3 ms·
Six questions will dictate the future of generative AI
- ProllyInfamous 3y agoThis MIT Press article was quite the review of 2023's genAI layperson boom! I particularly enjoyed Ilya's pre-Nov30 ChatGPT release comments that "it's not very good, and makes a lot of mistakes" [talking about GPT3.5 backend]. Also interesting were the links to 2 other studies; the Boston Consulting Group study shows that among these elite consultants, the bottom half [IQ wise] benefited +45% from access to GPT-4; the top half [IQ] still benefite +17% [completed tasks]. I have agreed with the basic premise of most of the paper authors, for about one year: these LLMs are NOT sentient, but are powerful in ways unknown to even the creators. Both myself and the paper authors seem to agree that JOB output will be, and already is, affected by publicly-accessible LLM-agents.
- o_nate 3y agoAlso, the fact that ChatGPT at least in its current iteration is more helpful for less knowledgeable users may have something to do with why OpenAI almost didn't release it. The elite engineers didn't get how revolutionary it would appear to the general public.