Testing #GPT4o, good but mac app has inference latency issues. #multimodal is the future. #LLMs are commoditized. Fine tuned #industry specific models with #device level tuning inference + cloud generalization seems more likely. Will only get better #AI #OpenAI.
Using #LLMs is showing that its explaining only what is in training #corpus nothing new. No where near #AGI its just #probability distribution #token sampling. Easy with #maths #statistics #python. But very good tool for #learning #research. #Human creativity still required #AI
Building #industry specific solutions using all #AI #ML #algorithms #tools, not just #LLMs, looks like the differentiator. Users don’t care what’s under the hood. Solve the problem, build the #business, that’s my view.
Lot of #automation in most #industries coming #toolkit #algorithms #data are there, getting better. Any task which is just copy pasting documents, basic #analysis #coding #consulting support etc will be like manual #farming. #Social implications yes, but people will learn adapt.
