OpenAI is a non-profit organization dedicated to the research and application of artificial intelligence. Its goal is to develop safe and universal artificial intelligence (AI) that benefits all people. To accomplish this, the original non-profit organization has partnered with a new for-profit arm.
In late 2015, the nonprofit organization OpenAI was founded with the goal of developing safe and useful artificial intelligence for the benefit of humanity. Because this project would be very capital-intensive, OpenAI launched with the goal of raising one $1 billion in donations. However, donations alone were not enough to cover the cost of the computing power and the specialists that would be needed to move the basic research forward.
In 2019, OpenAI entered into a strategic partnership with Microsoft. Although this partnership with Microsoft involves a multi-billion dollar investment, OpenAI remains a fully independent company governed by the non-profit OpenAI organization. In addition, OpenAI is explicitly excluded from all commercial and IP licensing agreements. Microsoft has a non-voting observer status on the board and exercises no control.
June 11, 2020: The company releases its first commercial product: OpenAI API, an API (application programming interface or programming interface, more specifically, an interface for programming applications) for accessing new AI models. Unlike most AI systems, which are designed for a specific use case, the API provides a universal „text in, text out“ interface. This allows users to test the technology for virtually any English-language task. The API features a powerful universal language model, GPT-3, and has received tens of thousands of submissions to date.
Why did OpenAI decide to release an API instead of releasing the models themselves?
There are three main reasons for this decision. First, commercializing the technology will help fund ongoing activities in AI research, security, and policy.
Second, many of the models underlying the API are very large and require a high level of expertise to develop and implement. This makes it difficult for all but the largest enterprises to benefit from the underlying technology. In this way, the API can make powerful AI systems more accessible to smaller companies and organizations.
Third, responding to misuse of the technology is easier with the API model. Since it is difficult to predict future use cases, it appears more secure than an open source model where access is not adjustable if malicious applications are discovered.
March 25, 2021: Nine months after the launch of the first commercial product, the OpenAI API, more than 300 applications are using GPT-3 and tens of thousands of developers worldwide are building on the platform. Currently, an average of 4.5 billion words are being generated per day.
A community of tens of thousands of developers is growing around the world, with most of them located in North America, Europe, Asia, and Australia. Many are from outside the traditional AI or software development fields and have had their first experience with an API or programming through the OpenAI interface.
November 30, 2022: ChatGPT launched
ChatGPT is an artificial intelligence trained to follow a prompt and provide a detailed response. ChatGPT has been refined on the basis of a model from the GPT-3.5 series, which completed its training in early 2022. ChatGPT and GPT-3.5 were trained on an Azure AI supercomputing infrastructure.
The new AI will be available for free and is designed to receive user feedback so AI developers can learn more about the product’s strengths and weaknesses.
September 25, 2023: New voice and image features have been added to ChatGPT. The user will now be able to have a conversation with an assistant by using voice input. A new text-to-speech model supports the new voice feature. This model is able to generate a human-like voice from text and a few seconds of samples spoken by professional voice actors. It also uses Whisper, an open source speech recognition system, for text-to-speech conversion.
With the help of one or more pictures, ChatGPT can now find out why, for example, the barbecue grill won’t start. It can examine the contents of the refrigerator in order to plan a meal, or it can analyze a complex graphic in order to obtain work-related data.
November 6, 2023: Without any programming skills, users can now create their own versions of ChatGPT, called GPTs. GPTs are a new way for everyone to create a customized version of ChatGPT that helps in everyday life, with specific tasks, at work or at home. For example, GPTs can be used to teach children mathematics or to design stickers.
To the German translation of this article: OpenAI