NEWS IN BRIEF: AI/ML FRESH UPDATES

Get your daily dose of global tech news and stay ahead in the industry! Read more about AI trends and breakthroughs from around the world

Mastering the Art of LLM Selection

Selecting the right large language model (LLM) is crucial but challenging. Moving from subjective vibes to comprehensive metrics is essential for accurate evaluation and model selection.

AI Sora Videos: Legal Experts Sound Alarm

Video app creates realistic deepfakes of Marx and MLK Jr. Some question use of 'historical figures' for legal testing. Night out with Henry VIII, DJing with Princess Diana, and shopping with Karl Marx in engaging scenarios.

Peak Power Rewards: Lowering US Bills

'Demand flexibility policy could ease strain on grid caused by AI datacenters, lower power bills, prevent blackouts, and help climate by rewarding customers for using less power or selling excess energy from solar panels.'

Revolutionary software creates recyclable clothing

MIT and Adobe's Refashion software aims to reduce textile waste by allowing users to design reconfigurable clothing items, adapting to trends and body changes. The tool breaks down fashion design into modules, enabling easy customization and prototyping of versatile garments in just 30 minutes.

GOP Releases Deepfake Attack Ad Targeting Chuck Schumer

The National Republican Senatorial Committee released a dystopian attack ad featuring a deepfake video of Chuck Schumer. The AI-generated Schumer robotically repeats a phrase about the government shutdown, with a small disclaimer acknowledging its artificial origins.

Enhancing Strategies with Amazon Bedrock Iterative Fine-Tuning

Amazon Bedrock now supports iterative fine-tuning for AI models, allowing incremental adjustments and continuous improvement without starting from scratch. This approach mitigates risks, optimizes performance based on real feedback, and accommodates evolving business requirements.

Personalized Object Detection: A Breakthrough in Generative AI

MIT and MIT-IBM Watson AI Lab researchers introduce a new training method to teach vision-language models to localize personalized objects in a scene, outperforming state-of-the-art systems. This approach could help AI systems track specific objects over time and assist visually impaired users in finding items in a room.

Enhancing Document AI with Amazon Nova Models

Fine-tuning Amazon Nova Lite for document processing tasks can significantly improve accuracy and reduce costs, especially for specialized documents like tax forms. The challenges of document processing, such as complex layouts and data quality variations, can be overcome by leveraging vision LLMs for structured information extraction.

Remembering John Searle

John Searle's 1984 lecture challenged the concept of AI thinking. Researchers continue exploring meaning and learning in AI today.