For several months now, only one company has been in the news: OpenAI. And with good reason, since it's the company behind ChatGPT and Dall-E, two hugely successful artificial intelligence tools.
You've probably seen this logo if you've ever used or heard of ChatGPT. It's actually the logo of OpenAI, the company behind ChatGPT and its language model, GPT. Who's behind OpenAI, one of the world's most powerful artificial intelligence companies?
What is OpenAI?
OpenAI presents itself as "an AI research and deployment company" with the aim of democratizing this technology. The company was originally a non-profit association, founded in December 2015, by two presidents: Sam Altman and multi-billionaire Elon Musk, CEO of Tesla, Twitter and Neuralink among others, along with several other members, all from tech backgrounds. Initially, Musk was to invest a total of $1 billion, but "only" invested $100 million. The reason: after results that were not good enough for him, Musk wanted to take control of OpenAI, which was refused by other founding members. He therefore left the board of directors in February 2018.
In 2019, it became a capped for-profit company: that is, the company can generate profits, but only up to a certain level. A decision potentially due to the departure of Elon Musk and, by extension, his investments.
The dual benefit of this status for OpenAI is to attract investors seeking a return on the company, while limiting the potential excesses of the dividend race. A "capped profit", but with a very high ceiling: one hundred times the value of the initial investment.
How was the company created?
Sociologist Antoine Goujon tells the story of OpenAI's creation in AOC. Elon Musk and Sam Altman commissioned Greg Brockman, former CTO of Stripe (an online payments company for businesses), to put together a team of artificial intelligence specialists, mixing engineers and researchers.
OpenAI offices in San Francisco // Source : OpenAI
To convince them, OpenAI, or rather its founding members, banked on the transparency of what was then an organization, particularly in scientific research. At the start of 2017, OpenAI had 45 employees, but today that's all changed, with nearly 400.
OpenAI philosophy: from openness to regulation
In its early days, then, OpenAI was intended to be open (it's in the name of the entity): publication of program source codes, dissemination of scientific research results, and so on. But, since 2015, OpenAI has made a pivot in its philosophy: while openness is not over, it has simultaneously suffered a brake.
OpenAI has carried out simulated examinations to differentiate between the two models // Source : OpenAI
The company recently admitted: "We were wrong in our initial thinking about openness, and we've changed our minds: we no longer think that we need to release everything (although we do open source some things and plan to do so in the future), but rather that we need to find a way to securely share access to systems and their benefits." An important change philosophically, but one that is intended to be secure: ChatGPT and Dall-E, the company's two most important tools, can be used for unfair purposes. Because of their power, OpenAI's responsibility is even greater.
Who funds OpenAI?
OpenAI's initial cash flow was $1 billion, $100 million of which was brought in by Elon Musk and some by Amazon Web Services, Amazon's hosting and cloud computing division. But it was in 2019 that things really accelerated, with Microsoft investing $1 billion in what had become a company.
Then, with the insane success of ChatGPT, Microsoft and OpenAI extended their partnership in early 2023, with the Redmond firm officially investing several billion dollars. The main rumor was ten billion dollars, according to Semafor. Microsoft would recover 75% of OpenAI's profits, until it had paid back the ten billion dollars. Although the firm would own "only" 49% of the shares, with 49% going to other investors and 2% to OpenAI's non-profit foundation.
At present, however, the precise ownership structure of OpenAI is not known. Nor is the company's valuation publicly known. But, at the time of a $300 million funding round documented by TechCrunch in April 2023, the valuation had been estimated at between $27 and $29 billion.
OpenAI's main artificial intelligence products
OpenAI's popularity and power are based primarily on two AI tools: ChatGPT and Dall-E.
ChatGPT and GPT, its language model
It's hard to introduce someone to ChatGPT for the first time, so much has been said about the chatbot. What OpenAI has developed is more like GPT, the large language model (or LLM), a kind of ChatGPT engine. The chatbot can answer a whole host of questions, without really understanding what it's saying.
In fact, GPT is essentially an algorithm that predicts the word to follow based on a given context and on previously generated words. For this reason, it can't be said to have the reflection of a human being (or even a living being).
Dall-E
Dall-E made the headlines about a year ago with the release of Dall-E 2, before seeing its popularity overtaken by ChatGPT, as well as by competitors such as Midjourney and Stable Diffusion. Dall-E is an artificial intelligence program for image generation.
Given a textual description, Dall-E is able to generate images based on a gigantic database with which it has been trained. To "understand" the indications it is given, it relies on GPT, the language model also used in ChatGPT.
Other OpenAI innovations
However, OpenAI is not a company known only for ChatGPT. It has also developed Microscope, an AI tool for visualizing neural networks, and Whisper, a speech recognition model.
Other artificial intelligence models are being developed within the company. These include Point-E, which generates 3D objects, and its successor, Shap-E.
Finally, we might mention Jukebox, described as "a neural network that generates music, including rudimentary vocals, as raw audio in a variety of genres and artist styles." The model was released in 2020, but since then nothing.
What does the future hold for OpenAI?
With the recent investments made by Microsoft and the trend towards artificial intelligence in the new technology industry, the future looks bright for OpenAI.
It's safe to assume that the company already has a next version of its language model used by ChatGPT in the pipeline. Although, for the moment, OpenAI claims not to be working on GPT-5.
Towards general artificial intelligence
From the outset, OpenAI has been a firm believer in "general artificial intelligence" (GAI). This term can be defined simply as AI capable of learning and performing any task. For the moment, no AI model can be described as such, but the company is working towards this. It even imagines an AI that could be smarter than the human brain. But it sees this in a positive light: AGI would have "the potential to give everyone incredible new capabilities; we can imagine a world in which every one of us has access to help with almost any cognitive task, greatly multiplying the power of human ingenuity and creativity," reads a recent blog post.
In it, the company explains that "as we create increasingly powerful systems, we want to deploy them and gain experience of how they work in the real world." A statement that remains vague, but which makes it clear that OpenAI wishes to integrate its tools and AI into human uses, to go beyond simple research and demonstration. OpenAI says it intends to be increasingly cautious in launching new AI models, especially as generative artificial intelligences will most likely be more heavily regulated by jurisdictions around the world.
0 Comments