Meta has introduced new security tools to enhance the Llama AI models, aimed primarily at assisting cybersecurity teams and developers in utilizing AI for defense purposes. These updates are part of Meta’s initiative to improve safety in AI development and use. The new tools include Llama Guard 4, an advanced customizable safety filter, which is […]
The Fundamental AI Research (FAIR) team at Meta has introduced five new projects aimed at advancing their pursuit of advanced machine intelligence (AMI). These initiatives focus significantly on improving AI perception, language modeling, robotics, and collaborative AI agents. Meta’s goal is to develop machines that can acquire, process, and interpret sensory information with human-like intelligence […]
Meta has announced that it will leverage content shared by its adult users in the EU to train its AI models. This decision follows the recent launch of Meta AI features in Europe, aiming to improve the relevance and capabilities of its AI systems for the region. In their statement, Meta outlined, “Today, we’re announcing […]
One of Meta’s latest flagship AI models, Maverick, has achieved the second-highest score in the LM Arena, a benchmark where human raters compare AI-generated outputs. However, there’s controversy surrounding this ranking as the version of Maverick tested on LM Arena seems to differ from the publicly available version intended for developers. In its announcement, Meta […]
Meta has launched a new collection of AI models known as Llama 4, building on its previous efforts in AI development. This release features three new models: Llama 4 Scout, Llama 4 Maverick, and Llama 4 Behemoth. Each model has been trained on extensive amounts of unlabeled text, images, and videos to enhance their visual […]
In July 2024, Meta CEO Mark Zuckerberg stated that selling access to Meta’s openly available Llama AI models was not part of the company’s business model. However, a newly unredacted court filing reveals that Meta is indeed earning some revenue through revenue-sharing agreements with companies that host its Llama models. This information emerged from the […]
Meta is currently piloting an in-house chip designed specifically for training artificial intelligence (AI) systems. This initiative is part of their broader strategy to lessen their dependency on external hardware providers like Nvidia. The chip has been developed in collaboration with Taiwan Semiconductor Manufacturing Company (TSMC) and is expected to handle AI-specific workloads effectively. The […]
Court documents recently revealed that Meta employees have been discussing the use of copyrighted works obtained through potentially questionable means for training the company’s AI models. Unsealed court filings from the case Kadrey v. Meta illustrate that Meta believes their use of copyrighted material is covered under "fair use," a claim disputed by the plaintiffs, […]
Amazon has announced a significant plan to invest over $100 billion in AI for 2025, marking a considerable increase from its earlier projection of $75 billion for the same year. This spending primarily focuses on enhancing their Amazon Web Services (AWS) cloud division’s AI capabilities. CEO Andy Jassy highlighted that the fourth-quarter capital expenditures of […]
U.S. markets experienced a sharp downturn on Monday following concerns that DeepSeek’s AI models might diminish the demand for GPUs, leading to a nearly 20% drop in Nvidia’s stock. Despite this, Meta’s CEO, Mark Zuckerberg, reaffirmed the company’s commitment to invest significantly in AI, projecting an expenditure of "hundreds of billions of dollars" as part […]