How quickly the Agents have become routine

I have been sitting in front of my screens watching various agents perform tasks for me, like I have been for several months now. It just struck me what a change this is compared to how I was working nine months ago, a year ago.

I wouldn’t say that we have reached anything resembling AGI (or whatever the catchword is these days) although these tools are remarkable, today.

I still feel it is important to retain a healthy dose of skepticism about the output of these models. There are still many instances of the models just going off on unhelpful tangents filled with hallucinations and suppositions.

At the same time, they increasingly have moments of what feel like remarkable insight and simulated genius in how they approach issues.

2016 me would have considered what we have today to be practically science fiction, even for so-called next word predictors¹.

1. Calling today’s models “next word predictors” just makes what they can do even more incredible. Imagine what future AI could do when it can actually think and reason more like we image AGI to be! ↩︎