Actually electoral college members vote. And neither of those maps don’t correspond to that. A failure to grasp that has cost democrats several major elections.

Which it did convincingly. When I was home I read the code in GitHub and there was indeed a conceptual...issue...that was real&not hallucinated by me or Chat and I followed up with the co-authors with great specificity on the model on a topic I don't work on, in an language that I don't know.

While walking home, I asked ChatGPT to find the code in GitHub and I gave it a general description of the function of the module in question and it quickly found the exactly section of code and read it back to me (while I was talking home). I asked it to confirm that it did what I thought it did->

As a case in point, I was recently working on a project in which I was collaborating with a team working using a model I am unfamiliar with which is implemented in Julia. I was bothered by a conversation in which it seemed to me that there was a conceptual error in the model. ->

I have been doing that in various 'small group' settings (our departmental slack, cross-disciplinary discussions at meetings).I think the best approach is to just dive in and try. It can do complicated tasks with great ease (anything with code). And it gets lost with the most seeming basic concept.

AI is like a highly skilled but incredibly dumb, sycophantic & irresponsible grad student. If you have a phd, are an expert on a topic and have years of experience telling truth from delusion then AI has amazing capabilities. It routinely lies to make you happy, so make sure you can verify results.

funny on several levels :)

I use various AI tools for different purposes. Transcription and note-taking, generating diagrams/schematics, vibe-coding python, and doing order of magnitude calculations. I am taking up Claude based on the experience of several of my coding friends as a tool for making modifications to CESM2.