On Saturday, Meta released the new open source Llama 4 series of models. As an open source model, Llama 4 has one major limitation - its massive size. The smallest Llama 4 Scout model in the series has 109 billion parameters, which is simply too large to run on a local system. But don't worry! Even if you don't have a GPU, we've found a way to use it for free through the web interface. If you have a powerful enough graphics card, you can also download the full model weights - this guide will explain all the ways to get them. 1. directly at Meta....

April 6, 2013 - Meta has released its latest Llama 4 series of AI models, including Llama 4 Scout, Llama 4 Maverick, and Llama 4 Behemoth, revealing that the models have been trained on "large amounts of unlabeled text, image, and video data" to give them "a wide range of visual comprehension capabilities," the company said. Meta reveals that the models have been trained on "large amounts of unlabeled text, image and video data" in order to provide them with "a wide range of visual comprehension capabilities". Meta has now uploaded Scout and Maverick from the series of models to Hugging Face (...

On April 2, the Guardian reported on March 31 that Richard Osman, Kazuo Ishiguro, Kate Moss, and Val McDermid were among the prominent authors who signed an open letter urging the British government to hold Meta accountable for using copyrighted books to train AI. The letter calls on the Secretary of State for Culture, Media and Sport, Lisa Nandy, to summon Meta executives to Parliament for questioning. Commenting on the reasons for signing, McDermid said, "The contract makes it clear that third parties must be paid for the use of an author's work. Whether it is an adaptation, translation or reproduction...

April 2, 2012 - Meta's head of artificial intelligence research, Joelle Pineau, announced on April 1, local time that she will leave the company on May 30th. Joelle Pineau joined Meta in 2017 and currently serves as vice president of AI research and head of FAIR, the basic artificial intelligence research department led by renowned AI expert Yann LeCun. Pino's departure comes at a time when Meta is investing more in AI. The company is said to be planning a 2025 ...

March 27, 2011 - Tech media outlet WinBuzzer published a blog post yesterday (March 26) reporting that new research shows Meta's involvement in the digital book piracy chain is far greater than expected. According to the report, Meta uses pirated books from LibGen, Z-Library and other "shadow libraries" to train AI models, and also re-uploads about 30% of downloaded books via BitTorrent, objectively prolonging the distribution cycle of pirated works. Experts point out that although BitTorrent is silent...

March 20, 2011 - After a nearly year-long hiatus, Meta's AI chatbot is finally returning to the European market. Starting this week, Meta AI will gradually roll out on WhatsApp, Facebook, Instagram, and Messenger in 41 European countries and 21 overseas territories, but is currently limited to text-based chat functionality. 1AI understands that Meta AI first went live in the U.S. in 2023, when the company planned to bring it to Europe, but due to Irish privacy regulator...

March 19, 2011 - Meta CEO Mark Zuckerberg announced on Threads that Llama, Meta's family of "open" artificial intelligence models, has reached 1 billion downloads, up from 650 million in December 2024, an increase of about 53% in just three months. Llama models are the core technology underpinning Meta AI, Meta's artificial intelligence assistant, which is widely used across Meta's platforms, including Faceboo...

France's leading publishers and authors' associations have filed a lawsuit against Meta, accusing it of unauthorized mass use of copyrighted content to train its AI models. The French National Publishing Union (SNE), the National Union of Authors and Composers (SNAC), and the Société des Grands Littéraires pour la Défense des Autres (SGDL) have filed a lawsuit against Meta, alleging copyright infringement and economic "parasitism," the three associations said in a press release on Wednesday. The three associations argued that as Facebook, Instagram and WhatsApp are social...

March 11 (Bloomberg) -- Social media company Meta, which owns Facebook, Instagram and WhatsApp, is testing its first in-house chip for training artificial intelligence systems, Reuters reported. According to two people familiar with the matter, the move marks a key step for Meta in reducing its reliance on outside vendors, such as NVIDIA, and moving toward designing more customized chips. Meta has begun deploying the chip on a small scale and plans to increase production for large-scale use if tests go well, the report said.Me...

March 9 (Bloomberg) - U.S. Federal Judge Vince Chhabria has ruled against an AI copyright lawsuit involving Meta, allowing the suit to move forward while dismissing some of the plaintiffs' claims. In Kadrey v. Meta, a number of authors, including Richard Kadrey, Sarah Silverman, and Ta-Nehisi Coates, accused the Meta public...

Social media giant Meta is accelerating the development of voice capabilities for its artificial intelligence, with plans to launch its latest open-source big language model, Llama 4, in the coming weeks, with voice interaction as one of its core features, according to a March 7 report in the Financial Times. Meta hopes to use this to drive the commercialization of AI technology and achieve a major breakthrough in its AI offerings by 2025, according to people familiar with the matter. Meta's development of voice features focuses on making users' conversations with AI closer to natural two-way conversations rather than traditional question-and-answer...

Feb. 28 - CNBC reported on Feb. 27 that Meta plans to launch a standalone AI app in Q2 this year as part of Zuckerberg's plan to overtake OpenAI and Google by the end of the year. Unlike rival AI tools like ChatGPT and Perplexity, Meta AI is currently limited to use through its website and its apps like Facebook and WhatsApp. While Meta's large user base can access these apps through...

In court filings this week, Meta claimed that although it downloaded 82 terabytes of pirated copyrighted material from a "shadow library" for use in training its LLaMA artificial intelligence models, company employees took steps to ensure that none of the files were "seeded" during the download process. seeding" during the download process. In file sharing terminology, "seeding" refers to sharing a file with other users while (or after) it is being downloaded. Since file sharing is based on a peer-to-peer network system, each user who downloads a file can upload a portion of the file to other users. Meta's law...

Feb. 15 (Bloomberg) -- Meta is investing heavily in AI-driven bionic humanoid robots, according to Bloomberg, which reports that Meta plans to "build robots that can perform tasks like humans and help people with physical tasks," and has created a new subteam within its Reality Labs hardware team to focus on the project. A new sub-team has been formed within the company's Reality Labs hardware team to focus on the project. Meta's first focus will be on robots that can do household chores, but the company's broader goal is to develop AI, sensors, and software for robotics and sell them...

Feb. 4 (Bloomberg) -- Meta CEO Mark Zuckerberg has promised to one day make general-purpose artificial intelligence (AGI) -- i.e., AI that is capable of doing anything a human can do -- publicly available to the general public. However, according to a newly released policy document from Meta, the company may limit the release of its internally developed, high-performance AI systems under certain circumstances. The document, called the "Frontier AI Framework," identifies two types of AI that Meta considers too risky to release...

January 24, Meta CEO Zuckerberg said on January 24 local time on the social platform Facebook, will achieve about 1GW of online computing in 2025, by the end of the year Meta will have more than 1.3 million GPUs. Zuckerberg said Meta plans to invest 60 billion to 65 billion U.S. dollars this year (Note: currently about 437.178 billion to 473.609 billion yuan) for capital expenditures, while significantly developing the artificial intelligence team. Zuckerberg said Meta plans to invest between $60 billion and $65 billion this year (note: currently about RMB 437.178 billion to RMB 473.609 billion) in capital expenditures, as well as significantly grow its artificial intelligence team. With Meta going ...

Jan. 22 (Bloomberg) -- Meta is upgrading its smart glasses and exploring new wearables such as watches and camera-equipped headsets, aiming to embed its AI capabilities into more products, Bloomberg's Mark Gurman reported today. Meta will develop Oakley-branded smart glasses for athletes this year, according to people familiar with the matter, and Reality Labs, Meta's device group, is also planning to release new high-end eyewear with built-in displays in 2025, though those initiatives have yet to be announced. The report also mentioned that Meta ...

On Tuesday, January 15, in an AI copyright case involving Meta (Kadrey v. Meta), the court made public records of internal communications between Meta executives and researchers. The documents show that as Meta developed its latest AI model, Llama 3, executives and researchers viewed outperforming OpenAI's GPT-4 as a central goal, and demonstrated a strong sense of competition in internal discussions. Ahmad Al-Dahle, Meta's vice president of generative AI ...

Jan. 4 - Meta has pulled a large number of AI-generated virtual profiles from Facebook and Instagram, Engadget reported today. The AI characters sparked widespread controversy and mockery on social media, and the company has since confirmed the decision. The dummy profiles, labeled "AI managed by Meta," launched in September 2023, alongside a celebrity AI chatbot that went live at the same time (note: now offline). However, the accounts have gone down in the last few months...

Meta Chief Scientist and Turing Award winner Yann LeCun (note: Yann LeCun, French) talked about his views on general artificial intelligence on the 29th "Into the Impossible" podcast. He said that the negative effects of AI are currently overblown and that its capabilities are still very limited. "In the most optimistic scenario, the realization of AGI is at least 5-6 years away." At present, society is generally worried about AI, and there is even a lack of views about the "possible doom" of AI, which Yang Likun believes has overlooked the fact that AI is ...

Meta is launching a series of AI products, including tools to help users create AI characters on Instagram and Facebook, in an effort to attract and retain younger users and compete with rivals, the FT reported today. Conor Hayes, Meta's vice president for generative AI, said the company expects these AI characters to gradually have an account-like presence on the platforms. These AIs will be equipped with profiles and avatars and will be able to generate and share content on the platform, signaling a shift in the way social media interactions...

Meta on Thursday announced the launch of an artificial intelligence model called Meta Motivo, which is designed to control the movements of humanoid digital intelligences to enhance metaverse experiences. Meta also released AI tools such as LCM, a large-scale conceptual model, and Video Seal, a video watermarking tool, and reiterated its commitment to continued investment in AI, AR and metaverse technologies. Note: Meta Motivo is a behavior-based base model trained in the Mujoco simulator using AMASS motion capture data...

Meta CEO Mark Zuckerberg posted an update on Threads three days ago stating that with the help of Llama 3.3 models, Meta AI is now approaching 600 million monthly users. The Meta AI service, which launches in the fall of 2023, announced in October that it had surpassed the 500 million monthly users mark, and the release of Llama 3.3 has further enticed users to use, experience, and download Meta's models like Llama. ...

December 10, 2011 - The bottleneck in training AI models is no longer just about architecture design, but also about data management efficiency - Meta AI has launched an open-source scalable and high-performance data loading (SPDL) tool that ultimately speeds up AI training by improving data loading efficiency. The SPDL tool uses multi-threading technology to achieve high throughput and lower resource usage in the regular Python interpreter (without the free-threading option enabled), and is compatible with the Free-Threaded Python...