A Harvard study finds ChatGPT-like AI diagnoses cancer with 96% accuracy, pioneering the first pathology-based model to predict patient outcomes globally.
Meta has introduced Meta Spirit LM, an open-source language model that enhances natural speech generation and supports ASR, TTS, and speech classification.
Researchers at EPFL discover that in-context learning effectively matches fine-tuning for large language models, particularly when data is limited.
A recent study reveals that safety features in large language models are more easily bypassed using keywords tied to marginalized groups, exposing hidden vulnerabilities.
A recent study finds that OpenAI’s latest o1-preview AI struggles with complex travel planning, identifying two main shortcomings and suggesting ways to improve its performance.
Swedish researchers find that improper use of AI tools like ChatGPT may exacerbate learning difficulties in students, particularly those already facing challenges.
Meta’s VP of AI, Joëlle Pineau, announced the company is developing unique AI reasoning techniques, setting them apart from OpenAI’s math-focused o1 model.
Former OpenAI CTO Mira Murati is reportedly seeking over $100 million to launch her own AI company, joining other ex-OpenAI leaders venturing into proprietary model development.
Researchers unveil Janus, a cutting-edge AI model that excels in both understanding and generating images through its innovative architecture.
**Google restructures, merging Gemini team into DeepMind to speed AI**
After briefly ousting Sam Altman as OpenAI’s CEO in November 2023, Microsoft is reevaluating its investment strategy, leading OpenAI to pursue alternative funding and computing resources.
Nvidia enhances Meta's Llama model with an innovative training method that merges evaluation and preference models, achieving top performance in alignment benchmarks.
