Meta, the company behind Facebook, announced on Friday that it's releasing new artificial intelligence (AI) models from its research team. One of the highlights is a tool called the "Self-Taught Evaluator," which could reduce the need for humans in developing AI. This tool builds on a method introduced in an August paper, which helps the AI break down complex problems into simpler steps. This approach, similar to what OpenAI has used, aims to make AI more accurate in tasks such as science, coding, and math.
How is this model different?
Interestingly, Meta's researchers trained this evaluator using only data generated by other AIs, meaning no human input was needed at that stage. This technology might pave the way for AI systems that can learn from their own mistakes, potentially becoming more autonomous.
What are its benefits?
Many experts in the AI field dream of creating digital assistants that can perform a range of tasks without human help. By using self-learning models, Meta hopes to improve the efficiency of AI training processes that currently require a lot of human oversight and expertise.
Jason Weston, one of the researchers, expressed optimism that as AI becomes more advanced, it will improve its ability to check its own work, potentially surpassing human performance in some areas. He pointed out that being able to learn and evaluate itself is vital for reaching a higher level of AI capability.
Other companies, like Google and Anthropic, are also exploring similar concepts; however, they usually don’t make their models available for public use.
Alongside the Self-Taught Evaluator, Meta released other tools, including an updated version of their image-recognition model and resources to help scientists discover new materials.
Meanwhile, Meta is implementing changes to its Facebook monetization program by consolidating its three creator monetization initiatives into a single program. This new approach aims to simplify the earning process for creators on the platform.
Currently, creators can earn through In-stream ads, Ads on Reels, and Performance bonuses, each with distinct eligibility requirements and application procedures. With the revised monetization program, creators will only need to apply once, streamlining the onboarding process into a single, unified experience.
ALSO READ: Flipkart's Big Diwali Sale announced, offers impressive discounts on smartphones and more
Inputs from IANS
