OpenAI, established in December 2015, is a prominent American AI research organization dedicated to developing "safe and beneficial" artificial general intelligence (AGI). Their research focuses on creating highly autonomous systems capable of surpassing human capabilities in economically valuable tasks. A key player in the AI boom, OpenAI has produced significant advancements like large language models, sophisticated image generation models, and previously, open-source models. The organization's release of ChatGPT is widely recognized as a catalyst for the current surge in AI development and interest.
December 2015: OpenAI Assembles Top AI Researchers
OpenAI successfully recruited nine leading AI researchers, identified by Greg Brockman and deep learning pioneer Yoshua Bengio, as its initial employees.
December 2015: Formation of OpenAI and Funding Announcement
OpenAI was publicly announced with prominent figures like Sam Altman, Elon Musk, and others pledging over $1 billion in funding. The organization committed to open collaboration, sharing its research and patents with the public.
December 2015: OpenAI Founded
OpenAI, an American artificial intelligence research organization, was founded with the goal of developing 'safe and beneficial' artificial general intelligence.
April 2016: OpenAI Gym Beta Release
OpenAI released a public beta version of 'OpenAI Gym,' a platform specifically designed for research in the field of reinforcement learning.
August 2016: Nvidia's Supercomputer Donation
Nvidia donated its first DGX-1 supercomputer to OpenAI, enabling the organization to train larger and more complex AI models more efficiently. This donation significantly reduced the processing time for training models.
December 2016: OpenAI Universe Released
OpenAI unveiled 'Universe,' a software platform designed to measure and train AI's general intelligence using a wide range of real-world applications, including games and websites.
2016: Gym Announcement
OpenAI announces Gym, a platform intended to serve as an easily implemented benchmark for general-intelligence across a wide range of environments. Similar to the ImageNet challenge for supervised learning, Gym aims to standardize the definition of environments in AI research, promoting reproducibility. The platform offers a user-friendly interface for interacting with these environments.
June 2017: Gym Python Compatibility
OpenAI's Gym platform, designed for reinforcement learning research, is made compatible with Python programming language. This allows researchers to use Python to develop and experiment with AI agents within various simulated environments provided by Gym.
September 2017: Shift of Gym Development Focus to GitHub
OpenAI shifts the focus of development and maintenance efforts for its Gym platform from the official documentation website to the platform's GitHub page. This suggests a change in the primary channel for updates, communication, and community engagement related to Gym.
2017: RoboSumo Release
OpenAI introduces RoboSumo, a virtual environment where simulated metalearning robots learn to navigate and compete. Initially lacking knowledge of basic movement, these humanoid robots are tasked with learning to walk and pushing their opponent out of a ring. Through adversarial learning, the robots adapt to changing conditions and exhibit generalization abilities. When transferred to a new environment with strong winds, the robots demonstrate an understanding of balance learned in RoboSumo, showcasing the potential for competition to drive the development of robust AI capabilities.
2017: OpenAI Five First Public Demonstration
OpenAI showcases its Dota 2 bot, later known as OpenAI Five, at The International 2017. In a one-on-one match, the bot defeats professional Dota 2 player Dendi. The demonstration highlights the bot's ability to learn through trial-and-error algorithms, using reinforcement learning to improve its gameplay over time by playing against itself and being rewarded for successful actions. The event marks a significant step towards developing AI systems capable of handling complex tasks.
2017: OpenAI's Significant Cloud Computing Costs
OpenAI incurred substantial expenses on cloud computing, spending $7.9 million, a quarter of its total functional expenses, solely on this resource. In comparison, DeepMind, another prominent AI research company, had total expenses of $442 million in the same year.
June 11, 2018: Generative Pre-Training of a Transformer-Based Language Model
Alec Radford and colleagues publish a paper detailing a groundbreaking approach to generative pre-training of a transformer-based language model. The paper, published on OpenAI's website, demonstrates how this generative language model can acquire world knowledge and process long-range dependencies through pre-training on a diverse and extensive text corpus. This research significantly advances the field of natural language processing, paving the way for more sophisticated language models.
June 2018: OpenAI Five Plays as Full Team
OpenAI Five expands its capabilities to play as a full team of five bots in Dota 2, demonstrating its ability to coordinate and strategize against human opponents. The team showcases its progress by defeating amateur and semi-professional Dota 2 teams, marking a significant advancement in the development of AI for team-based competitive games.
2018: Elon Musk Resigns from OpenAI Board
Elon Musk stepped down from his position on OpenAI's Board of Directors, citing potential conflicts of interest with his role as CEO of Tesla, particularly concerning Tesla's development of AI for self-driving cars. While Musk stated he would remain a donor, he did not make any further donations after his departure.
2018: Dactyl Development
OpenAI develops Dactyl, a system using machine learning to train a Shadow Hand, a human-like robot hand, to manipulate physical objects. Dactyl's learning process occurs entirely in simulation, employing the same reinforcement learning algorithms and training code used for OpenAI Five. The project addresses the challenge of object orientation through domain randomization, exposing the AI to diverse simulated experiences to enhance its adaptability in real-world scenarios. Dactyl's setup includes RGB cameras, enabling the robot to perceive and interact with arbitrary objects by sight. OpenAI demonstrates the system's capability to manipulate a cube and an octagonal prism.
2018: Debate Game Launch
OpenAI launched the Debate Game, which trains machines to debate simplified problems in front of a human judge. The goal is to research how this approach can help audit AI decisions and develop explainable AI.
2018: Gym Retro Release
OpenAI releases Gym Retro, a platform specifically designed for reinforcement learning (RL) research on video games. It allows researchers to train RL algorithms and study their ability to generalize across different games. Gym Retro focuses on enabling agents to transfer knowledge from one game to another, even if the games have similar concepts but different visual appearances.
2018: Resource-Intensive Training for Dota 2 Bots
Training OpenAI's Dota 2 bots demanded significant computational resources, requiring the rental of 128,000 CPUs and 256 GPUs from Google for several weeks.
February 2019: GPT-2 Announcement and Limited Release
OpenAI announced GPT-2, a more powerful unsupervised transformer language model compared to its predecessor GPT-1. Due to concerns about potential misuse, only limited demonstrative versions were initially released to the public.
February 2019: GPT-2 Announcement
OpenAI announced the development of GPT-2, a language model capable of generating human-quality text, garnering significant attention for its advancements in natural language processing.
April 2019: OpenAI Five Defeats Dota 2 World Champions
In a live exhibition match, OpenAI Five defeats OG, the reigning Dota 2 world champions, with a score of 2-0. The victory highlights the substantial progress made in AI capabilities for complex and strategic games, showcasing the potential of AI to surpass human expertise in challenging domains.
November 2019: Full GPT-2 Release
OpenAI released the complete version of the GPT-2 language model. This followed the development of tools to detect "neural fake news" and discussions about the potential impact of GPT-2 on online information.
2019: Dactyl Solves Rubik's Cube
OpenAI demonstrates Dactyl's ability to solve a Rubik's Cube, showcasing advancements in the robot's dexterity and problem-solving skills. The robot successfully solves the puzzle 60% of the time. This achievement tackles the complex physics involved in manipulating objects like the Rubik's Cube, which are difficult to model accurately. OpenAI achieves this by enhancing Dactyl's robustness to perturbations using Automatic Domain Randomization (ADR). ADR automatically generates progressively challenging simulated environments, eliminating the need for manual specification of randomization ranges and further improving the robot's adaptability.
2019: OpenAI Transitions to Capped-Profit Model
OpenAI underwent a significant structural change, shifting from a non-profit to a 'capped' for-profit organization. This transition limited profits to 100 times any investment, enabling OpenAI Global LLC to attract investments from venture funds and offer employees stakes in the company.
2019: OpenAI Funding Clarification
Investigations revealed that the initial pledged funding for OpenAI was significantly less than $1 billion, with Elon Musk being the primary donor. Notably, YC Research did not contribute despite initial claims.
2019: Microsoft Invests in OpenAI
Microsoft made a significant investment in OpenAI Global LLC, providing $1 billion to support the development and research of artificial intelligence.
2019: MuseNet Release
OpenAI released MuseNet, a deep neural network trained to predict subsequent musical notes in MIDI music files. It can generate songs with 10 instruments in 15 different styles.
May 2020: GPT-3 Description
OpenAI first described GPT-3, an unsupervised transformer language model significantly larger than GPT-2, containing 175 billion parameters.
June 2020: OpenAI API Launch
OpenAI announces the launch of a multi-purpose API, providing developers with access to its advanced AI models. This API enables developers to integrate OpenAI's AI capabilities into their own applications, facilitating a wide range of English language AI tasks and expanding the accessibility of OpenAI's technology.
June 2020: GPT-3 Private Beta
OpenAI initiated a two-month free private beta for GPT-3 through a paid cloud API. The full model was not immediately released to the public due to concerns about potential misuse.
September 23, 2020: Microsoft Exclusive License for GPT-3
OpenAI granted Microsoft an exclusive license for GPT-3.
2020: OpenAI Releases Microscope
OpenAI released Microscope, a collection of visualizations for analyzing the features of eight neural network models commonly studied in interpretability. The models include AlexNet, VGG 19, different versions of Inception, and different versions of CLIP Resnet.
2020: Jukebox Release
OpenAI released Jukebox, an open-sourced algorithm for generating music with vocals. Trained on 1.2 million samples, Jukebox can create song samples based on genre, artist, and lyrics snippets. OpenAI acknowledges limitations in the model's ability to create complex musical structures.
2020: GPT-3 and API Launch
OpenAI introduced GPT-3, a powerful language model trained on massive internet datasets, showcasing its ability to answer questions, translate languages, and generate creative text formats. OpenAI also launched its commercial API, powered by GPT-3, as its first commercial product.
2020: MuseNet in Pop Culture
MuseNet was used in the internet psychological thriller Ben Drowned to create music for the titular character, demonstrating early application in pop culture.
March 2021: DALL-E Reveal
OpenAI introduced DALL-E, a Transformer model that generates images from textual descriptions using a 12-billion-parameter version of GPT-3. At this time, no API or code was made available.
August 2021: Codex API Private Beta Release
OpenAI released the API for Codex, a GPT-3 descendant trained on code from 54 million GitHub repositories, in private beta. Codex powers GitHub Copilot.
2021: DALL-E Unveiled
OpenAI introduced DALL-E, a specialized deep learning model capable of generating intricate digital images from textual descriptions using a variation of the GPT-3 architecture.
2021: CLIP Unveiling
OpenAI revealed CLIP, a model trained to analyze the semantic similarity between text and images, with applications in image classification.
April 2022: DALL-E 2 Announcement
OpenAI announced DALL-E 2, an enhanced version of DALL-E capable of producing more realistic images.
November 2022: OpenAI Launches ChatGPT
OpenAI launched ChatGPT, a conversational AI tool built on GPT-3, allowing users to ask questions and receive answers in natural language. ChatGPT gained 1 million users within 5 days of its launch.
December 2022: ChatGPT Public Preview Launch
OpenAI released a free preview of ChatGPT, its new AI chatbot powered by GPT-3.5, attracting widespread media attention and over a million signups within its first five days.
December 2022: Point-E Release
OpenAI published the software for Point-E on GitHub, a system designed to convert text descriptions into 3D models.
2022: OpenAI Publishes Approach to AGI Alignment Problem
OpenAI releases its strategy for addressing the challenge of aligning artificial general intelligence (AGI) with human values. The organization acknowledges the potential risks of unaligned AGI, emphasizing the difficulty of the alignment problem and suggesting it may require global cooperation to solve. OpenAI outlines its plans to explore methods for leveraging human feedback in AI training and using AI to progressively automate alignment research.
2022: Whisper Release
OpenAI released Whisper, a general-purpose speech recognition model trained on a large and diverse audio dataset. Whisper is a multi-task model that can perform multilingual speech recognition, translation, and language identification.
January 2023: OpenAI Faces Criticism Over Data Annotation Practices
OpenAI was criticized for outsourcing data annotation to Sama, a company employing workers in Kenya. The annotations, used to train an AI model for detecting toxicity, exposed workers to graphic content, raising ethical concerns and prompting discussions about worker compensation.
January 23, 2023: Microsoft Invests $10 Billion in OpenAI
Microsoft announced a significant investment of $10 billion in OpenAI, aiming to integrate ChatGPT into its products and services.
February 7, 2023: Microsoft Integrates AI into its products
Microsoft revealed its plans to incorporate AI technology based on ChatGPT into various products, including Bing, Edge, and Microsoft 365.
March 3, 2023: Reid Hoffman Resigns from OpenAI Board
Reid Hoffman stepped down from his position on OpenAI's board to prevent conflicts of interest due to his investments in other AI companies.
March 14, 2023: OpenAI Releases GPT-4
OpenAI launched GPT-4, a more advanced language model, available through an API and as a feature in ChatGPT Plus.
March 2023: OpenAI Criticized for Lack of Transparency
OpenAI faced criticism for limited transparency regarding technical details of models like GPT-4, deviating from their initial commitment to openness and hindering independent research and safety development. OpenAI cited competitiveness and safety concerns as justification for this shift.
March 23, 2023: Codex API Support Discontinuation
OpenAI announced the discontinuation of support for the Codex API.
April 2023: EU Forms ChatGPT Task Force
The European Data Protection Board (EDPB) established a task force dedicated to ChatGPT to investigate potential GDPR violations and coordinate enforcement actions among data protection authorities.
May 2023: ChatGPT App Launched on iOS
OpenAI launched a dedicated ChatGPT app for iOS, supporting chat history syncing and voice input powered by their speech recognition model, Whisper.
May 22, 2023: OpenAI leaders propose Governance of Superintelligence
OpenAI leadership published recommendations for governing superintelligence, emphasizing the need for international oversight and increased safety research.
July 2023: Authors Sue OpenAI for Copyright Infringement
Authors Sarah Silverman, Matthew Butterick, Paul Tremblay, and Mona Awad filed a lawsuit against OpenAI for copyright infringement, alleging unauthorized use of their work.
July 2023: ChatGPT App Launched on Android
OpenAI launched a ChatGPT app for Android, expanding its accessibility to a wider user base.
August 2023: OpenAI Sued for Violating EU GDPR
OpenAI faced a lawsuit for allegedly violating EU General Data Protection Regulations, raising concerns about data privacy and compliance.
August 2023: OpenAI Acquires Global Illumination
OpenAI acquired Global Illumination, a startup specializing in AI-driven creative tools and digital infrastructure.
September 2023: ChatGPT Gains Multimodal Capabilities
OpenAI announced that ChatGPT can now "see, hear, and speak". ChatGPT Plus users gained the ability to upload images, while mobile app users could interact with the chatbot using voice.
September 2023: Authors Guild Joins Copyright Lawsuit Against OpenAI
Seventeen authors, including prominent names like George R. R. Martin and John Grisham, joined the Authors Guild in a class action lawsuit against OpenAI, alleging illegal use of their copyrighted works in training OpenAI's technology.
September 2023: DALL-E 3 Announcement
OpenAI announced DALL-E 3, a more advanced model able to generate images from complex descriptions without manual prompt engineering, and render complex details.
September 21, 2023: Microsoft Rebrands Copilot Products
Microsoft consolidated its various Copilot products under the single brand "Microsoft Copilot," including Bing Chat and Microsoft 365 Copilot.
October 2023: OpenAI partners with G42
OpenAI announced a partnership with G42, an Emirati AI company, allowing G42 to utilize OpenAI technology.
October 2023: DALL-E 3 Integrated into ChatGPT
OpenAI integrated DALL-E 3, their latest image generation model, into ChatGPT Plus and ChatGPT Enterprise. ChatGPT was used to write prompts for DALL-E based on user conversations.
November 6, 2023: OpenAI Launches GPTs
OpenAI introduced GPTs, enabling users to create custom versions of ChatGPT for specific purposes.
November 14, 2023: OpenAI Suspends New ChatGPT Plus Sign-Ups
OpenAI temporarily halted new subscriptions for ChatGPT Plus due to high demand, with access restored a month later.
November 2023: OpenAI Reorganization
OpenAI undergoes a significant reorganization, including the return of Sam Altman as CEO and changes to the board of directors. This event is perceived by some as a shift towards a greater business focus and a potential reduction in the influence of individuals prioritizing caution within the organization.
November 17, 2023: Leadership Shakeup and Reinstatement at OpenAI
In a surprising turn of events, the OpenAI board removed CEO Sam Altman and Chairman Greg Brockman from their positions. However, just four days later, both Altman and Brockman returned to their roles after negotiations with the board, leading to the resignation of most board members. A new board was formed with Bret Taylor, former Salesforce co-CEO, as chairman, and Microsoft gained a non-voting board seat.
November 18, 2023: Talks of Sam Altman's Return as OpenAI CEO Begin
Discussions emerged regarding Sam Altman's potential return as OpenAI's CEO due to pressure from investors like Microsoft and Thrive Capital who opposed his removal.
November 19, 2023: Negotiations for Altman's Return Fail; Emmett Shear Appointed Interim CEO
Negotiations for Sam Altman's return to OpenAI broke down, leading to Emmett Shear's appointment as interim CEO. The board's attempts to replace Altman with Anthropic CEO Dario Amodei and to merge the companies were declined.
November 20, 2023: Altman and Brockman Join Microsoft, Negotiations with OpenAI Resume
Microsoft announced that Sam Altman and Greg Brockman would lead a new AI research team, while still maintaining their commitment to OpenAI. In response to a potential mass resignation of OpenAI employees, negotiations with Altman and the board restarted.
December 2023: The New York Times Sues OpenAI
The New York Times filed a lawsuit against OpenAI for copyright infringement.
December 2023: Microsoft Expands Copilot Integration
Microsoft extended the availability of Microsoft Copilot to Windows operating systems and released standalone apps for Android and iOS.
2023: OpenAI Delays GPT Store Launch
OpenAI's GPT Store, initially planned for a 2023 launch, was postponed to early 2024, likely due to leadership changes following the initial announcement.
2023: OpenAI Revenue Projections
Anonymous sources indicated that OpenAI Global LLC projected revenue of $200 million for 2023.
2023: Microsoft's $10 Billion Investment
Microsoft further solidified its partnership with OpenAI by investing an additional $10 billion, a large portion of which was dedicated to computational resources on Microsoft's Azure cloud service.
2023: OpenAI Introduces ChatGPT Plus and GPT-4 API
OpenAI introduced ChatGPT Plus, a GPT-4 powered version of ChatGPT, available for a monthly subscription fee. They also made GPT-4 accessible to select applicants through their API waitlist, charging fees based on token usage.
February 2024: The Intercept and Others Sue OpenAI for Copyright Infringement
The Intercept, Raw Story, and Alternate Media Inc. filed a lawsuit against OpenAI for copyright infringement, introducing a new legal strategy for digital publishers in copyright litigation against OpenAI.
April 2024: NOYB Files GDPR Complaint Against OpenAI
NOYB filed a complaint with the Austrian Datenschutzbehörde (Data Protection Authority) against OpenAI for violating the GDPR. The complaint centered around ChatGPT generating inaccurate personal data and the lack of transparency and user control over data processing.
May 2024: OpenAI Admits Destruction of Training Datasets
OpenAI revealed that they had destroyed their Books1 and Books2 training datasets, which were used to train GPT-3 and were believed to contain over 100,000 copyrighted books, further escalating the copyright controversy.
2024: OpenAI Revenue Projections
Anonymous sources indicated that OpenAI Global LLC projected revenue of $1 billion for 2024.
2024: GPT Store Launch Further Delayed
The launch of OpenAI's GPT Store was further delayed, with no confirmed date set.