Artificial Intelligence #259

Hey, in this issue: China’s cheap, open AI model DeepSeek thrills scientists (learn how it works); Do generative video models learn physical principles from watching videos?; The Hundred-Page Language Models Book is now available to order; tool touted as 'first AI software engineer' is bad at its job; lessons learned from building an AI sales assistant; use LLMs in Excel formulas; and more.

• [Model] China’s cheap, open AI model DeepSeek thrills scientists
• [Explained] How Deepseek R1 was trained
• [ArXiv] Do generative video models learn physical principles from watching videos? Not yet
• The new book "The Hundred-Page Language Models Book" is now available to order

The Hundred-Page Language Models Book by Andriy Burkov, the follow-up to his bestselling The Hundred-Page Machine Learning Book (now in 12 languages), offers a concise yet thorough journey from language modeling fundamentals to the cutting edge of modern Large Language Models (LLMs). Within Andriy's famous "hundred-page" format, readers will master both theoretical concepts and practical implementations, making it an invaluable resource for developers, data scientists, and machine learning engineers.

• Tool touted as 'first AI software engineer' is bad at its job, testers claim
• [Nvidia] Lessons learned from building an AI sales assistant
• [Tool] TabPFN: Foundation model for tabular data
• [New Book] The Hundred-Page Language Models Book by Andriy Burkov
• [Tool] Cellm: Use LLMs in Excel formulas
• [MIT] MIT researchers introduce Boltz-1, a fully open-source model for predicting biomolecular structures

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.