Artificial Intelligence #260

Hey, in this issue: DeepSeek R1 and R1-Zero explained; AI and Machine Learning: A historical perspective; AI haters build tarpits to trap and trick AI scrapers that ignore robots.txt; how evaluates AI models and LLMs; how to align open LLMs with DPO & and synthetic data; state of open video generation models in Diffusers; and more.

The sponsor of this issue is Spice AI.

• AI and Machine Learning: A historical perspective
• [Ars Technica] AI haters build tarpits to trap and trick AI scrapers that ignore robots.txt
• [Explained] DeepSeek R1 and R1-Zero
• [Sponsored] Spice.ai Open Source 1.0-stable announced!

Spice.ai OSS is ready for production. It’s a single-node data and AI inference engine written in Rust and built on Apache DataFusion, Arrow, and DuckDB. It unifies SQL query, retrieval, and AI into one runtime for data-grounded apps across on-prem, cloud, and edge. Apache 2.0. Read the announcement.

• [Model] SmolVLM Grows Smaller – Introducing the 250M & 500M Models!
• [GitHub] How we evaluate AI models and LLMs for GitHub Copilot
• [Tutorial] How to align open LLMs with DPO & and synthetic data
• [New LLM Book] The Hundred-Page Language Models Book is available for order
• [Hugging Face] State of open video generation models in Diffusers
• [Project] TinyZero: Clean, accessible reproduction of DeepSeek R1-Zero

More than 850,000 subscribers are reading this newsletter. If you are building an AI or a data product or service, you can become a sponsor of one of the future newsletter issues and get your business featured in the newsletter. Feel free to reach out to [email protected] for more details on sponsorships.

Enjoy the newsletter? Please help us make it bigger and better by sharing it with colleagues and friends.