Mike McMurdie’s Post

On this AI Appreciation day: for those interested in learning more about what happens behind the scenes on AI, I found this approach to using spreadsheets for GPT 2.0 very enlightening. Kudos to Ishan Anand!

View profile for Cannon Brooke, graphic

Advancement Data Sommelier | Leveraging data to help nonprofits turn insights into actionable results.

If you find LLMs intimidating or confusing, check out this GitHub from Ishan Anand. He managed to squeeze GPT 2.0 into a spreadsheet format. Yes, it's got its fair share of drawbacks—you can't exactly have a back-and-forth chat since it's all about entering prompts. However, it's still a pretty awesome way to glimpse what makes these models tick, especially the transformer architecture and how it decides the next word to throw at you (next-token prediction). Warning: The file's huge and doesn't play nice with Macs, so you'll need a PC to dive into this one. https://lnkd.in/gkuYsxvv #prospectresearch #fundraising #aifundraising

Release Spreadsheets-are-all-you-need v0.5.0 · ianand/spreadsheets-are-all-you-need

Release Spreadsheets-are-all-you-need v0.5.0 · ianand/spreadsheets-are-all-you-need

github.com

Ishan Anand

VP, Product for the Edgio Applications platform

1mo

Thanks Mike and glad you found it educational! You might enjoy this talk I gave at the AI Engineer World's Fair in San Francisco last month. It uses the spreadsheets-are-all-you-need as a quick 20 min crash course in LLMs and mechanistic interpretability: https://www.youtube.com/watch?v=5zE2sMka620&t=27878s

To view or add a comment, sign in

Explore topics