Hardware
High Gear for Llama 3.1 405B: SambaNova boosts Llama 3.1 performance with fast, free access to largest model
SambaNova raised the speed limit for access to the largest model in the Llama 3.1 family — and it’s free.
Analytics DeepLearning.AI
Machine Learning Research
OpenAI Forges Chains of Thought: OpenAI’s o1 models excel in reasoning, outperform GPT-4o in math and coding
Preliminary versions of OpenAI’s new model family were trained explicitly to think step-by-step, yielding outstanding marks in math, science, and coding — but users can’t see their reasoning steps.
Data Points
A new language model tool for web scraping and conversion: Plus, a plan to combat AI sexual abuse imagery
Hugging Face open-sources an LLM evaluation suite, Adobe announces its Firefly Video model, Meta researchers blend image diffusion with text transformers, NotebookLM can now generate synthetic podcasts.
Data Points
OpenAI’s o1 models recognize and fix mistakes: Plus, explaining Reflection 70B’s replication controversy
Copilot adds fine-tuning for faster code completion, DataGemma uses RAG and RIG for fact-retrieval, Mistral introduces its open multimodal model, results of the latest summit on military AI.
Letters
Why Science-Fiction Scenarios of AI’s Emergent Behavior Are Likely to Remain Fictional: The sudden apparance of “emergent” AI capabilities may be an artifact of the metrics you study
Over the weekend, my two kids colluded in a hilariously bad attempt to mislead me to look in the wrong place during a game of hide-and-seek.
The Batch Newsletter
Nations Sign Binding AI Treaty, Waymo Reveals Safety Record, 2D to 3D Goes Mainstream, Balancing Web Data Distributions
The Batch AI News and Insights: Over the weekend, my two kids colluded in a hilariously bad attempt to mislead me to look in the wrong place during a game of hide-and-seek.
Machine Learning Research
2D-to-3D Goes Mainstream: AI systems from Stability AI and Shutterstock transform 2D images into 3D meshes in seconds
Traditionally, building 3D meshes for gaming, animation, product design, architecture, and the like has been labor-intensive. Now the ability to generate 3D meshes from a single image is widely available.
Machine Learning Research
Balancing Web Data Distributions: Automated method organizes large datasets for better model performance
Datasets that were scraped from the web tend to be unbalanced, meaning examples of some classes (say, cats) are plentiful while examples of others (say, caterpillars) are scarce.
Tech & Society
Western Powers Sign AI Treaty: International AI treaty sets standards to support innovation and human rights
The European Union, United Kingdom, United States, and other countries signed a legally binding treaty that regulates artificial intelligence.
Business
Waymo Spotlights Safety Record: Waymo claims its robotaxis are safer than human drivers, citing new safety data
Waymo, the autonomous vehicle division of Alphabet, released an analysis of its own safety data. It suggests that the company’s self-driving cars are safer than human drivers on the same roads.
Data Points
Replit Agent builds and deploys applications using natural language prompts: DeepSeek-V2.5’s open model blends coding and chat
NVIDIA’s Blackwell chips impress on hardware tests, fine-tuned versions of Llama 3.1 add reflection, most AI jailbreaks may not amount to much, new architecture extends context windows to 100M tokens.
Data Points
LAION cleans up its image dataset: Plus, OLMoE outcompetes smaller open models
AlphaProteo, a DeepMind system that designs novel proteins, updates and price drops for Command-R and Command-R+, Anthropic shows off easy software projects in Claude, YouTube builds system to detect synthetic music and faces.