OpenAI has announced its new models, o1-preview and o1-mini, which have improved capabilities in areas like maths and science. The models have been rated as having a medium risk of chemical, biological, radiological and nuclear weapon risk. They have exceeded human PhD-level accuracy on a benchmark of physics, biology, and chemistry problems, and even placed among the top 500 students in the US in a qualifier for the USA Math Olympiad.
Source: https://www.techmeme.com/river
Duplicated with: | 2024-09-14 08:51:06 AM |
Meta is updating how it labels content on Instagram, Facebook, and Threads, introducing a new menu in the top-right corner of images and videos edited with AI. The "AI Info" tag will appear within this menu, allowing users to check if AI information is available and read what may have been adjusted. The company previously applied the "AI Info" tag to all AI-related content, but is now making changes to better reflect the extent of AI used across images and videos on the platforms.
Source: https://www.techmeme.com/river
2024-09-14 08:51:05 AM |
Meta CEO Mark Zuckerberg appeared on the Acquired podcast at the Chase Center in downtown San Francisco, where he joked about needing to schedule his next appearance to apologize for something. He then stated that his days of apologizing are over. The event took place on Tuesday evening.
Source: https://www.techmeme.com/river
2024-09-14 08:51:04 AM |
The White House has extracted voluntary commitments from several major AI vendors, including OpenAI, Microsoft, Adobe, Cohere, and Common Crawl, to combat nonconsensual deepfakes and child sexual abuse material. The organizations have agreed to responsibly source and safeguard datasets, incorporate feedback loops in their development processes, and remove nude images from training datasets when appropriate.
Source: https://www.techmeme.com/river
2024-09-14 08:51:03 AM |
