OpenAI is already working on GPT-5

Artificial Intelligence

OpenAI LP is working on GPT-5 and intends to raise more capital from Microsoft to support its development efforts, Chief Executive Officer Sam Altman revealed in an interview. The interview came out in the Financial Times today and landed a week after OpenAI held its first annual developer conference.

OpenAI unveiled new customization options for GPT-4 and other features during the event. GPT-5 was also on the agenda, as during the event, Altman said that OpenAI expects its next flagship language model to offer advanced features that current networks aren’t capable of but didn’t offer details. Altman also said OpenAI engineers are at work on the next iteration, which he believes will leap beyond GPT-4.

However, he said it is still hard to tell with certainty how it will be better. Altman did not have a release date to offer either but said GPT-5 will need more data to train than its predecessors. The company intends to outsource the information it needs from other organizations in the field and publicly available sources.

Raising money

Last week, OpenAI announced a partner initiative through which companies can contribute training data to its myriad AI projects. The company’s current flagship, which debuted in March, GPT-4, represented a significant leap compared to the company’s earlier neural network.

At last week’s event, the company introduced an improved variant of GPT-4 that reportedly uses prompts even more reliably. During the interview, Altman repeated that building large-scale language models is expensive and that OpenAI hopes to raise more over time from Microsoft and other investors to power development.

Microsoft has obliged and invested $10 billion in OpenAI earlier this year at a $29 billion valuation. According to Bloomberg last month, OpenAI is considering a secondary sale at a valuation of $86 billion. A company’s current investors sell stock in a secondary sale, but it takes in no new capital. The Bloomberg report said OpenAI’s next funding round could conclude at a much higher valuation than the recent $10 billion raise.

Costs and profitability

Most of the costs come from the company’s need for graphics processing units. Nvidia’s flagship data center GPU, the H100, sells for about $40,000. Even with the crunch, Altman has already started getting H100s and expects more before year’s end, and he even said, “Next year looks already like it’s going to be better” in terms of supply.

The CEO also discussed the company’s finances and disclosed that the company isn’t profitable yet but added that revenue growth has “been good” in 2023. According to data from The Information, OpenAI’s annual revenue run-rate rose to $1.3 billion in October from $1 billion two months before that and only $28 million last year.

Altman says the company’s ultimate goal is to develop artificial general intelligence, a term used to describe a hypothetical AI model capable of performing a wide range of tasks and learning new ones like humans do.

Share to...