Foundation Models - News Articles

Structured Guides Boost Coding Productivity Over Raw Documentation

Sep 11, 2025

LangChain Blog

Structured guides like Claude.md significantly boost coding productivity and quality compared to raw documentation, with the best results achieved by combining a condensed guide with access to comprehensive documentation.

Coding & Developer Tools Foundation Models

AI Models Lack Human-Level Language Understanding, Capabilities Emerge

Sep 11, 2025

MachineLearningMastery.com

Powerful AI language models demonstrate remarkable abilities, including reasoning and coding, yet still lack the deep language comprehension of humans, relying on statistical pattern matching rather than true understanding, while smaller models sometimes outperform larger ones on specific tasks.

Foundation Models Research

Floating-Point Math Quirk and Batch Variability Affect LLM Outputs

Sep 10, 2025

Thinking Machines Lab

A quirk in floating-point arithmetic and batch variability in large language models (LLMs) can lead to inconsistent outputs, but achieving deterministic inference through batch-invariant kernels for key operations like RMSNorm, matrix multiplication, and attention resolves these issues.

Research Foundation Models

Groundbreaking AI Model Learns from Physical World Interactions

Sep 10, 2025

GitHub

A pioneering artificial intelligence model, trained on real-world physical interactions, learns to adapt and generalize through a feedback loop between its virtual embodiment and the environment, with open-source WALL models enabling further research and applications.

Robotics Foundation Models Open Source

Lilly Unveils TuneLab: AI-Driven Drug Discovery Platform with $1 Billion Research Investment

Sep 10, 2025

Eli Lilly and Company

Eli Lilly unveils TuneLab, an AI-driven drug discovery platform backed by a $1 billion research investment, offering biotech companies access to AI models built on Lilly's extensive drug data and joining Lilly Catalyze360's offerings for biotech partners.

Pharma Foundation Models Funding

Anthropic Pays Record $1.5 Billion to Settle AI Book Piracy Claims

Sep 09, 2025

The Authors Guild

Anthropic pays a record-breaking $1.5 billion settlement for using around 500,000 pirated books to train its AI systems, destroying all copies and prohibiting future use without permission.

Copyright & IP Foundation Models Legal

Google Expands AI Mode to 5 New Languages, Embracing Multimodal AI Search

Sep 09, 2025

TechCrunch

Google embraces multimodal AI search, expanding AI Mode with advanced Gemini 2.5 AI to 5 new languages - Hindi, Indonesian, Japanese, Korean, and Brazilian Portuguese - paving the way for an AI-led search experience as the default 'soon'.

Search Foundation Models Geopolitics

GPT-5 Unleashes Powerful 'Research Goblin' for Comprehensive Online Searches

Sep 08, 2025

Simon Willison’s Weblog

The latest GPT-5 update introduces a powerful 'Research Goblin' feature that excels at comprehensive online searches, analyzing results, and iteratively refining queries, though its trustworthiness remains uncertain.

Foundation Models Search Consumer Products

AI Models Struggle with Hallucinations, Researchers Propose Evaluation Updates

Sep 08, 2025

TechCrunch

Large AI language models continue to hallucinate plausible but false statements, prompting researchers to propose updated evaluations that discourage confident guessing by penalizing confident errors more severely and rewarding expressed uncertainty.

Research Foundation Models AI Safety

Anthropic AI Pays $1.5 Billion to Settle Authors' Copyright Lawsuit

Sep 08, 2025

NPR

In a landmark settlement, Anthropic AI pays $1.5 billion to resolve copyright claims from authors over an estimated 500,000 books used to train its AI models, compensating writers around $3,000 per book.

Copyright & IP Foundation Models Legal