AI News

Westlake University's Natural Language Processing Lab has released the DeepScientist system, the first AI scientist with complete scientific research capabilities. It can proactively identify research limitations, propose new ideas, write code, execute experiments, and write papers without human intervention. In AI text detection tasks, DeepScientist achieved in two weeks what humans take three years to accomplish, achieving a 7.9% AUROC improvement, surpassing the state-of-the-art (SOTA) human solutions.

OpenAI has partnered with streaming music platform Spotify, allowing users to log in to their accounts by mentioning Spotify on the ChatGPT web or mobile app and receive personalized music and podcast recommendations. ChatGPT will then use Spotify controls based on the conversation to recommend songs, artists, albums, etc., and clicking on them will automatically open the Spotify app.

The Bailing team officially released Ling-1T, the first flagship non-thinking model in the Ling 2.0 series. The model boasts trillions of parameters, based on the Ling 2.0 architecture, with each token activating approximately 50 billion parameters. Ling-1T was pre-trained on a high-quality, high-inference-density corpus of over 20T tokens, supports a maximum of 128K context windows, and significantly enhances its efficient reasoning capabilities through an Evolutionary CoT (Evo-CoT) approach combining mid-training and post-training, achieving state-of-the-art (SOTA) performance on multiple complex inference benchmarks.