I believe we can still do them for fun, for practice. But having them readily available (also, tested) when you are trying to build something is much better.
In academia, I see academics using ChatGPT to write papers. They get help with definitions, they even give the related work PDFs to it and make it write the part. No one fact checks. Students; they use it to write reports, homeworks and code.
GPT may be good for learning, but not for total beginners. That is key. As many people stated here, it can be good for those with experience. Those without, should seek for those experienced people. Then, when they have the basics they can get help from GPT to go further.
It is very hard to do testing on theoretical side of things. It is also possible to cheat on experimental results. On top of that, it eliminates the possibility of redoing (or publishing previously done results) if there is no significant contribution.
I think academic publishing is not in a good place right now and should be improved, but I don't personally see any problem with the concept. Thank you for your insights on the matter.