Unlock Editor's Digest without cost
Roula Khalaf, editor of the FT, picks her favorite stories on this weekly newsletter.
OpenAI has unveiled advances in its flagship artificial intelligence model, intensifying competition with Google and other Big Tech groups pushing for breakthroughs within the technology.
The San Francisco startup demonstrated a lot of improvements to its GPT-4 model at an event on Monday, including the power to interpret voice, video, images and code in a single interface – even though it didn’t do anything exciting expected to unveil recent model.
The update, announced as GPT-4o, “provides GPT-4 level intelligence but is far faster and improves text, image and audio capabilities,” said Mira Murati, chief technology officer, before introducing live voice translation in demonstrated in several languages. Shares of language learning tool Duolingo fell about 5 percent after the announcement.
The updates come only a day before Google's annual developer conference, which is anticipated to feature a lot of AI-related announcements, creating fierce competition between corporations on the forefront of technology developing models that may interpolate between text, images and audio and program and complete certain tasks independently.
Murati said her team had “no idea what Google was doing” and that the timing of the event was coincidental. The updates announced Monday would “change the paradigm” of interaction between humans and machines, she added.
The company is offering its recent model without cost. “This is a really tangible way for us to further our mission of providing these advantages to everyone. This just isn’t trivial,” said Murati.
OpenAI has set the pace within the race to construct super-powerful AI systems since launching its ChatGPT chatbot in November 2022, and its early dominance within the space has pushed its valuation to over $80 billion.
Sam Altman, co-founder and CEO of OpenAI, said the corporate's next AI model will probably be “significantly higher” than GPT-4. But OpenAI has sought to temper expectations that the launch of the brand new model – which was widely expected to be released in the primary half of 2024 – is imminent.
According to Murati, OpenAI is anticipated to release an update to GPT-5 sometime in 2024, although she didn’t provide an actual date.
Meanwhile, startups like Anthropic and Mistral, in addition to major tech corporations Google and Meta, have narrowed OpenAI's early lead and developed AI tools that may handle complex tasks and generate lines of code, text or images.
There can be OpenAI’s closest partner, Microsoft. In addition to committing $13 billion to OpenAI and providing the startup with computing power and access to chips, Microsoft has struck deals with rival startups comparable to Inflection and Mistral and is working by itself AI models that use the technology of OpenAI would compete.
Competition between corporations has been intensified by the provision of increasingly powerful semiconductors, particularly Nvidia's graphics processors. This has given the chip manufacturer a central role within the further development of AI.
Murati concluded her presentation on Monday by thanking Nvidia boss Jensen Huang “for providing us with probably the most advanced GPUs possible” to run his demos.