By Julia Love and Davey Alba
Google has unveiled a series of updates to its artificial intelligence offerings for cloud computing customers, emphasizing that the technology is safe and ready for use in the enterprise domain, despite recent stumbles in consumer-facing tools.
On Tuesday, at the company’s annual cloud computing conference in Las Vegas, Cloud CEO Thomas Kurian showed how Google’s most powerful AI model, Gemini, can be used to create ads, fend off cybersecurity threats and broadcast short videos and podcasts.
Enterprise customers will be able to match responses to Gemini queries with trusted sources of information, known as grounding. The company is deploying the use of Google search results as a source of AI model answers, providing greater accuracy and freshness, Kurian said.
“Companies have piloted a number of scenarios with us with generative AI; now they are deploying them in production,” Kurian said in an interview with Bloomberg before the announcements. “The capabilities to do things like grounding, improving response accuracy – all of that, step by step, people have become familiar with, they see the value and they deploy accordingly.”
Google, a unit of Alphabet Inc., lags Amazon.com Inc. and Microsoft Corp. in cloud computing, but the market represents one of the tech giant’s best growth bets as its core search advertising business matures. Google has announced its cloud unit’s first full year of profitability in 2023 and hopes to use its AI prowess to narrow the gap with rivals. After OpenAI’s ChatGPT debuted in late 2022 and was quickly adopted by students and the general public, Google and its cloud competitors see 2024 as the year the technology takes over the enterprise world.
The race between technological powers is on. Google’s main rival in artificial intelligence, Microsoft-backed startup OpenAI, is also courting enterprise customers. OpenAI now has more than 600,000 people signed up to use ChatGPT Enterprise, up from about 150,000 in January, Chief Operating Officer Brad Lightcap said last week.
Google’s pro-business efforts follow embarrassing setbacks in the consumer market. In February, Gemini, its flagship artificial intelligence product, which ingests huge volumes of digital media to train software that predicts and generates content in response to a prompt or query, was heavily criticized after coughing up images historically inaccurate. CEO Sundar Pichai called the responses “totally unacceptable” and the Mountain View, Calif.-based company has stopped accepting prompts from people in its image generator while it works to address the concerns .
Yet Kurian presented generative AI in the corporate world as a very different story. Businesses can use Gemini to create images for ad campaigns, but the professional tool comes with 19 different controls to help marketers ensure content is consistent with their brand, Kurian said.
Despite the fallout from Gemini images, Google Cloud continued to allow enterprise customers to generate images of people using the enterprise version of the tool — and no customers complained about the results, Kurian said.
“We have had no issues with any consumer reported issues with Gemini for Workspace,” he said. “No single customer was affected, as our enterprise platform has the enterprise capability to control various elements of factuality, security, model safety and accountability.”
These controls will now be strengthened by the ability for enterprise customers to anchor Gemini responses in Google Search. When this feature is enabled, the AI model will produce citations for each phrase in its results, based on its retrieval of information from Google search results. In a demonstration with Bloomberg on Friday, hours after an earthquake struck New Jersey, a Google employee showed how the default version of the model indicated there was no earthquake. recent land in the region; the version of the model based on Google search results correctly gave the magnitude of the shaking and indicated that no major damage was reported.
Business clients can also anchor the model’s answers in their company’s data, or even in a specific part of an employee handbook, unlike the consumer version of Gemini, which is more of a universal tool.
Google Cloud’s application development platform, called Vertex AI, adds new features backed by Gemini 1.5 Pro, which Google says has the “longest popup” of any large-scale AI model . Gemini 1.5 Pro can process up to 1 million “tokens” — essentially words or chunks of words — at a time, according to the company, including audio. This means developers can ask the AI model for answers based on hundreds or even thousands of images, videos, documents and audio files.
In a demo for Bloomberg, a Google Vertex AI product manager showed how Gemini 1.5 Pro works with Google Workspace. Cloud customers can upload marketing images and other media to Google Drive and ask the AI model to create new content such as a slideshow or podcast based on a brand’s style. Users can also ask the AI model for “live images,” a four-second animated image showing a particular product in a scene. For example, Nenshad Bardoliwalla, Vertex AI product manager, generated an image of a yellow camping tent against the backdrop of a gently babbling stream. Google said images generated by its Vertex AI platform would include digital watermarks to indicate they were generated by AI.
While last year Google touted how its AI tools could be used to accomplish everyday business tasks like writing emails and marketing copy, this year the company expanded its capabilities to include more behind-the-scenes work.
The company has also rolled out a series of Gemini applications for cybersecurity, which Google says will help customers analyze threats and remediate potential vulnerabilities. These features build on Google’s acquisition of cybersecurity company Mandiant for $5.4 billion in 2022. Google’s AI-based security features can help businesses be more proactive in fighting bad actors, said Eric Doerr, vice president of cloud security engineering. “What would otherwise be very manual research tasks” can be facilitated by AI, he said.
Google has been keen to highlight how it is working with the growing number of AI startups, which it sees as a key source of cloud business. Many of the most high-profile AI startups were founded by former Google employees and are sought-after customers given the enormous computing power they require.
Google Cloud has seen an increase in activity from startups using its platform to build generative AI applications and services, Kurian said. More than 60% of generative AI startups that have raised money now pay for Google’s cloud computing services, Kurian said. Of those valued at more than $1 billion – colloquially called “unicorns” – 90% are Google Cloud customers, up from 70% in August, he said. New clients include AssemblyAI, a company that creates AI models for transcription, and Writer, a startup focused on custom generative AI applications such as chatbots.
Competition for these customers is fierce, and many startups use multiple cloud providers. Amazon said more than 5,000 generative AI startups were customers of its AWS platform as of September.
Kurian said the product updates stem from close collaboration between its cloud unit and Google DeepMind, the company’s first AI lab, led by Demis Hassabis. The lab is the product of the merger of two research groups last year, a move Google made to put all of its talent to use in the intensifying AI race. Cloud engineers and research organizations work closely – in some cases side by side – to refine product direction, Kurian said.
“This collaboration extends from the top to the bottom of the organization,” Kurian said. “We have a very close working relationship in the Bay Area, in London, in Seattle with the DeepMind team.”
