OpenAI Japan shares vision for much more powerful “GPT Next” coming in 2024
OpenAI Japan has announced significant performance improvements for OpenAI’s upcoming AI models, expected before the end of this year.
At the KDDI Summit, OpenAI Japan provided insights into the company’s next generation of AI models. A new version dubbed “GPT Next” is planned for 2024, promising a substantial leap in capabilities.
The graph presented by OpenAI Japan shows a significant increase in performance. While GPT-3 and GPT-4 are relatively close in capability, GPT Next is projected to make a much larger jump, increasing performance by a factor of 100, according to OpenAI Japan CEO Tadao Nagasaki. The slide is titled “OpenAI Vision,” suggesting that it’s not actual math – but still.
According to Japanese ITmedia, Nagasaki, who became head of OpenAI Japan in April 2024 after serving as President of Amazon Web Services Japan, emphasized that AI technology is developing exponentially compared to conventional software.
Ad
These statements echo earlier announcements from OpenAI and its partners. At Microsoft’s Build developer conference in May, CTO Kevin Scott showed a similar graphic suggesting a much more powerful OpenAI model by the end of 2024.
At the Viva Technology festival in Paris in May, an OpenAI researcher also referred to the 2024 model as “GPT Next,” ranking it well above GPT-4. And OpenAI CEO Sam Altman has repeatedly promised further major advances in AI, saying in early May that GPT-5 will be “a lot smarter” than GPT-4.
When will the GPT-4 era come to an end?
According to a report by The Information, OpenAI is currently working on two new AI systems: “Strawberry,” an AI model with improved capabilities in areas such as mathematics and programming, and “Orion,” which could become the actual successor to GPT-4 and thus “GPT Next.”
Orion is expected to act more reliably and logically thanks to high-quality training data generated by Strawberry. OpenAI has not yet commented on these rumors.
