To expand on my comment wrt "promising for overall advancement": My daughter, in her math class: Her teacher- I'll reserve overall judgement on their teaching: she may be perfectly adequate as a teach for other students, which is part of my point- simply doesn't teach in the same sense other teachers do: present topic, leave details of "figuring out how to apply methods" to the students. Doesn't work for my daughter, who has never done less than excellent in math previously. She realized she ChatGPT (we monitor usage) for any way of explaining things that "simply worked" for how she can engage with explanations. Math has never been as easy for her, even more so than before, and her internalization of the material is achieving a near-intuitive understanding.

Now consider: the above process is available and cheap to every person in the world with a web browser (we don't need to pay for her to have a plus account). If/when ChatGPT starts doing ridiculous intrusive ads, a simple Gemma 3 1b model will do nearly as good a job) This is faster and easier and available in more languages than anything else, ever, with respect to individual-user tailored customization simply by talking to the model.

I don't care how many pointless messages get sent. This is more valuable than any single thing Google has done before, and I am grateful to Rob Pike for the part his work has played in bring it about.

Seconded — "AI" is a great teaching resource. All bigger models are great at explaining stuff and being good tutors, I'd say easily up to the second year of graduate studies. I use them regularly when working with my kid and I'm trying to teach them to use the technology, because it is truly like a bicycle for the mind.