OpenAI’s building GPT-5 – but it hasn’t got a clue what it will do
OpenAI is hard at work developing GPT-5 – the next generation of its AI large language model – but even the company’s CEO admits he can’t say what it will do.
OpenAI’s GPT-4 is one of the most prominent AI language models in the world, powering AI services including Microsoft’s Copilot offerings, the Bing AI chatbot and, of course, the company’s own ChatGPT.
GPT-4 was released in March and development work is well underway on GPT-5, but the company’s CEO Sam Altman says it’s too early to say how it will differ from what’s gone before.
“Until we go train that model, it’s like a fun guessing game for us,” Altman said in an interview with the FT. “We’re trying to get better at it, because I think it’s important from a safety perspective to predict the capabilities. But I can’t tell you here’s exactly what it’s going to do that GPT-4 didn’t.”
Recommended reading: What is Sora? Even the AI experts aren’t sure
Paying for GPT-5 and AGI
Developing large language models is a blisteringly expensive business, made even more costly by recent shortages of the Nvidia H100 chips that most of the big AI companies use to build their AI systems.
Also in short supply is high quality data to train the next generation of AI with, with OpenAI recently putting out a call for large-scale data sets, particularly those focusing on long-form writing.
The cost of acquiring the supercomputing power and the data necessary to improve AI models means OpenAI will once again be turning to partners such as Microsoft for additional investment, Altman claims. Microsoft has already ploughed $10 billion into OpenAI as part of a multi-year partnership, but the OpenAI CEO said the company will need more money to reach so-called AGI – artificial general intelligence – where AI achieves human-like levels of sophistication.
When asked by the FT whether Microsoft would continue to invest in his company, Altman replied “I hope so,” before adding: “There’s a long way to go, and a lot of compute to build out between here and AGI… training expenses are just huge.”
The big problem for OpenAI is it’s far from the only player in the game. Google, Amazon, Elon Musk’s x.AI and others are all investing heavily, which not only pushes up prices for those in-demand AI processors, it increases demand for training data and boosts competition for customers to pay for the resulting services.
NEXT UP
Ghostbusters proton packs in real life
Would Ghostbusters proton packs be useful in the real world? Richard Trenholm speaks to scientist James Maxwell to find out.
OpenAI’s “magic” GPT-4o update is an iterative release that may herald less AI hype
OpenAI unveils update to GPT-4o, wider access for free users and a desktop version – so does this mark the end of generative AI hype?
What is GPT-4o?
Fresh from the announcement of GPT-4o, we explain what it is, what it can do and how much it costs