Meta, Facebook's parent company, has just launched a new AI model with self-learning evaluation (STE) feature, a remarkable step forward in the field of artificial intelligence.
Meta previously introduced this model in August, using a “thought chain” similar to the o1 model recently released by OpenAI, which helps the algorithm make more reliable judgments.
The principle of this technique is to break down complex problems into smaller logical steps, which has the potential to improve response accuracy in difficult fields such as science, programming, and mathematics.
 |
| Meta has announced a slew of new AI products |
One of the major challenges facing current AI models is their reliance on human feedback (RLAIF – Reinforcement Learning from AI Feedback) to learn. This process requires the involvement of experts to label and verify data, which is not only costly but also time-consuming.
Meta isn’t the only company pursuing reinforcement learning from AI feedback. Other tech giants like Google and Anthropic have also done their own research in this area. But Facebook’s parent company is the first to publicly release AI models that can “self-assess learning.”
“We hope that as AI approaches superhuman levels, it will be able to self-check its own work,” said Jason Weston of Meta's research team.
Additionally, Meta has also released an update to its Segment Anything image recognition model, which speeds up the generation of data that can be used in the research of new inorganic materials.
Last week, Meta also partnered with Blumhouse Productions — the film production house known for horror franchises like “The Purge” and “Get Out” — to put its Movie Gen video-generating AI model to the test.
Movie Gen is said to be able to create realistic video and audio that can compete with services from other industry-leading startups like OpenAI and ElevenLabs.
Source
Comment (0)