Exclusive: Google Cloud CEO on Gemini's value for businesses
Add Axios as your preferred source to
see more of our stories on Google.
Thomas Kurian. Photo: Cayce Clifford/Bloomberg via Getty Images
Google Cloud CEO Thomas Kurian says this week's launch of Gemini for businesses comes as many companies are looking to reduce their dependence on a single vendor, such as OpenAI.
Why it matters: Google, though a pioneer in the research that led to the current generative AI boom, has been seen playing catch-up to Microsoft and OpenAI.
Kurian told Axios that Google has seen a strong influx of interest — and gained new customers — since the recent OpenAI drama that saw CEO Sam Altman fired and rehired in less than a week.
- "I think customers want choice," Kurian said in an interview with Axios.
- For its part, OpenAI has said it didn't lose any customers during the turmoil.
Driving the news: Google said Wednesday that the Gemini Pro model announced last week is now available for businesses to use from within both its Google AI Studio and Vertex AI developer tools.
- A higher-end model, Gemini Ultra, is being tested with an already determined select group of customers, Kurian said.
Between the lines: Kurian said Gemini breaks new ground as a model that can deal with voice, video, text and images simultaneously, while also improving output of computer code, an area where others had been ahead.
- Google also has advantages, he said, because it builds not just models like Gemini, but also the cloud infrastructure used to train and operate the models. The result, he said, are gains in both cost and performance.
- "We're passing along those saving to customers," he said.
- OpenAI uses Microsoft's Azure to train and host GPT-4 and other models, though the two companies work closely to optimize the models for Microsoft's cloud hardware. Amazon Web Services, by contrast, touts the fact that its Bedrock works with a wide range of models from different companies.
Yes, but: While Gemini is multimodal by design, the standard version of Gemini Pro uses only text for input and output. Another version, Gemini Pro Vision, can also accept images as input, while the fully multimodal version has yet to be released.
Editor's note: This story has been corrected to reflect that Gemini Ultra (not Gemini Pro) is the higher-end model being tested.
