OpenAI has started building ChatGPT 5 — its next-generation AI model GPT-5. CEO Sam Altman confirmed this in a recent interview, and claimed it could possess superintelligence, but the company would need further investment from its long-time partner Microsoft to make it a reality.
Speaking to the Financial Times, Altman said the partnership with Microsoft is working really well, and that he expects to raise a lot more money over time from the Windows creator and other investors.
Building a major AI model like ChatGPT requires billions of dollars and masses of computer resources, training on billions or trillions of pages of data, and extensive fine-tuning and safety testing.
Going beyond the human
While GPT-4 is an impressive artificial intelligence tool, its capabilities come close to or mirror the human in terms of knowledge and understanding. The next generation of AI models is expected to not only surpass humans in terms of knowledge, but also match humanity’s ability to reason and process complex ideas.
This is also known as artificial general intelligence (AGI), which goes beyond simply parroting a new version of what it is given and provides an ability to express something new and original. It is this type of model that has had governments, regulators and even big tech companies themselves debating how to ensure they don’t go rogue and destroy humanity.
Billions to train the next AI
Microsoft has already pumped more than $10 billion into OpenAI as part of a multi-year agreement, with Altman saying he hopes more will follow as "there is a long way to go and a lot of compute to build out between here and AGI."
Microsoft has shifted its entire business model around the use of AI with Copilot running front and center in Windows and various applications. So you can see how the investment will benefit the company's huge move into this field.
However, for Altman the true product is intelligence. He told FT: “Microsoft has shifted its entire business model around the use of AI, embedding its Copilot system — built on top of GPT-4 — into Windows, Microsoft 365 and other products.”
Should we be worried?
GPT-5 will require more processing power and more data than ever before, which Altman says will come from a combination of publicly available data found online, as well as data it buys from companies. It has called out for datasets not widely available including written conversations and long-form writing.
He hasn't set a timeline for GPT-5 or exactly what capabilities it might have as it is impossible to tell until it is finished. Training the model is expected to take months if not years with availability to the public unlikely for some time after it is finished training — so there is still time to build a bunker, get offline and hide from Skynet.
More from Tom's Guide
