Building a super-intelligent AI is the stated goal of major AI companies. It may or may not be possible, but due to the recent surge in capabilities of frontier models I would not take any risk.
Frontier models are, indeed, already too capable and too dangerous. I would have been more comfortable if a stop of the development occurred 12 months ago. However, the second best moment to press the "stop" button is now.
Frontier models are, indeed, already too capable and too dangerous. I would have been more comfortable if a stop of the development occurred 12 months ago. However, the second best moment to press the "stop" button is now.