|
|
|
|
|
|
by sgt101
1191 days ago
|
|
|
Perhaps - but normally you'll have a narrowly defined and very specific technical topic/hypothesis that you're working on, and many/most of these aren't going to be closed off by ChatGPT4 Will this effect the job market (both academic and commercial) for these folks? It's very hard to say. Clearly lots of value will be generated by the new generation of models. There will be a lot of catchup and utilisation work where people will want to have models in house and with specific features that the hyperscale models don't have (for example constrained training sets). I'm wondering how many commercial illustrators have had their practices disrupted by Stable Diffusion? Will the same dynamics (what ever they are) apply for the use of LLM's? |
|
|
Pretty hard disagree. Even if your NLP PhD topic is looking at hypotheses on underlying processes about how languages work (and LLMs can't give you this insight), 9 times out of 10 it's with an eye for some sort of "applicability" of this for the future. GPT-4 just cut off the applicability parts of this for huge swaths of NLP research.