Information

Today, December 12, OpenAI updated its incident report to say that the API, ChatGPT, and Sora services are now fully operational. The outage lasted approximately 4 hours and 10 minutes, and OpenAI said it will conduct a full root cause analysis and investigation into the outage. Earlier today, OpenAI confirmed that its chatbot ChatGPT was experiencing global downtime, and that Sora and its APIs were also affected. The downtime issue is also known to have affected iOS 18.2 ......

Recently歸藏 conducted a test of the AI video generation models of Pixverse, Pika and Runway. Overall, Pixverse and Pika were fairly close in terms of video generation quality, receiving high scores of 74.5 and 73.5 respectively, while Runway was relatively poor at 64.5. Pixverse's model is considered to be the most balanced of the three, able to handle a greater magnitude of motion while maintaining a good level of consistency. In particular, it performed amazingly well for 2.5D shots. However, for purely secondary images, Pixverse ......

December 16th, 2011 - No Questions Asked Core Dome announced today that it has open sourced Megrez-3B-Omni, the full-modal comprehension mini-model of No Questions Asked Core Dome's end-side solution, and its language-only model version, Megrez-3B-Instruct. Officially, Megrez-3B-Omni is a full-modal comprehension model for the end-side, with the ability to process three types of modal data: image comprehension, audio comprehension, and text comprehension. In terms of image understanding, Megrez-3B-Omni is currently the best model in OpenCompass, MME, MMMU, and O.......

Late Tuesday, Microsoft, Alphabet and Supermicro reported quarterly results that disappointed investors, causing AI-related companies to vaporize $190 billion in market value. The sell-off after these tech giants reported after the closing bell underscored the high expectations of investors, who have pushed the companies' shares to record highs in the AI-driven stock market rally of recent months, in anticipation of integrating the technology across the corporate sector. Google's parent company, Alphabet, saw its shares fall 5.6% after its advertising revenue missed expectations in December. Alphabet also said that in order to support ......

Dec. 20, 2011 - Large language models (LLMs), represented by ChatGPT and others, have become increasingly adept at processing and generating human language over the past few years, but the extent to which these models mimic the neural processes of the human brain that support language processing remains to be further elucidated. As reported by Tech Xplore 18, a team of researchers from Columbia University and the Feinstein Institute for Medical Research recently conducted a study that explored the similarities between LLMs and neural responses in the brain. The study showed that with advances in LLM technology, these models are not only performing ......

WHEE, a subsidiary of Meitu, has announced that the AI Ultra HD feature is now available, which allows you to restore old photos and generate ultra HD images with one click. Simply import the image, choose the style that suits the image, and you can generate ultra-high-definition images with one click. In addition to generic scenes, the function supports a variety of scene adaptations, such as portraits, 3D assets and natural scenery, and can adjust the parameters for picture redrawing and re-creation....

December 25, 2012 - OpenAI insiders recently revealed that the company is exploring the possibility of developing humanoid robots, The Information reported today. The outlet learned of the news from two people "with direct knowledge of the matter". However, details about the potential project are still scarce, and it's unclear when OpenAI plans to launch the robot, as well as its target application scenarios. Currently, the few startups developing humanoid robots are focusing on two application scenarios: first, automating tasks such as cargo handling within warehouses to improve ......'s performance.

Meta's second-generation in-house AI chip, Artemis, went into production this year, according to media reports. It is understood that the new chip will be used in data center reasoning tasks, working in tandem with GPUs from vendors such as NVIDIA. A Meta spokesperson previously said, We believe that our self-developed gas pedal will complement commercially available GPUs to provide the best balance of performance and efficiency for Meta's tasks. In addition to running recommendation models more efficiently, Meta will also need to provide arithmetic for its own generative AI applications, as well as for Llama3, the GPT-4 open source competitor it is training. ......