3 ms·
LLM sizes have been increasing 10X every year for the last few years. Source is NVidia. That pace is staggering. GPT4’s lead is temporary and it’s a shiny ball
by mmaunder 4y ago
LLM sizes have been increasing 10X every year for the last few years. Source is NVidia. That pace is staggering. GPT4’s lead is temporary and it’s a shiny ball distraction from doing dev on your own projects rather than being a user on someone else’s. Get to it and get over it. There are plenty of breakthroughs to be had.
- api 4y agoThe work being done to make these LLMs possible to run locally is terribly important. The same needs to happen for training. The most urgent breakthrough we need is distributed training algorithms that allow training to be done in the style of Seti@Home or even more cooperatively.