New Delhi, October 21: Meta reportedly introduced a new series of advanced AI models to improve the evaluation of artificial intelligence systems. The development comes from Meta's research division, known as FAIR (Fundamental AI Research). Among these innovations is a model, which is reportedly called the “Self-Taught Evaluator”, which can assess the performance of other AI models with less human input. Additionally, the “Spirit LM” model is expected to enhance the integration of speech and text to make interactions more natural.
As per a report of Indian Express, Meta has launched an AI model that can evaluate other AI models’ work. The Spirit LM is another open-source language model developed to smoothly integrate speech and text. The recent announcements follow a paper released by Meta in August, which explained that their new models would use a method called the “chain of thought.” Meta and MSDE Partner To Integrate AI, VR, MR Into Skill India Ecosystem at National Skill Training Institutes Across Multiple Cities.
Meta Self-Taught Evaluator
The new models that have been introduced include several updates and new features. These consist of the improved Segment Anything Model 2, which works with images and videos, as well as the Self-Taught Evaluator, Meta Spirit LM, Layer Skip, SALSA, Meta Lingua, OMat24, and MEXMA. The Self-Taught Evaluator is a new approach that creates its own data to train reward models without needing humans to label the information.
The model produces various results from different AI systems and then uses another AI to evaluate and enhance those results. The process continuously improves over time. According to Meta, the model is powerful and performs better than other models that depend on data labelled by humans, like GPT-4 and similar systems. Elon Musk’s Private Jet Tracking Accounts Still Active on Mark Zuckerberg-Run Meta Platforms, While Taylor Swift’s Tracker Suspended.
Meta Spirit LM
Meta has reportedly created Spirit LM, its first open-source model that can handle both text and speech in a more natural manner. Additionally, it can learn new tasks related to automatic speech recognition (ASR), text-to-speech (TTS), and speech classification. Meta has developed two versions of this model. The Spirit LM Base is made to focus on the sounds of speech, while the Spirit LM version is developed to capture the tone and emotion in speech.
(The above story first appeared on LatestLY on Oct 21, 2024 03:11 PM IST. For more news and updates on politics, world, sports, entertainment and lifestyle, log on to our website latestly.com).