Not big tech (or PhD level research), but half the work I did on my side project (subtitles for Chinese learning/OCR) is sort of obsolete now, most of the rest

Not big tech (or PhD level research), but half the work I did on my side project (subtitles for Chinese learning/OCR) is sort of obsolete now, most of the rest of it within a year or two. I put months into an NLP pipeline to segment Chinese sentences, classifying pinyin and translating words in-context, something ChatGPT is great at out the box. My painstaking heuristic for determining show difficulty using word frequencies and comparing distributions to children's shows is now the simple task of giving part of the transcript and asking ChatGPT how difficult it is. Next up, the OCR I did will probably be solved by ChatGPT4. It seems the writing is on the wall: most tasks on standard media (text/images/video), will be "good enough" for non-critical use. The only remaining advantage of bespoke solutions is speed and cost and that will also be a fleeting advantage. But it's also extremely exciting, we'll be able to build really great things very easily, and focus our efforts elsewhere. Today anyone can throw together a language learning tutor to rival Duolingo. As long as you're in it for solving problems you shouldn't be too threatened by whatever tool set you're currently becoming obsolete.