# Meta AI > Mediated Wiki article. Canonical URL: https://mediated.wiki/source/Meta_AI > Markdown URL: https://mediated.wiki/source/Meta_AI.md > Source: https://en.wikipedia.org/wiki/Meta_AI > Source revision: 1353663462 > License: Creative Commons Attribution-ShareAlike 4.0 International (https://creativecommons.org/licenses/by-sa/4.0/) {{Short description|Artificial intelligence division of Meta Platforms}} {{Multiple issues| {{Self-published|article|date=July 2025}} {{Update|date=November 2025}} }} {{Infobox company | name = Meta AI | logo = Meta AI Logo (2026).svg | logo_size = 110px | caption = | image = Meta AI screenshot.jpg | type = [Division](/source/Division_(business)) | industry = [Artificial intelligence](/source/Artificial_intelligence) | founded = {{Start date and age|2013}} | founders = {{hlist|[Rob Fergus](/source/Rob_Fergus)|[Yann LeCun](/source/Yann_LeCun)|[Serkan Piantino](/source/Serkan_Piantino)|[Mark Zuckerberg](/source/Mark_Zuckerberg)}} | dissolved = | hq_location = [Astor Place](/source/Astor_Place), [New York City](/source/New_York_City), [New York](/source/New_York_(state)), U.S. | area_served = | key_people = | products = [Llama](/source/Llama_(language_model)) | services = | revenue = | equity = | equity_year = | owner = [Meta Platforms](/source/Meta_Platforms) | num_employees = | subsid = | homepage = {{URL|https://ai.meta.com/}} | footnotes = | former name = | board_of_directors = | focus = | motto = | method = | origins = | registration_id = | tax_id = | leader_title = }} {{Meta sidebar}} {{Artificial intelligence}} '''Meta AI''' is a research division of [Meta](/source/Meta_Platforms) (formerly Facebook) that develops [artificial intelligence](/source/artificial_intelligence) and [augmented reality](/source/augmented_reality) technologies. == History == Meta AI was founded in 2013 as Facebook Artificial Intelligence Research (FAIR).{{Cite web |date=9 December 2013 |title=NYU "Deep Learning" Professor LeCun Will Head Facebook's New Artificial Intelligence Lab |url=https://techcrunch.com/2013/12/09/facebook-artificial-intelligence-lab-lecun/ |access-date=2022-05-08 |website=TechCrunch |language=en-US}}{{Cite web |date=2014-11-25 |title=Facebook's AI team hires Vladimir Vapnik, father of the popular support vector machine algorithm |url=https://venturebeat.com/2014/11/25/facebooks-ai-team-hires-vladimir-vapnik-father-of-the-popular-support-vector-machine-algorithm/ |url-status=live |archive-url=https://web.archive.org/web/20141127001517/http://venturebeat.com/2014/11/25/facebooks-ai-team-hires-vladimir-vapnik-father-of-the-popular-support-vector-machine-algorithm/ |archive-date=2014-11-27 |access-date=2022-05-08 |website=VentureBeat |language=en-US}} It has workspaces in [Menlo Park](/source/Menlo_Park%2C_California), London, New York City, Paris, Seattle, [Pittsburgh](/source/Pittsburgh), [Tel Aviv](/source/Tel_Aviv), and [Montreal](/source/Montreal) as of 2025.{{Cite web |title=Facebook Opens New AI Research Center In Paris |url=https://techcrunch.com/2015/06/02/facebook-opens-new-ai-research-center-in-paris/ |access-date=2022-05-08 |website=TechCrunch |date=2 June 2015 |language=en-US }}{{Cite news |last=Dillet |first=Romain |date=June 2, 2015 |title=Facebook Opens New AI Research Center in Paris |url=https://techcrunch.com/2015/06/02/facebook-opens-new-ai-research-center-in-paris/ |access-date=May 7, 2022 |work=TechCrunch}} In 2016, FAIR partnered with [Google](/source/Google), [Amazon](/source/Amazon_(company)), [IBM](/source/IBM), and [Microsoft](/source/Microsoft) in creating the [Partnership on Artificial Intelligence to Benefit People and Society](/source/Partnership_on_AI). Meta AI was directed by [Yann LeCun](/source/Yann_LeCun) until 2018, when Jérôme Pesenti succeeded the role. Pesenti is formerly the [CTO](/source/Chief_technology_officer) of [IBM's](/source/IBM) [big data](/source/big_data) group.{{Cite news |last=Dave |first=Greshgorn |date=January 23, 2018 |title=The head of Facebook's AI research is stepping into a new role as it shakes up management |work=Quartz |url=https://qz.com/1186806/yann-lecun-is-stepping-down-as-facebooks-head-of-ai-research/ |access-date=May 7, 2022 |archive-date=May 8, 2022 |archive-url=https://web.archive.org/web/20220508180652/https://qz.com/1186806/yann-lecun-is-stepping-down-as-facebooks-head-of-ai-research/ |url-status=live }} FAIR's research includes [self-supervised learning](/source/self-supervised_learning), [generative adversarial network](/source/generative_adversarial_network)s, [document classification](/source/document_classification) and translation, and [computer vision](/source/computer_vision).{{Cite web |date=2018-12-05 |title=FAIR turns five: What we've accomplished and where we're headed |url=https://engineering.fb.com/2018/12/05/ai-research/fair-fifth-anniversary/ |url-status=live |archive-url=https://web.archive.org/web/20220511012517/https://engineering.fb.com/2018/12/05/ai-research/fair-fifth-anniversary/ |archive-date=2022-05-11 |access-date=2022-05-08 |website=Engineering at Meta |language=en-US}} FAIR released [Torch](/source/Torch_(machine_learning)) deep-learning modules as well as [PyTorch](/source/PyTorch) in 2017, an [open-source](/source/Open_source) machine learning framework, which was subsequently used in several deep learning technologies, such as [Tesla](/source/Tesla%2C_Inc.)'s autopilot {{Cite web |last=Karpathy |first=Andrej |title=PyTorch at Tesla - Andrej Karpathy, Tesla |website=[YouTube](/source/YouTube) |date=6 November 2019 |url=https://www.youtube.com/watch?v=oBklltKXtDE |access-date=2022-05-08 |archive-date=2023-03-24 |archive-url=https://web.archive.org/web/20230324144838/https://www.youtube.com/watch?v=oBklltKXtDE |url-status=live }} and [Uber](/source/Uber)'s Pyro.{{Cite web |title=Pyro |url=https://pyro.ai/ |access-date=2022-05-08 |website=pyro.ai |archive-date=2022-05-06 |archive-url=https://web.archive.org/web/20220506094900/http://pyro.ai/ |url-status=live }} That same year, a pair of chatbots were falsely rumored{{Cite web |date=2017-07-31 |title=Facebook researchers shut down AI bots that started speaking in a language unintelligible to humans |url=https://www.firstpost.com/tech/news-analysis/facebook-researchers-shut-down-ai-bots-that-started-speaking-in-a-language-unintelligible-to-humans-3876197.html |url-status=live |archive-url=https://web.archive.org/web/20220508180652/https://www.firstpost.com/tech/news-analysis/facebook-researchers-shut-down-ai-bots-that-started-speaking-in-a-language-unintelligible-to-humans-3876197.html |archive-date=2022-05-08 |access-date=2022-05-08 |website=Tech2}} to be discontinued for developing a language that was unintelligible to humans.{{Cite web |last=McKay |first=Tom |date=2017-08-01 |title=No, Facebook Did Not Panic and Shut Down an AI Program That Was Getting Dangerously Smart |url=https://gizmodo.com/no-facebook-did-not-panic-and-shut-down-an-ai-program-1797414922 |access-date=2025-05-27 |website=Gizmodo |language=en-US |quote=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.}} FAIR clarified that the research had been shut down because they had accomplished their initial goal to understand how languages are generated by their models, rather than out of fear. FAIR was renamed Meta AI following the rebranding that changed Facebook, Inc. to Meta Platforms Inc.{{Cite news |last=Murphy Kelly |first=Samantha |date=October 29, 2021 |title=Facebook changes its company name to Meta |url=https://edition.cnn.com/2021/10/28/tech/facebook-mark-zuckerberg-keynote-announcements/index.html |access-date=May 7, 2022 |work=CNN Business |archive-date=May 7, 2022 |archive-url=https://web.archive.org/web/20220507032946/https://edition.cnn.com/2021/10/28/tech/facebook-mark-zuckerberg-keynote-announcements/index.html |url-status=live }} On October 1, 2025, Facebook announced "We will soon use your interactions with AI at Meta to personalize the content and ads you see".{{cite web |title=Improving Your Recommendations on Our Apps With AI at Meta |url=https://about.fb.com/news/2025/10/improving-your-recommendations-apps-ai-meta/ |website=Meta Newsroom |archive-url=https://archive.today/20251001210742/https://about.fb.com/news/2025/10/improving-your-recommendations-apps-ai-meta/ |archive-date=1 October 2025 |date=1 October 2025}} == Virtual assistant == Meta AI is also the name of the [virtual assistant](/source/virtual_assistant) developed by the team, now integrated as a [chatbot](/source/chatbot) into Meta's [social networking](/source/Social_network) products.{{Cite web |last=Kawale |first=Ajinkya |date=20 February 2025 |title=India among largest Meta AI adopters, backs open-source innovation |url=https://www.business-standard.com/technology/tech-news/india-among-largest-meta-ai-adopters-backs-open-source-innovation-125022000986_1.html |archive-url=https://web.archive.org/web/20250220130626/https://www.business-standard.com/technology/tech-news/india-among-largest-meta-ai-adopters-backs-open-source-innovation-125022000986_1.html |archive-date=2025-02-20 |access-date=2025-02-22 |website=Business Standard |language=en-US}} It is also available as a [subscription-based](/source/Subscription_business_model) stand-alone [app](/source/Mobile_app).{{Cite web |date=2024-02-27 |title=Meta to Launch Standalone AI App with Premium Features Amid Growing Competition |url=https://www.livemint.com/technology/tech-news/meta-ai-to-get-a-standalone-app-launch-its-premium-subscription-soon-report-11740706441071.html |access-date=2025-02-28 |website=Mint}}{{Cite web |date=2024-02-27 |title=Meta AI Expansion: Standalone App and Subscription Model in the Works |url=https://www.livemint.com/technology/tech-news/meta-ai-to-get-a-standalone-app-launch-its-premium-subscription-soon-report-11740706441071.html |access-date=2025-02-28 |website=Mint}} The virtual assistant was [pre-installed](/source/Pre-installed_software) on the second generation of [Ray-Ban Meta](/source/Ray-Ban_Meta) smartglasses, and can incorporate inputs from the glasses' cameras after an update.{{Cite web |last= |date=2024-04-23 |title=Smart(er) Glasses: Introducing New Ray-Ban {{!}} Meta Styles + Expanding Access to Meta AI with Vision |url=https://www.meta.com/blog/quest/ray-ban-meta-smart-glasses-new-styles-multimodal-ai-ferrari/ |url-status=live |archive-url=https://web.archive.org/web/20240727025149/https://www.meta.com/blog/quest/ray-ban-meta-smart-glasses-new-styles-multimodal-ai-ferrari/ |archive-date=2024-07-27 |website=Meta Quest Blog}} It is also available on [Quest 2](/source/Quest_2) and newer [HMDs](/source/Head-mounted_display).{{Cite web |last=Meta Quest Blog |date=July 23, 2024 |title=Introducing Meta AI on Meta Quest—Your Smart MR Assistant |url=https://www.meta.com/blog/quest/meta-ai-on-meta-quest-3 |website=Meta Blog}} Since May 2024, the 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 and [attribution](/source/Attribution_(copyright)) has raised ethical and legal concerns, especially as Meta continues to reduce news visibility on its platforms.{{cite news |date=21 May 2024 |title=Meta walked away from news. Now the company's using it for AI content |url=https://www.washingtonpost.com/technology/2024/05/22/meta-ai-news-summaries/ |archive-url=https://web.archive.org/web/20240521131313/https://www.washingtontimes.com/news/2024/may/21/slovakias-parliament-votes-to-condemn-political-vi/ |archive-date=21 May 2024 |access-date=22 May 2024 |newspaper=The Washington Post}} == Current research == {{Original research|section|date=July 2025|reason=cited articles are all original researches instead of e.g., reviews.}} === Natural language processing and chatbot === {{See also|Natural language processing}} Natural language processing is the ability for machines to [understand](/source/natural_language_understanding) and [generate natural language](/source/natural_language_generation). The team is also researching [unsupervised machine translation](/source/Unsupervised_machine_learning) and [multilingual](/source/Multilingualism) chatbots.{{Cite web |title=Meta AI Research Topic - Natural Language Processing |url=https://ai.facebook.com/research/topics/nlp |access-date=2022-05-08 |website=ai.facebook.com |language=en |archive-date=2022-05-08 |archive-url=https://web.archive.org/web/20220508180655/https://ai.facebook.com/research/topics/nlp/ |url-status=live }}{{cite arXiv |last1=Lample |first1=Guillaume |last2=Ott |first2=Myle |last3=Conneau |first3=Alexis |last4=Denoyer |first4=Ludovic |last5=Ranzato |first5=Marc'Aurelio |date=2018-08-13 |title=Phrase-Based & Neural Unsupervised Machine Translation |class=cs.CL |eprint=1804.07755}}{{cite arXiv |last1=Conneau |first1=Alexis |last2=Lample |first2=Guillaume |last3=Rinott |first3=Ruty |last4=Williams |first4=Adina |last5=Bowman |first5=Samuel R. |last6=Schwenk |first6=Holger |last7=Stoyanov |first7=Veselin |date=2018-09-13 |title=XNLI: Evaluating Cross-lingual Sentence Representations |class=cs.CL |eprint=1809.05053}}{{Anchor|Galactica}} ==== Galactica ==== Galactica is a [large language model (LLM)](/source/large_language_model) designed for generating scientific text. It was available for three days from 15 November 2022, before being withdrawn for generating [racist](/source/Racism) and [inaccurate](/source/Inaccurate_claim) content.{{Cite web |title=Why Meta's latest large language model survived only three days online |url=https://www.technologyreview.com/2022/11/18/1063487/meta-large-language-model-ai-only-survived-three-days-gpt-3-science/ |access-date=2025-07-18 |website=MIT Technology Review |language=en}}{{cite news |last1=Edwards |first1=Benj |date=18 November 2022 |title=New Meta AI demo writes racist and inaccurate scientific literature, gets pulled |url=https://arstechnica.com/information-technology/2022/11/after-controversy-meta-pulls-demo-of-ai-model-that-writes-scientific-papers/ |access-date=30 December 2022 |work=[Ars Technica](/source/Ars_Technica) |language=en-us}} ==== Llama ==== {{Main|Llama (language model)}} [Llama](/source/Llama_(language_model)) is an LLM released in February 2023.{{Cite web |last=Leswing |first=Kif |date=2023-02-24 |title=Mark Zuckerberg announces Meta's new large language model as A.I. race heats up |url=https://www.cnbc.com/2023/02/24/mark-zuckerberg-announces-meta-llama-large-language-model.html |access-date=2025-04-14 |website=CNBC |language=en}} As of January 2026, the most recent release is the [Llama 4](/source/Llama_(language_model)).{{Cite web |last=Wiggers |first=Kyle |date=2025-04-05 |title=Meta releases Llama 4, a new crop of flagship AI models |url=https://techcrunch.com/2025/04/05/meta-releases-llama-4-a-new-crop-of-flagship-ai-models/ |access-date=2025-04-14 |website=TechCrunch |language=en-US}} === Hardware === Meta used [CPUs](/source/Central_processing_unit) and in-house custom [chips](/source/Integrated_circuit) before 2022; they switched to [Nvidia](/source/Nvidia) [GPUs](/source/Graphics_processing_unit) since then.{{Cite news |date=2023-04-26 |title=Insight: Inside Meta's scramble to catch up on AI |url=https://www.reuters.com/technology/inside-metas-scramble-catch-up-ai-2023-04-25/ |work=Reuters}} MTIA v1, one of their early chips, is designed for the company's [content recommendation](/source/content_recommendation) algorithms. It was fabricated on [TSMC](/source/TSMC)'s [7 nm process](/source/7_nm_process) technology and consumed 25W, capable of 51.2 [TFlops](/source/Floating_point_operations_per_second) [FP16](/source/Half-precision_floating-point_format).{{Cite web |last=Peters |first=Jay |date=2023-05-19 |title=Meta is working on a new chip for AI |url=https://www.theverge.com/2023/5/18/23728678/meta-ai-new-chip-mtia-msvp-datacenter |access-date=2023-06-07 |website=The Verge |language=en-US |archive-date=2023-06-07 |archive-url=https://web.archive.org/web/20230607123218/https://www.theverge.com/2023/5/18/23728678/meta-ai-new-chip-mtia-msvp-datacenter |url-status=live }} == Controversy == The French media outlet [Mediapart](/source/Mediapart) reports that in 2022, Facebook's parent company illegally used works accumulated by the pirate site [LibGen](/source/LibGen) to train its artificial intelligence.{{Cite web |date=December 22, 2025 |title=Comment un cofondateur de Mistral AI a piraté des millions de livres quand il travaillait chez Meta |url=https://www.mediapart.fr/journal/economie-et-social/221225/comment-un-cofondateur-de-mistral-ai-pirate-des-millions-de-livres-quand-il-travaillait-chez-m |access-date=December 23, 2025 |website=Mediapart |language=fr}} == References == {{reflist}} == External links == *{{Official website|https://ai.meta.com/|name=Official website}} {{Meta Platforms}} {{Meta Horizon}} {{Generative AI}} {{Generative AI chatbots}} {{Virtual assistants}} {{Authority control}} Category:Artificial intelligence laboratories Category:Chatbots Category:Meta Platforms Category:Virtual assistants Category:2013 establishments in California Category:Artificial intelligence industry in the United States Category:Meta Platforms criticism and controversies --- Adapted from the Wikipedia article [Meta AI](https://en.wikipedia.org/wiki/Meta_AI) by Wikipedia contributors ([contributor history](https://en.wikipedia.org/wiki/Meta_AI?action=history)). Available under [Creative Commons Attribution-ShareAlike 4.0 International](https://creativecommons.org/licenses/by-sa/4.0/). Changes may have been made.