razorguymania

9 ms·

1. 1.▲
   by razorguymania 3y ago

   https://wow.groq.com/hey-elon-its-time-to-cease-de-grok/
2. 2.▲
   by razorguymania 3y ago

   See this reply: https://news.ycombinator.com/item?id=39437239
3. 3.▲
   by razorguymania 3y ago

   yes
4. 4.▲
   by razorguymania 3y ago

   I think this post is output from the chatbot itself
5. 5.▲
   by razorguymania 3y ago

   GPUs can't scale single user performance beyond a certain limit. You can throw 100s of GPUs at it but the latency will never be as good.
6. 6.▲
   by razorguymania 3y ago

   Its using vanilla llama-2 from Meta with no fine tuning. The point here is the speed and responsiveness of the underlying HW and SW.
7. 7.▲
   by razorguymania 3y ago

   Its not using GPUs. Its using Groq's own HW - LPUs