Meta to Launch AI Models with Capability to Check Work by Other Models

Last week, Meta Platforms Inc. (NASDAQ: META) announced that it’d be launching new artificial intelligence models from its research division. The new models, which would include an evaluator that’s self-taught, may reduce the involvement of humans in the AI development process.

They also support Meta’s objective of achieving advanced machine intelligence while also supporting reproducibility and open science.

The launch follows the company’s introduction of the tool in a paper which explained how it depended on the same technique used by the recently rolled out o1 models to make reliable judgments about responses given by a model. This technique breaks down issues into smaller steps and may enhance the accuracy of responses on difficult problems in subjects like math, coding, and science.

Researchers at Meta under Fundamental AI Research (FAIR) used data generated from artificial intelligence to train the evaluator model, which did away with human input at that stage. The FAIR team is committed to using advanced machine intelligence to power innovation and products for the benefit of every individual.

Two researchers behind the project explained in an interview that the use of AI in assessing itself provided a glimpse into a potential pathway toward designing autonomous artificial agents that can learn from their own errors.

It is expected that self-improving models will eliminate the need for reinforcement learning from human feedback, which is a costly and inefficient process that needs input from humans who must have the expertise required to accurately label data and verify that answers to writing and math queries are correct.

One Meta researcher, Jason Weston, hopes that as artificial intelligence becomes more super-human it will get better at verifying its work so that it’s better than the average human. Weston adds that the idea of being able to self-evaluate and be self-taught is pivotal to getting artificial intelligence to the super-human level.

Other artificial intelligence tools released by Meta include a tool that speeds up the generation times of large language models’ responses, an update to its Segment Anything image-identification model, and datasets that can assist in discovering new inorganic materials.

Other firms that have published research on Reinforcement Learning from Artificial Intelligence Feedback include Anthropic and Google. Unlike Meta though, these firms haven’t released their models for use by the public.

FAIR researchers hope that their work will inspire iteration and help advance artificial intelligence in a responsible way, helping bring the world closer together. They believe that access to state-of-the-art AI creates opportunities for all.

About AINewsWire

AINewsWire (“AINW”) is a specialized communications platform with a focus on the latest advancements in artificial intelligence (“AI”), including the technologies, trends and trailblazers driving innovation forward. It is one of 70+ brands within the Dynamic Brand Portfolio @ IBN that delivers: (1) access to a vast network of wire solutions via InvestorWire to efficiently and effectively reach a myriad of target markets, demographics and diverse industries; (2) article and editorial syndication to 5,000+ outlets; (3) enhanced press release enhancement to ensure maximum impact; (4) social media distribution via IBN to millions of social media followers; and (5) a full array of tailored corporate communications solutions. With broad reach and a seasoned team of contributing journalists and writers, AINW is uniquely positioned to best serve private and public companies that want to reach a wide audience of investors, influencers, consumers, journalists, and the general public. By cutting through the overload of information in today’s market, AINW brings its clients unparalleled recognition and brand awareness.

AINW is where breaking news, insightful content and actionable information converge.

To receive SMS alerts from AINewsWire, text “AI” to 888-902-4192 (U.S. Mobile Phones Only)

For more information, please visit www.AINewsWire.com

Please see full terms of use and disclaimers on the AINewsWire website applicable to all content provided by AINW, wherever published or re-published: https://www.AINewsWire.com/Disclaimer

AINewsWire
Los Angeles, CA
www.AINewsWire.com
310.299.1717 Office
Editor@AINewsWire.com

AINewsWire is powered by IBN