3 ms·

At least, people working on ML models that handle these tasks must be feeling terrible. They know their models will be abandoned sooner or later and composed on

At least, people working on ML models that handle these tasks must be feeling terrible. They know their models will be abandoned sooner or later and composed on top an LLM. 1. Classification 2. Named Entity Recognition (NER) 3. Dialog Engine 4. Sentiment Analysis 5. Tone Analysis 6. Language Translation 7. Summarization 8. Tokenization 9. Simple NLP Tasks (part-of-speech tagging, dependency parsing, lemmatization, morphological analysis) 10. Sentence Segmentation 11. Content Parsing 12. Question Answering (Structured & Unstructured) 13. Similarity 14. Grammar Correction 15. Speech to Text (ASR) 16. Text to Speech (TTS) 17. OCR 18. Image Recognition 19. Text Test Data Generation

espe 4y ago
actually that might not be the case. don't underestimate the value of older, better understood and much smaller models. also, why not call bert-style (encoder) models LLMs as well. i would expect last-gen models to give us an edge in controlling the effects of the latest ones (cf. the alignment discussion).
- chopete3 4y ago
  BERT models are also LLMs. I referred to LLM more as an API based access, hosted by Microsoft, Google or AWS for large scale isolated/production consumption, like RDS (MySQL, Postgres). There will always be custom models, with controlled training data and specific use cases.
lysecret 4y ago
Couldn't agree more. I see a lot of parallels to the move to cloud. Yes, some big comps will still have an AI engineering team and build their own models but the vast majority will move to an API. The same way the vast majority used to have a sys admin type who set up and maintained a local server who then moved to the cloud. Of course the SuperStar AI engineers will now find much more interesting work at the big comps, just like there are super interesting challanges building the cloud for everyone.

espe 4y ago

actually that might not be the case. don't underestimate the value of older, better understood and much smaller models. also, why not call bert-style (encoder) models LLMs as well. i would expect last-gen models to give us an edge in controlling the effects of the latest ones (cf. the alignment discussion).

chopete3 4y ago
BERT models are also LLMs. I referred to LLM more as an API based access, hosted by Microsoft, Google or AWS for large scale isolated/production consumption, like RDS (MySQL, Postgres). There will always be custom models, with controlled training data and specific use cases.

lysecret 4y ago

Couldn't agree more. I see a lot of parallels to the move to cloud. Yes, some big comps will still have an AI engineering team and build their own models but the vast majority will move to an API. The same way the vast majority used to have a sys admin type who set up and maintained a local server who then moved to the cloud. Of course the SuperStar AI engineers will now find much more interesting work at the big comps, just like there are super interesting challanges building the cloud for everyone.