RTFM
haha, I have been saying LLMs have no value outside of entertainment and possibly some time to market benefit on first run stuff. Eventually humans always get it better and iterative time to market I don’t think shifts left with AI assistance.
ChatGPT was the oh my fuck we can juice consumers moment for ML/AI, what people realize is it’s not nearly as valuable and the pure entertainment value of it isn’t worth it for the consumer at its current astronomical operating cost and development cost. Effectively if the model isn’t constantly improved and retrained it gets “boring” to the human using it which gets them to stop using it and stop paying for it which makes the consumer use case moot.
all the commercial or business use cases are effectively trying to sell the new word processor to companies who may or may not even need to write stuff in a word processor. The traditional ML/AI use cases to do things like tuning and optimization are also just taking a backseat here for a lot of folks and those are things that actually deliver real and measurable business value and impact.
having NPUs in consumer equipment is not a new thing, but everyone is acting like now that they are in laptops it’s gonna change the way everyone uses computers and it’s just laughable. Make a bot that can do complex shit that a human needs without them having to learn excel and you’ve got something of extreme value, the problem is that’s not what they are even trying to solve they are trying to gimmick their way to success with brain tingles instead of measurable and demonstrable improvement of efficiency and work quality.