Jump to content

Meta AI

From Wikipedia, the free encyclopedia
Meta AI
Company typeDivision
IndustryArtificial intelligence
FoundedDecember 11, 2015; 9 years ago (2015-12-11)
Founders
HeadquartersAstor Place, New York City, New York, U.S.
Key people
ProductsLLaMA
OwnerMeta Platforms
Websiteai.meta.com

Meta AI is a research division of Meta (formerly Facebook) that develops artificial intelligence and augmented reality technologies.

History

[edit]

The foundation of laboratory was announced in 2013, under the name Facebook Artificial Intelligence Research (FAIR).[1] FAIR has workspaces in Menlo Park, California, London, United Kingdom, and Manhattan. FAIR was first directed by New York University's Yann LeCun, a deep learning professor and Turing Award winner.[2] Working with NYU's Center for Data Science, FAIR's initial goal was to research data science, machine learning, and artificial intelligence.[3][self-published source?] Vladimir Vapnik, a pioneer in statistical learning, joined FAIR[4] in 2014.

FAIR opened a research center in Paris, France in 2015,[5] and subsequently launched smaller satellite research labs in Seattle, Pittsburgh, Tel Aviv, Montreal and London.[6] In 2016, FAIR partnered with Google, Amazon, IBM, and Microsoft in creating the Partnership on Artificial Intelligence to Benefit People and Society.

In 2018, Jérôme Pesenti, former CTO of IBM's big data group, assumed the role of president of FAIR, while LeCun stepped down to serve as chief AI scientist.[7] FAIR had approximately 200 staff in 2018.[8]

FAIR's initial work included research in learning-model enabled memory networks, self-supervised learning and generative adversarial networks, document classification and translation, as well as computer vision.[3] FAIR released Torch deep-learning modules as well as PyTorch in 2017, an open-source machine learning framework,[3] which was subsequently used in several deep learning technologies, such as Tesla's autopilot [9] and Uber's Pyro.[10] That same year, a pair of chatbots were falsely rumored[11] to be discontinued for developing a language that was unintelligible to humans.[12] However, FAIR clarified that the research had been shut down because they had accomplished their initial goal to understand how languages are generated, rather than out of fear.[11]

FAIR was renamed Meta AI following the rebranding that changed Facebook, Inc. to Meta Platforms Inc.[13]

In 2022, Meta AI predicted the 3D shape of 600 million potential proteins in two weeks.[14]

Meta AI assistant

[edit]

Meta AI is also the name of an AI assistant developed by the research division.[15]

Meta AI was pre-installed on the second generation of Ray-Ban Meta on September 27, 2023, as a voice assistant.[16] On April 23, 2024, Meta announced an update to Meta AI on the smart glasses to enable multimodal input via Computer vision.[17] On July 23, 2024, Meta announced that Meta AI with Vision would be incorporated into v68 of Horizon OS on the Meta Quest 3 and Quest Pro for detection of physical objects in passthrough mode, replacing the older voice assistant software in Horizon OS.[18] In addition, Meta AI without Vision is supported on the Quest 2.

On February 27, 2024, Meta announced the launch of a standalone AI app, expanding its AI ecosystem beyond social media integration.[19] The app aims to enhance user experience with advanced AI capabilities and introduce a premium subscription model.[20]

Meta AI has been progressively rolled out internationally.[21][22] India is among the nations with the highest usage of Meta AI on WhatsApp and Instagram.[21]

Current research

[edit]

Natural language processing and conversational AI

[edit]

Artificial intelligence communication requires a machine to understand natural language and to generate language that is natural. Meta AI seeks to improve these technologies to improve safe communication regardless of what language the user might speak.[23] Thus, a central task involves the generalization of natural language processing (NLP) technology to other languages. As such, Meta AI actively works on unsupervised machine translation.[24][25] Meta AI seeks to improve Natural-language user interface by developing aspects of chitchat dialogue such as repetition, specificity, response-relatedness and question-asking,[26] incorporating personality into image captioning,[27] and generating creativity-based language.[28]

In November 2022, a large language model designed for generating scientific text, Galactica, was released.[29] Meta withdrew Galactica on 17 November due to offensiveness and inaccuracy.[30] Before the cancellation, researchers were working on Galactica Instruct, which would use Large language model to allow the model to follow instructions to manipulate LaTeX documents on Overleaf.[31]

Llama

[edit]

In February 2023, Meta AI launched LLaMA (Large Language Model Meta AI), a large language model ranging from 7B to 65B parameters.[32] On April 5, 2025, Meta released two of the three Llama 4 models, Scout and Maverick (the biggest model, Behemoth, was still in training).[33]

Hardware

[edit]

Until 2022, Meta AI mainly used CPU and in-house custom chip as hardware, before finally switching to Nvidia GPU. This necessitated a complete redesign of several data centers, since they needed 24 to 32 times the networking capacity and new liquid cooling systems.[34]

MTIA v1

[edit]

The MTIA v1 is Meta's first-generation AI training and inference accelerator, developed specifically for Meta's recommendation workloads. It was fabricated using TSMC's 7 nm process technology and operates at a frequency of 800 MHz. In terms of processing power, the accelerator provides 102.4 TOPS at INT8 precision and 51.2 TFLOPS at FP16 precision, while maintaining a thermal design power (TDP) of 25 W.[35]

Meta AI offers options for users to customize their interaction with its features. Users are able to mute the AI chatbot on platforms like Facebook, Instagram, and WhatsApp,[36] temporarily halting notifications from the chatbot. Some platforms also offer the ability to hide certain AI elements from their interface. To locate the relevant settings, users can consult the platform's help documentation or settings menu.

Mathematical theorem proving

[edit]

In 2022, Meta created a method for proving mathematical theorems called HyperTree Proof Search (HTPS), which successfully generated proofs of 10 International Mathematical Olympiad problems in Lean.[37]

Since May 2024, the Meta AI chatbot has summarized news from various outlets without linking directly to original articles, including in Canada, where news links are banned on its platforms. This use of news content without compensation has raised ethical and legal concerns, especially as Meta continues to reduce news visibility on its platforms.[38]

References

[edit]
  1. ^ "NYU "Deep Learning" Professor LeCun Will Head Facebook's New Artificial Intelligence Lab". TechCrunch. 9 December 2013. Retrieved 2022-05-08.
  2. ^ "Yann LeCun - A.M. Turing Award Laureate". amturing.acm.org. Archived from the original on 2023-03-27. Retrieved 2022-05-08.
  3. ^ a b c "FAIR turns five: What we've accomplished and where we're headed". Engineering at Meta. 2018-12-05. Archived from the original on 2022-05-11. Retrieved 2022-05-08.
  4. ^ "Facebook's AI team hires Vladimir Vapnik, father of the popular support vector machine algorithm". VentureBeat. 2014-11-25. Archived from the original on 2014-11-27. Retrieved 2022-05-08.
  5. ^ Dillet, Romain (June 2, 2015). "Facebook Opens New AI Research Center in Paris". TechCrunch. Retrieved May 7, 2022.
  6. ^ "Facebook Opens New AI Research Center In Paris". TechCrunch. 2 June 2015. Retrieved 2022-05-08.
  7. ^ Dave, Greshgorn (January 23, 2018). "The head of Facebook's AI research is stepping into a new role as it shakes up management". Quartz. Archived from the original on May 8, 2022. Retrieved May 7, 2022.
  8. ^ Shead, Sam. "Facebook Plans To Double Size Of AI Research Unit By 2020". Forbes. Archived from the original on 2022-05-08. Retrieved 2022-05-08.
  9. ^ Karpathy, Andrej (6 November 2019). "PyTorch at Tesla - Andrej Karpathy, Tesla". YouTube. Archived from the original on 2023-03-24. Retrieved 2022-05-08.
  10. ^ "Pyro". pyro.ai. Archived from the original on 2022-05-06. Retrieved 2022-05-08.
  11. ^ a b "Facebook researchers shut down AI bots that started speaking in a language unintelligible to humans". Tech2. 2017-07-31. Archived from the original on 2022-05-08. Retrieved 2022-05-08.
  12. ^ McKay, Tom (2017-08-01). "No, Facebook Did Not Panic and Shut Down an AI Program That Was Getting Dangerously Smart". Gizmodo. Retrieved 2025-05-27. When Facebook directed two of these semi-intelligent bots to talk to each other, FastCo reported, the programmers realized they had made an error by not incentivizing the chatbots to communicate according to human-comprehensible rules of the English language. In their attempts to learn from each other, the bots thus began chatting back and forth in a derived shorthand—but while it might look creepy, that's all it was.
  13. ^ Murphy Kelly, Samantha (October 29, 2021). "Facebook changes its company name to Meta". CNN Business. Archived from the original on May 7, 2022. Retrieved May 7, 2022.
  14. ^ "Meta's new AI just predicted the shape of 600 million proteins in 2 weeks". Live Science. November 4, 2022. Archived from the original on November 5, 2022. Retrieved November 5, 2022.
  15. ^ "Meta AI rolls out in 6 more countries including UK and Brazil". India Today. 2024-10-10. Retrieved 2025-02-24.
  16. ^ "Introducing the New Ray-Ban | Meta Smart Glasses". Meta. 2023-09-27. Retrieved 2024-01-08.
  17. ^ "Smart(er) Glasses: Introducing New Ray-Ban | Meta Styles + Expanding Access to Meta AI with Vision". Meta Quest Blog. 2024-04-23. Archived from the original on 2024-07-27.
  18. ^ Meta Quest Blog (July 23, 2024). "Introducing Meta AI on Meta Quest—Your Smart MR Assistant". Meta Blog.
  19. ^ "Meta to Launch Standalone AI App with Premium Features Amid Growing Competition". Mint. 2024-02-27. Retrieved 2025-02-28.
  20. ^ "Meta AI Expansion: Standalone App and Subscription Model in the Works". Mint. 2024-02-27. Retrieved 2025-02-28.
  21. ^ a b Kawale, Ajinkya (20 February 2025). "India among largest Meta AI adopters, backs open-source innovation". Business Standard. Archived from the original on 2025-02-20. Retrieved 2025-02-22.
  22. ^ "Meet Your New Assistant: Meta AI, Built With Llama 3". 18 April 2024.
  23. ^ "Meta AI Research Topic - Natural Language Processing". ai.facebook.com. Archived from the original on 2022-05-08. Retrieved 2022-05-08.
  24. ^ Lample, Guillaume; Ott, Myle; Conneau, Alexis; Denoyer, Ludovic; Ranzato, Marc'Aurelio (2018-08-13). "Phrase-Based & Neural Unsupervised Machine Translation". arXiv:1804.07755 [cs.CL].
  25. ^ Conneau, Alexis; Lample, Guillaume; Rinott, Ruty; Williams, Adina; Bowman, Samuel R.; Schwenk, Holger; Stoyanov, Veselin (2018-09-13). "XNLI: Evaluating Cross-lingual Sentence Representations". arXiv:1809.05053 [cs.CL].
  26. ^ See, Abigail; Roller, Stephen; Kiela, Douwe; Weston, Jason (2019-04-10). "What makes a good conversation? How controllable attributes affect human judgments". arXiv:1902.08654 [cs.CL].
  27. ^ Shuster, Kurt; Humeau, Samuel; Hu, Hexiang; Bordes, Antoine; Weston, Jason (2019-03-20). "Engaging Image Captioning Via Personality". arXiv:1810.10665 [cs.CV].
  28. ^ Fan, Angela; Lewis, Mike; Dauphin, Yann (2018-05-13). "Hierarchical Neural Story Generation". arXiv:1805.04833 [cs.CL].
  29. ^ Taylor, Ross; Kardas, Marcin; Cucurull, Guillem; Scialom, Thomas; Hartshorn, Anthony; Saravia, Elvis; Poulton, Andrew; Kerkez, Viktor; Stojnic, Robert (2022-11-16). "Galactica: A Large Language Model for Science". arXiv:2211.09085 [cs.CL].
  30. ^ Edwards, Benj (18 November 2022). "New Meta AI demo writes racist and inaccurate scientific literature, gets pulled". Ars Technica. Retrieved 30 December 2022.
  31. ^ Scialom, Thomas (July 23, 2024). "Llama 2, 3 & 4: Synthetic Data, RLHF, Agents on the path to Open Source AGI". Latent Space (Interview). Interviewed by swyx & Alessio. Archived from the original on July 24, 2024.
  32. ^ Leswing, Kif (2023-02-24). "Mark Zuckerberg announces Meta's new large language model as A.I. race heats up". CNBC. Retrieved 2025-04-14.
  33. ^ Wiggers, Kyle (2025-04-05). "Meta releases Llama 4, a new crop of flagship AI models". TechCrunch. Retrieved 2025-04-14.
  34. ^ Insight: Inside Meta's scramble to catch up on AI By Katie Paul, Krystal Hu, Stephen Nellis and Anna Tong April 25, 20233:06 PM PDT
  35. ^ Peters, Jay (2023-05-19). "Meta is working on a new chip for AI". The Verge. Archived from the original on 2023-06-07. Retrieved 2023-06-07.
  36. ^ UBB, Ajit (May 2, 2024). "How to Turn OFF Meta AI Facebook". UBB.
  37. ^ "Teaching AI advanced mathematical reasoning". ai.meta.com. November 3, 2022.
  38. ^ "Meta walked away from news. Now the company's using it for AI content". The Washington Post. 21 May 2024. Archived from the original on 21 May 2024. Retrieved 22 May 2024.
[edit]