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

Optimizing GPT-2: Hardware Edition

Summary: Learn how to optimize hardware for faster GPT-2 training on NVIDIA GPUs, with insights on timing code and setting batch sizes for maximum efficiency. Achieve significant speed gains (up to 10x) using an Ampere-series Nvidia GPU.

Rising Scam Threat: Brace Yourself

AI is making it easier for scammers to fool anyone, even tech-savvy individuals like Arwa Mahdawi in NYC. The story highlights the growing sophistication of AI in perpetuating scams, posing a challenge for individuals to stay vigilant.

Measuring Success: Classification Model Metrics

Machine learning model predictions in credit card fraud detection evaluated using confusion matrix and metrics. Understanding true positives, false positives, false negatives, and true negatives crucial for model performance assessment.

Striving for Gold: Olympic Performance Optimization

MIT startup Striv developed tactile sensing technology for shoe inserts, used by elite athletes like USA marathoner Clayton Young and Jamaican Olympian Damar Forbes. Founder Axl Chen aims to bring this tech to the public after Paris 2024 Olympics, following success in VR gaming and interest from various industries.

ChatGPT wows testers with advanced voice mode

OpenAI introduces Advanced Voice Mode for ChatGPT Plus subscribers, enabling natural, real-time conversations with AI. Users impressed by feature's responsiveness, emotional cues, and realistic voice simulations.

AI Humility: Preventing Overconfidence in Wrong Answers

Researchers from MIT and the MIT-IBM Watson AI Lab have developed Thermometer, a calibration method tailored to large language models, ensuring accurate and reliable responses across diverse tasks. Thermometer involves building a smaller model on top of the LLM, preserving accuracy while reducing computational costs, ultimately providing users with clear signals to determine a model's reliability.

Revolutionizing Japanese LLMs with AWS Trainium

AWS Japan's LLM Development Support Program aids innovative companies in leveraging large language models (LLMs) to drive progress and boost productivity. Ricoh's bilingual LLM training strategy showcases how organizations are transforming possibilities with generative AI on AWS.

Google Play's Hidden Malware Family

Android malware Mandrake resurfaces in Google Play after years of stealth, targeting victims with intricate spying activities. Bitdefender unveils Mandrake's advanced evasion tactics, including kill switch and decoy apps, affecting tens of thousands of users.