The Drum · Weekly AI news, freeSubscribe →
Timeline
Meta AI emerged from Facebook's pivot toward artificial intelligence, beginning with foundational research efforts in the early 2010s and accelerating dramatically after the company's 2021 rebrand. Under the leadership of Yann LeCun, Chief AI Scientist, Meta built one of the world's largest AI research organizations, developing large language models, computer vision systems, and multimodal AI. The division has pursued an open-source strategy with models like Llama, contrasting with competitors' proprietary approaches. Meta AI has faced scrutiny over content moderation, bias in AI systems, and the environmental costs of training massive models. Today, Meta AI represents a critical bet on artificial intelligence as central to the company's metaverse vision and long-term competitive positioning.
The Story
Meta's AI journey began as a research initiative within Facebook, gradually expanding as the company recognized AI's importance to content recommendation and user engagement. The 2021 rebrand to Meta accelerated this shift, positioning AI as central to the metaverse vision and attracting top research talent. A major inflection point came with the release of Llama in 2023, signaling Meta's commitment to open-source AI and challenging the closed-model dominance of OpenAI and Google. The narrative remains unresolved: whether Meta can leverage AI to drive metaverse adoption, whether open-source models will prove commercially viable against proprietary competitors, and how regulators will treat Meta's AI capabilities given the company's history with data and content moderation.
2013
Facebook AI Research Lab Founded
Facebook established FAIR, its dedicated AI research division, in Menlo Park. The lab focused on fundamental machine learning research and attracted leading researchers in deep learning and computer vision.
2015
FAIR Expands to Multiple Locations
Facebook opened FAIR offices in New York and Paris, signaling commitment to global AI talent acquisition. The expansion reflected growing competition for top researchers in machine learning.
2016
Yann LeCun Joins as Chief AI Scientist
Renowned deep learning pioneer Yann LeCun joined Facebook as Chief AI Scientist, leading the company's AI research strategy. His arrival elevated FAIR's profile and attracted additional top-tier researchers.
2017
PyTorch Framework Released
Facebook released PyTorch, an open-source machine learning framework that became widely adopted in AI research and industry. The framework's flexibility and ease of use made it a competitor to TensorFlow.
2018
AI Powers Content Recommendation
Facebook deployed AI systems to improve feed ranking and content recommendation at scale. These systems processed billions of user interactions daily to personalize content delivery.
2019
Bias Concerns in AI Content Moderation
Researchers and advocacy groups raised concerns about bias in Facebook's AI-powered content moderation systems. Studies showed the systems disproportionately flagged content from marginalized communities.
2020
AI Research Accelerates During Pandemic
Facebook increased AI research funding and hiring as remote work enabled global talent recruitment. The company invested heavily in recommendation systems and computer vision technologies.
2020-10
Jigsaw Partnership for Harmful Content Detection
Facebook partnered with Jigsaw, Google's incubator for addressing online harms, to develop AI systems for detecting harmful content. The collaboration aimed to improve content moderation across platforms.
2021-06
Meta Rebrand Announced
Facebook announced its rebrand to Meta, positioning the company around the metaverse concept. AI became central to the company's strategic narrative and future vision.
2021-10
Meta AI Becomes Standalone Division
Meta reorganized its AI efforts into a dedicated division, consolidating FAIR and applied AI teams. The restructuring reflected AI's growing importance to the company's core business.
2021-11
Massive AI Infrastructure Investment
Meta announced plans to invest billions in AI infrastructure and computing capacity. The company built specialized data centers to support training of large-scale models.
2022-01
AI Research on Multimodal Learning
Meta AI published research on multimodal models combining vision and language. These systems could understand and generate content across multiple modalities simultaneously.
2022-03
Make-A-Scene Image Generation Model
Meta AI released Make-A-Scene, a model for generating images from text descriptions with spatial control. The model demonstrated advances in controllable generative AI.
2022-06
Galactica Large Language Model Released
Meta AI released Galactica, a large language model trained on scientific papers and academic text. The model was designed for scientific reasoning and knowledge synthesis.
2022-07
Galactica Withdrawn After Criticism
Meta withdrew Galactica from public access following criticism about factual errors and potential misuse. The incident highlighted challenges in releasing large language models responsibly.
2022-09
AI for Metaverse Avatar Creation
Meta AI developed systems for generating realistic avatars and virtual environments using AI. These tools aimed to accelerate metaverse content creation and user engagement.
2022-11
ChatGPT Launch Shifts AI Landscape
OpenAI released ChatGPT, capturing public attention and accelerating the large language model race. Meta faced pressure to accelerate its own language model development.
2023-02
Llama Large Language Model Announced
Meta AI announced Llama, a family of large language models ranging from 7 billion to 65 billion parameters. The models were initially released to researchers but quickly leaked to the public.
2023-03
Llama Becomes Open-Source Standard
Despite initial restrictions, Llama became widely available and adopted across the AI community. The model's open-source nature democratized access to frontier language models.
2023-04
Llama 2 Released with Commercial License
Meta released Llama 2, an improved version with explicit commercial use rights. The move positioned Meta as a champion of open-source AI against proprietary competitors.
2023-05
Segment Anything Model Released
Meta AI released SAM, a foundation model for image segmentation trained on billions of images. The model demonstrated Meta's advances in computer vision and transfer learning.
2023-07
AI Regulation Discussions Intensify
Meta engaged with policymakers on AI regulation as governments worldwide considered frameworks. The company advocated for balanced regulation that wouldn't stifle innovation.
2023-08
Llama 2 Adoption Accelerates
Llama 2 became the most widely adopted open-source large language model. Companies and researchers built applications and fine-tuned versions for specific domains.
2023-09
Meta AI Hiring Surge
Meta announced aggressive hiring for AI research and engineering roles. The company competed intensely with OpenAI, Google, and Anthropic for top talent.
2023-10
Multimodal AI Research Advances
Meta AI published research on multimodal models combining text, images, and video. These systems could understand and reason across different types of content.
2023-11
AI for Content Moderation Improvements
Meta deployed improved AI systems for detecting hate speech and harmful content. The systems used multimodal understanding to catch violations across text and images.
2024-01
Llama 2 Reaches 10 Million Downloads
Llama 2 surpassed 10 million downloads, becoming a foundational model for the open-source AI ecosystem. The milestone demonstrated the success of Meta's open-source strategy.
2024-02
AI for Metaverse Interaction
Meta AI developed systems for natural language interaction within metaverse environments. Users could communicate with AI agents and control virtual spaces through conversation.
2024-03
Environmental Impact Concerns Raised
Environmental groups raised concerns about the energy consumption of Meta's large-scale AI training. The company committed to using renewable energy for AI infrastructure.
2024-04
Llama 3 Development Announced
Meta announced Llama 3, the next generation of its language model family. The model was expected to match or exceed the capabilities of leading proprietary models.
© 2026 Cedar & Bloom. All rights reserved.
