"if you don't build me a world eating data center my shitty autocomplete algorithm will run slow :(" meanwhile in the 90s if your code needed more than a few bytes of ram they just took you into the parking lot and shot you
kinda odd how people on social media don't think that <<progress>> is possible
eg NVIDIA chips that use 50x less power
or better cooling or wind/solar power
@failedLyndonLaRouchite @computer Are you implying that LLMs are progress compared to standard autocomplete?
yes, they are enormous progress
at least, that is what all the smart, hardworking people (engineers, PhDs) say at the biotech company where I worked last year
@failedLyndonLaRouchite @computer Smart people are not immune to advertising induced delusion.
LLMs are not a good technology for the tool they are marketed as. Every tool they have been marketed as has been done before for cheaper & with better performance. The autocomplete on my phone keyboard has been flawless for ten years. My shell can read $PATH, the working directory, & a bunch of prewritten completions files to make it so it can finish whatever I'm typing all by itself when I press the tab key. My text editor reads open documents for words I've typed & suggests a few guesses on what I'm typing right now so I don't have to copy & paste or type it all out again. Search engines have been incredibly useful for decades, only failing when LLMs were integrated & when randomly generated text by LLMs polluted their databases. They can't replace spreadsheets at all either. They just don't work for anything they're advertised for, it's just not what the technology does.
@failedLyndonLaRouchite @computer They're only any good for the statistics work they were designed for. Even then it's better to just use the original data that you already have.