I found @simon’s summary very helpful in pulling together what’s happened in LLMs in 2024. Worth reading, particularly if you’re also following more skeptical voices like @timnitGebru @baldur @emilymbender
I downloaded Simon’s llm CLI and was impressed (and overwhelmed) by the breadth and scope of models out there. I tried to use some of the latest tools (Cursor, Gemini 1.5) to write code and analyze videos and was sadly unimpressed with current capabilities. https://me.dm/@anildash/113757562876292774
How much better is Gemini Advanced compared to regular #gemini ? I was trying to have it summarize some YouTube videos and was singularly unimpressed. Am I doing something wrong? #llm #genai
Every time I raise an example like this, I inevitably get told I am not using the latest models (“did you see the latest frontier math results?”). Also, even if this is true today, it won’t be true tomorrow because LLMs are moving so fast. https://mathstodon.xyz/@jonmsterling/113555911887640230
Every year I give a little to grow STEM and other literacy in underserved communities through @donorschoose — will you help fund this library in Texas? https://www.donorschoose.org/project/a-universe-of-possibilities/8759955/?utm_source=dc&utm_medium=directlink&utm_campaign=project&utm_term=donor_330136&rf=directlink-dc-2024-12-project-donor_330136
Trying to write some code with Copilot reminds me a bit of the Super Size Me documentary. “Regular or copilot?” “I think I’m going to have to go copilot!” https://youtu.be/as2zMlxeOkw