OpenAI is a non-profit organization that has brought Generative-AI to the mainstream despite being a pioneer in the industry. The company was founded in 2015 in San Francisco as a joint initiative. Since then, they have been promoting AI research and the deployment of Friendly General Intelligence (AGI) that allows users to interact in natural language.
Apart from AI, OpenAI is also working on Natural Language Processing, Robotics, Computer Vision, and Reinforcement learning, but they are primarily focused on developing AGI. According to Reuters, this year, OpenAI reached roughly 100 million active users in January.
Today, we will talk in-depth about OpenAI, its impact, and its future. The company is still young among its competitors, holding various projects that have the potential to revolutionize how we discover things, analyze data, and boost individual productivity.
OpenAI Founding Members
Twelve prominent figures in the tech industry played a vital role as founding members, serving as initial board members and investing over $1 billion in the company.
Courtesy: Sam Altman
- Elon Musk
- Sam Altman
- Ilya Sutskever
- Greg Brockman
- Trevor Blackwell
- Vicki Cheung
- Andrej Karpathy
- Durk Kingma
- Jessica Livingston
- John Schulman
- Pamela Vagata
- Wojciech Zaremba.
The company has also secured a $1 billion grant from the US National Science Foundation. OpenAI has collaborated with several organizations for its research and development, including Amazon Web Services (AWS), Infosys, YC Research, IBM, Google, and Microsoft.
OpenAI AI-Model
A handful of tool companies have released these models, which have undergone multiple iterations since their initial release.
- DALL·E
- The company’s first-ever product for the public is an AI image generator.
- ChatGPT
- Last year, OpenAI released the ChatGPT model, making OpenAI one of the hottest tech brands in the industry.
First OpenAI Model
OpenAI first released an AI tool called Universe for video games to train AI agents to perform complex moves like throwing kicks and punches in 2016. That same year, they released OpenAI Gym for reinforcement learning research.
Bill Gates says it is a huge milestone in AI.
History of OpenAI
Let us first know how it all began. It all started with John McCarthy, who used the term “Artificial Intelligence” in 1956 after Alan Turing founded the first modern computer in the early 1950s. After that, it continued to develop interest among research institutions, and they developed the first machine learning and natural language processing (NLP) in 1997. After this foundation, in the early 2000s, artificial intelligence saw significant advancements with more refined algorithms and computing power to process big data faster.
OpenAI came into existence when former Y Combinator CEO Sam Altman and Elon Musk came together to build a safe AGI. Before building AGI, they were aware of the threat to humanity and the opportunities of AI technology. The company did not begin to start leveraging the open-source toolkit until 2018. After the first release of their first OpenAI tool in 2016, the company released a research paper titled “Improving Language Understanding by Generative Pre-Training,” which was the first draft of a generative AI model inspired by the structure and function of the human brain. This model generated things by performing many functions based on a prompt. In this paper, they shared the concept of a Generative Pre-Trained Transformer (GPT), a machine-learning model trained on a large dataset of human-generated text.
Founder Timeline
- 2018: Elon Musk resigned due to a conflict of interest with Tesla’s AI self-driving cars.
- 2023: Reid Hoffman resigned due to a conflict of interest with his investment in Greylock Partners.
Timeline
- 2015: Founded
- 2016: Launched OpenAI GYM and Universe.
- 2017: Spent $7.9 million on cloud computing and launched RoboSumo.
- 2018: Started training OpenAI’s Dota 2 Bots on Google Cloud (specs: 128,000 CPUs and 256 GPUs).
- Announced GYM Retro, Debate Game, and Dactyl.
- 2019: Established OpenAI LP (a hybrid of non-profit and for-profit).
- Announced GPT-2, MuseNet, and Jukebox.
- Partnered with Microsoft to run OpenAI models on Azure-based supercomputers.
- After transitioning from non-profit to a hybrid of non-profit and for-profit, OpenAI spent $1 billion and compensated with equity as legally required.
- 2020: Announced GPT-3 and an API to let developers develop their OpenAI-based products and services.
- 2021: Announced DALL-E.
- 2022: Released ChatGPT 3.5 to the public and exclusively licensed it to Microsoft.
- Received 3 million sign-ups within five days.
- Projected $200 million in revenue in 2023 and $1 billion in 2024, according to reports.
- Announced Whisper (speech recognition model) and Microscope.
- 2023: Announced a multi-year $10 billion investment with Microsoft.
- Reportedly, Microsoft acquired 49% stakes in OpenAI, and Microsoft receives 75% profit until the investment returns.
- Announced DALL-E 2, GPT-4, ChatGPT Plus, and API.
OpenAI Fundraise
- 2019: OpenAI announced a $1 billion venture, as reported by its founders and partners.
- 2019: Microsoft invested $1 billion.
- 2023: Microsoft invested $10 billion.
Evolution of the GPT Model
The company released the GPT Model to the public when GPT-3 because they feared people would start misusing it. The company implemented a never-before-seen database with every iteration that was able to generate text, and now it is also able to adjust based on conversation history.
How OpenAI GPT Works
It uses machine learning algorithms and deep learning to process large datasets. The more data it has, the more capable it becomes.
- GPT-1
- It was first trained on BookCorpus’s 7000 unpublished books.
- GPT-2
- Trained on 8 million web pages, OpenAI gathered more than 1.5 billion parameters that help to make text predictions.
- GPT-3
- This was the first release to the public, and was trained on 45TB of text data, which gave the company 175B parameters. To handle this massive data, Microsoft helped OpenAI with a supercomputer with 285,000 CPU cores and 10,000 GPUs.
- GPT-4
- This is the latest GPT model, which was released after GPT-3. It was 40% better than GPT-3 and can also understand visual input, meaning it can now handle creative tasks. This model is only accessible to ChatGPT Plus Subscribers or via API.
OpenAI has also released DALL·E, which was trained on a similar architecture as the GPT-2 model but is used to generate images instead of text. It was trained on 12 billion parameters.
How can I use the GPT model?
It is available for everyone to use. To get started, head to OpenAI Chat, create a new account and then begin using ChatGPT by entering “promoted” into the text box and hitting enter to get an answer.
Is OpenAI a non-profit or for-profit organization?
This was the primary reason Elon Musk left the OpenAI Board in 2018. OpenAI transitioned from a non-profit organization to a hybrid of for-profit and non-profit organizations. The company is slowly moving away from its original idea of developing open, accessible, and safe AI development because it is hard to run OpenAI as a non-profit since they have been paying $1.9 million to OpenAI employees, as reported, which is not comparable to Facebook or Google. However, employees were ready to switch from Google Meta to OpenAI because of its mission.
OpenAI is starting to prioritize commercial applications and needs to focus on the risks associated with generative AI. The company announced partnerships with Microsoft in February 2023, where OpenAI models will be integrated within Microsoft’s products and services like Edge, Bing Search Engine, GitHub, Windows, Microsoft 365, and others. For this, Microsoft has funded the company with $10 billion as a strategic partnership.
OpenAI’s Impact on Society
AI systems are generally more intelligent than humans, unlocking limitless possibilities for everyone and allowing access to new scientific knowledge. They also contribute to the global economy and increase individual productivity. However, it is essential to remember that “with great power comes great responsibility.”
It is also acknowledged that some individuals may misuse OpenAI’s capabilities to create societal disruption that can harm humanity or unduly concentrate power. Therefore, the company has implemented filters on top of its AI to reduce the risk and is working with policymakers and institutes to understand the benefits and downsides of AI systems. As AGI continues to overgrow, OpenAI upholds the responsibility to take more caution with creating and developing its AI models.
OpenAI also shares its research by making its patents and research open to the public.
It is not easy to predict the future of AGI, but OpenAI tools have shown the potential of AGI to the world. AI has been used for years, but generative AI benefits education, marketing, finance, and other fields. It is helpful, especially with repetitive tasks.
OpenAI Competitors
Before OpenAI released its model, Meta and Microsoft had already released their AI models. However, they were pulled down after a couple of days due to being racist and criticized for being inaccurate. Tech giants like Meta and Google are also working on similar models but currently offer limited access due to exhibiting problematic behaviour. After the release of Microsoft’s Bing Chatbot, which used the OpenAI GPT model, Google has also released its Bard based on LamDA and PALM models. Meta has also released AI VoiceBox, and there are plans to bring AI to its core products and services soon. OpenAI also allows new tech startups to emerge, such as Midjourney, Playground.ai, and more.