Unlock the Editor’s Digest for free
Roula Khalaf, Editor of the FT, selects her favourite stories in this weekly newsletter.
OpenAI has unveiled advances to its flagship artificial intelligence model, intensifying the competition with Google and other Big Tech groups pushing for breakthroughs in the technology.
The San Francisco start-up demonstrated a series of improvements to its GPT-4 model at an event on Monday, including the ability to interpret voice, video, images and code in a single interface — though it stopped short of revealing a much-anticipated new model.
The update, billed as GPT-4o, “provides GPT-4 level intelligence, but it’s much faster and improves on capabilities across text, vision and audio”, chief technology officer Mira Murati said, before demonstrating live voice translation across languages. Shares in language-learning tool Duolingo fell about 5 per cent following the announcement.
The updates come just one day before Google’s annual developer conference, which is expected to include a number of AI-related announcements, adding to a fierce contest among companies at the frontier of the technology to build models that can interpolate between text, images, audio and code, and complete certain tasks autonomously.
Murati said her team have “no idea what Google is doing”, and that the event’s timing was coincidental. The updates announced on Monday would “shift the paradigm” of interaction between humans and machines, she added.
The company will offer its new model for free. “This is a very concrete way for us to advance our mission to provide these benefits to everyone. That’s not trivial,” said Murati.
OpenAI has set the pace in the race to build super-powerful AI systems since the launch of its ChatGPT chatbot in November 2022, and its early dominance in the space has pushed its valuation up to more than $80bn.
Sam Altman, OpenAI’s co-founder and chief executive, has said the company’s next AI model would be “materially better” than GPT-4. But OpenAI has sought to temper expectations that the launch of the new model — which was widely expected to be released in the first half of 2024 — was imminent.
OpenAI was likely to share an update on GPT-5 at some point in 2024, according to Murati, although she did not specify when.
In the meantime, start-ups including Anthropic and Mistral, as well as big tech companies Google and Meta, have narrowed OpenAI’s early lead, developing AI tools that can complete complex tasks and generate lines of code, text or images.
There is also OpenAI’s closest partner, Microsoft. As well as committing $13bn to OpenAI and supplying the start-up with computing power and access to chips, Microsoft has struck deals with rival start-ups including Inflection and Mistral, and is working on its own AI models which would compete with OpenAI’s technology.
Competition between the companies has been fuelled by a supply of increasingly powerful semiconductors, particularly Nvidia’s Graphics Processing Units. That has given the chip company a central role in the push to advance AI.
Murati closed her presentation on Monday by thanking Nvidia chief Jensen Huang “for bringing us the most advanced GPUs possible” to run its demos.
Additional reporting by Madhumita Murgia in London
Read the full article here
