Google AI

How Google AI’s New Method Reduces LLM Training Data by 10,000x Google Research has unveiled a groundbreaking method for fine-tuning large language models (LLMs) that reduces required training data by up to 10,000 times. This innovative approach utilizes active learning to focus expert labeling on the most informative examples, especially “boundary cases” where model uncertainty is highest. In experiments with Gemini Nano models, this technique matched or surpassed the quality of models trained on 100,000 random labels with as few as 250 to 450 targeted examples. This development promises to make AI model development significantly leaner, more agile, and more cost-effective. ...

AI Discovers Stronger, More Durable Plastics in Materials Science Breakthrough Researchers at MIT and Duke University have utilized machine learning to develop significantly more durable plastics, a breakthrough that could lead to longer-lasting products and a reduction in plastic waste. The AI model identified novel crosslinker molecules, known as mechanophores, that can be integrated into polymer materials to make them more resistant to tearing. These mechanophores respond to mechanical force, allowing the material to withstand greater stress before breaking. After the AI model performed a rapid computational analysis of thousands of potential molecules, a task that would have been incredibly time-consuming with traditional methods, researchers synthesized a polymer with one of the identified candidates. The resulting material was found to be approximately four times tougher than plastics made with standard crosslinkers. This AI-driven approach to polymer science not only accelerated the discovery process but also uncovered new design principles for creating more resilient polymers. ...