When you start following Machine Learning research

some people aren’t happy unless they live in an active war zone

Claude 2.1 (200K Tokens) - Pressure Testing Long Context Recall We all love increasing context lengths - but what's performance like? Anthropic reached out with early access to Claude 2.1 so I repeated the “needle in a haystack” analysis I did on GPT-4 Here's what I found:

Just the beginning 👀

There is so much to build. My god.

why not > why

Great analysis of llms artificialanalysis.ai

They should guarantee turbo models to always have < 30 day knowledge cutoff. gpt-4 turbo takes < 25 days to train

“Coding” was never the source of value, and people shouldn’t get overly attached to it. Problem solving is the core skill. The discipline and precision demanded by traditional programming will remain valuable transferable attributes, but they won’t be a barrier to entry. Many

Jason is a phenomenal person. He deserves all the recognition. We’ve got more discoveries to show the world… coming soon

Andres is a hero, Microsoft give him yolo runs

I have cardiology models more accurate than any cardiologist on my mac

Does this mean 1980??

Are prompts eligible for the The Nobel Prize in literature?

ai is going to take their jobs

Interesting read, excited to see how it scales

first upscaler I’ve used that doesn’t look terrible

Spotify knows how I spend my Saturday nights

Building APIs has ruined all 3 digit numbers for me

You don’t get to hate academic publishing unless you love it