News for hackers. Stuff that matters.
Re:Gemini Ultra Released by mark_l_watson on Thursday February 08, 2024 @11:19AM (#39303768)
Good comments. As much as I am personally engaged in small LLMs that I can run on my own computer, and integrate into software that I write for myself, I think the future of large scale adoption of AI belongs to Google, Microsoft, and Apple in western countries (and China is doing wonderful things in their markets).

The old Bard/Gemini integration with Gmail, Google Docs, etc. is pretty good. I signed up for a minute for Microsoft’s $20/month AI Office integrations, but cancelled and will try again in 2 months. I am an Apple customer and I expect spectacular things from Apple. I expect Apple, Google, Samsung, etc., to offer fantastic on device AI.

I would like to see a money saving Google bundling family plan. I find Apple’s super bundle family plan a pretty good deal.

Re:Gemini Ultra Released by msoad on Thursday February 08, 2024 @12:28PM (#39304797)
small LLM? Small Large Language Model lol
Re:Gemini Ultra Released by mark_l_watson on Thursday February 08, 2024 @02:38PM (#39306561)
Good joke, thanks, but I will explain anyway: to me 30 billion parameters or smaller is small since I can run it using Ollama on my home computer. I managed a deep learning team at Capital One and our ‘large’ models were less than 20 million parameters. :-)
Re:Gemini Ultra Released by falcor84 on Thursday February 08, 2024 @03:12PM (#39307015)
I suppose we could call them Medium Language Models, but unfortunately that TLA is already taken
Re:Gemini Ultra Released by IanCal on Thursday February 08, 2024 @02:40PM (#39306572)
A large language model the size of a small language model.