OpenAI is working on the GPT-5 artificial intelligence model and plans to attract additional investments from Microsoft that will contribute to the development of its projects. The CEO of the company “Sam Altman” (Sam Altman) told about this in an interview with the Financial Times.

According to Altman, OpenAI engineers are already working on GPT-5. The new artificial intelligence model will be more advanced compared to GPT-4, but technically it is still difficult to predict in what respect. The company’s president declined to specify the platform’s release date, but noted that more data is needed for training than when building previous OpenAI models. The developer plans to obtain the necessary information from publicly available sources and other companies: Last week, the company launched a partnership initiative in which third-party organizations can provide datasets for training their own projects. AI field.

The current OpenAI flagship product, GPT-4, introduced in March, was a significant step forward compared to its predecessors: the neural network learned to answer a wider range of questions and began to do so more accurately. Last week, the company introduced an improved version of GPT-4 Turbo.

Developing large language models is an expensive task. OpenAI expects to attract “many more funds over time” from Microsoft and other investors to implement its projects. The software giant has already invested $10 billion in the startup at a valuation of $29 billion. OpenAI is also considering the possibility of a secondary sale of assets valued at $86 billion. itself does not attract new capital.

A significant portion of the cost of AI training relates to hardware accelerators: The flagship model of NVIDIA H100 currently costs around $40,000. Despite ““severe crisis” As for the supply of accelerators, Altman reported that OpenAI has started receiving them and expects a new batch by the end of the year. According to him, the supply situation may improve next year. Last year was financially very successful for the company, but it has not yet reached profitability.

The long-term goal of OpenAI’s engineering investments is the development of powerful artificial intelligence (Artificial General Intelligence – AGI). This term refers to a hypothetical artificial intelligence model that can perform a wide range of tasks at the same level as humans and learn new ones. The release of large language models is an important step on the path to AGI, “but there will be many other things beyond that.”