This line jumped out at me:

“In other words, AI is on track to perform at base competence on most tasks in about three years. These researchers believe agents will need another few years to outperform humans”

LLMs have shown steady improvements, but it’s a big assumption they will continue their performance gains, and specifically, that it’s a simple matter of time before they surpass human level performance on “most tasks”. (In six years no less!) The author is claiming LLMs are fundamentally AGI capable.

This is not going to end well. At least I will be able to pick up used DDR5 memory and GPUs on the cheap in the not too distant future.