marius.vision - Posts in ai

Collapse posts

Showing posts tagged with: ai

Math is important for AI, but not the way you think

I.

So we all know AI is taking over coding. It's obvious, and undeniable. Both the quality and speed of production for AI generated code is currently very high, and it's only getting better. I can make frogger in a one-shot prompt on my GPU, and from what people are saying Claude can basically whip up a spinning hexagon program that cures cancer in under 20 ...

Opening the AI Museum

So benchmarking LLMs is kind of an unsolved problem. The metrics used to evaluate models are either too narrow or too gameable. Cross-entropy may be a useful statistic to tune a training run, but in practice it doesn't tell me if a model can understand my amazon e-mails or write cool haikus. Practical approaches exist, like the LM Arena. This is going in the ...