OpenAI reveals new details about its AI development roadmap and fundraising plans

OpenAI LP is working on GPT-5 and plans to raise more capital from Microsoft Corp. to support its development efforts, Chief Executive Officer Sam Altman has disclosed in a new interview.

The interview, which was published today in the Financial Times, comes a week after OpenAI’s first annual developer conference. The event saw the company introduce new customization options for GPT-4 and a raft of other features. GPT-5 was also on the agenda: Altman said during the event that OpenAI expects its next flagship language model to offer more advanced features than current neural networks, but didn’t go into detail.

Altman told the Financial Times today that OpenAI engineers are already working on GPT-5. He reiterated his view that the model will be more advanced than GPT-4, but said it’s technically difficult to predict in exactly what way. Altman also declined to provide a release date.

One detail the executive did divulge is that GPT-5 will take more data to train than OpenAI’s earlier models. The company plans to source that information from other organizations and publicly available sources, Altman stated. Last week, OpenAI launched a partner initiative through which companies can contribute training data to its artificial intelligence projects. 

GPT-4, OpenAI’s current flagship AI, represented a major upgrade over the company’s earlier neural networks when it debuted in March. Compared with its predecessors, GPT-4 can answer a broader range of questions and in a more accurate manner. At its developer event last week, OpenAI introduced an improved version of GPT-4 that it says processes user prompts even more reliably. 

Building large-scale language models is an expensive endeavor. Altman told the Financial Times that OpenAI expects “to raise a lot more over time” from Microsoft Corp. and other investors to finance its development efforts. Earlier this year, Microsoft reportedly invested $10 billion in the AI developer at a $29 billion valuation.

Sources told Bloomberg last month that OpenAI is exploring a secondary sale at a $86 billion valuation. During a secondary sale, a company’s existing investors sell stock but the company itself doesn’t take in new capital. Based on the Bloomberg report, OpenAI’s next funding round could close at a much higher valuation than its most recent $10 billion raise.

Graphics processing units likely account for a sizable portion of OpenAI’s development costs. Nvidia Corp.’s flagship data center GPU, the H100, currently sells for about $40,000. Altman told the Financial Times that despite a “brutal crunch” in the supply of Nvidia chips, OpenAI has already started receiving H100s and expects more to arrive by year’s end. “Next year looks already like it’s going to be better” in terms of supply, he added.

The executive also briefly discussed OpenAI’s finances. He disclosed that the company is not yet profitable, but added “revenue growth had been good this year.” According to The Information, OpenAI’s annual revenue run-rate climbed to $1.3 billion in October from $1 billion two months earlier and just $28 million last year.

The long-term goal of OpenAI’s engineering investments, Altman detailed, is to develop artificial general intelligence. This is the term for a hypothetical AI model that can perform a wide range of tasks and learn new ones in a manner on par with humans. Altman told the Financial Times that large models are “one of the core pieces . . . for how to build AGI, but there’ll be a lot of other pieces on top of it.”

Image: OpenAI