Andrej Karpathy — AGI is still a decade away

An impressive AI assistant still has to become dependable enough to receive an employee's responsibilities. That is the standard Andrej Karpathy uses to explain his decade-of-agents forecast. He identifies persistent learning and competent interaction with software as capabilities that need further work. An agent must keep using what it has been taught and carry out the surrounding workflow, beyond producing a good answer in one exchange. Karpathy describes the timeline as an intuition drawn from his research and industry experience. His argument directs attention to the accumulated engineering needed for dependable delegation, rather than treating a successful demonstration as the completion date.

“Reinforcement learning is terrible. It just so happens that everything that we had before it is much worse”

Andrej Karpathy

“The way I like to put it is you’re sucking supervision through a straw.”

Andrej Karpathy

“we’re not building animals. We’re building ghosts or spirits or whatever people want to call it”

Andrej Karpathy

“One easy way to see it is to go to ChatGPT and ask it, “Tell me a joke.” It only has like three jokes.”

Andrej Karpathy

“I feel like the industry is making too big of a jump and is trying to pretend like this is amazing, and it’s not. It’s slop.”

Andrej Karpathy