraordinary evidence of GPT4 displaying human-level creativity and reasoning. Nonetheless, the conversation around AI is dominated by pieces drawing out errant, simple behavior.
I’d like to say that the alarmist and ridiculous examples such as Kevin Roose’s experience in the New York Times are a trick. I’d like to say that the only evidence against human-level AI is lackluster performance on AP English exams. I’d like to say that making arbitrary distinctions based on the robot’s IQ is a way to get slapped in the face when superintelligence comes.
I can’t say any of that, really, because I’d need undeniable evidence to make what could only be an undeniable claim.
Why specifying human-level AI requires undeniable evidence
The existence of human-level AI would mark an international priority, and kick off an arms race between the US and China toward AI superintelligence.
Second, and probably most important, our country’s inequality relies on a system of merit and intellectual achievement, and artificial intelligence is a threat to the ruling class.
For these reasons, AI intelligence will be a contrarian take until we pass the threshold of superintelligence. The future of artificial intelligence remains the bland, mealy stuff of Last Week Tonight pieces.
For now.