My overall thoughts on AI boil down to:

  1. AI is a powerful tool.
  2. Engineers should use the right tool for the job. In other words, engineers should always be driven by problems, not solutions.
  3. Engineers should use AI when it’s the best tool to solve a given problem.

To add slightly more color, I think AI can be the right tool for the job when (a) a suitably generalized dataset is available and (b) the consequences of hallucination are acceptable. (b) often implies that a human is in the loop. I enjoyed reading Microsoft CTO Kevin Scott’s take here on how AI copilots are an ideal model for AI tools.

I think Geslin et al. is a fantastic overview of AI for battery lifetime prediction.

UPDATE 2024-06-24: Another framework I’ve found useful is replacing “AI” with “statistical model”. Could a statistical model write a good draft of an offer letter? Sure. How about a scientific paper? Unless your work is unoriginal, probably not. Can it predict the future? Depends on if the future follows the same distribution as the past.