Google Is Rolling Out More AI Features for Customers on the Cloud
Google announced a raft of new artificial intelligence-powered features for customers of its cloud-computing business, as the technology giant jostles for dominance in the burgeoning field with rivals such as Microsoft Corp. and startup OpenAI.
As Silicon Valley buzzes about so-called generative AI — software that can create images, text and video based on user prompts — Google Cloud offered a glimpse of what it’s doing to keep up in the race. In a demonstration, the company showed how cloud customers will be able to use its AI tools to create presentations and sales-training documents, take notes during meetings and draft emails to colleagues. The company also made some of its underlying AI models available to developers so they can build their own applications using Google’s technology.
Alphabet Inc.-owned Google also said Tuesday it had signed up a flurry of AI startups as customers for its cloud service, including Midjourney, which offers an image-generation system, and AI21, which specializes in technology known as large language models. Google is offering young AI-focused businesses $250,000 in free use of its cloud — which provides computing horsepower and storage — for the first year, which the company said is 2 1/2 times what it typically offers.
“We believe in having a broad, vibrant partner ecosystem for AI,” Thomas Kurian, chief executive officer of Google Cloud, said in an interview.
Alphabet CEO Sundar Pichai declared Google an AI-first company years ago, but the tech giant has been scrambling to reassert its leadership in the industry since OpenAI released its viral chatbot, called ChatGPT, in late November. As part of a “code red” decree inside the company, senior management told staffers that all of its most important products — those with more than a billion users — must incorporate generative AI within months, Bloomberg has reported. The pressure to deliver may be particularly acute for Google Cloud, which is viewed as one of the company’s best bets for growth, though it still trails Amazon.com Inc. and Microsoft in the market.
Google said it opened the new cloud tools this week to a pool of “trusted testers,” and hasn’t yet determined pricing. Kurian said during a press event that Google will also make the programs available to consumers who use the Google One subscription service, which offers online storage, though he didn’t provide further details.
Cloud-computing companies are dangling perks and services to AI startups to lure them to their platforms. Training AI models and running related services requires a large number of chips and intense computing power, making AI startups attractive customers.
As Microsoft deepens its relationship with OpenAI, whose technology is being woven into the software giant’s Bing search engine, Google Cloud is pitching itself to other upstarts as a more neutral computing platform. In addition to Google’s own technology, AI models from young companies such as Anthropic and Cohere will also be available through Google Cloud.
To woo startups, Google also offers them built-in systems for business tasks like billing, as well as access to Google’s sales team, which helps sell the companies’ programs, so they can focus on product development. “They get instant access to a sales organization, which — having built one at Google myself — is not the easiest thing to do,” Kurian said.
Among the startups that have signed on for Google Cloud is Osmo, which is working on ways to let computers understand smells to aid in tasks like disease detection and safety. Osmo was spun out of Google Brain — but the firm’s CEO said it considered other cloud providers before deciding on Google to gain access to that company’s Big Query data warehouse service and Bigtable database, which help keep track of the chemicals Osmo is working with.
“What we do almost every hour of the day is we look for new, interesting chemicals to synthesize, new ingredients to acquire. And that means scanning through, I think now over a billion ingredients that we could potentially make,” said Osmo CEO Alex Wiltschko.
Google had previously signed up Cohere, which is working on language models for large companies and startups, and Anthropic, a research lab founded by former OpenAI leaders, in which Google invested in last month.
Among new features for enterprise clients, Google showcased a Generative AI App Builder that it said would help companies build applications such as chatbots and digital assistants “in minutes or hours.” In a product called Vision AI Generation, the company demonstrated how its technology could, for instance, generate polished marketing images from text descriptions, similar to OpenAI’s popular text-to-image product Dall-E.
To mitigate concerns about data privacy, Kurian in the demonstration said Google will provide “end-to-end sovereignty” on cloud customers’ data. Companies that use Google Cloud’s new AI suite will have the ability to encrypt their data, Kurian said, allowing enterprises to meet customer compliance requirements, even in industries with strict regulatory oversight. Neither user input nor the output that the custom AI generates will be used to train Google’s underlying models, Kurian added, and Google won’t be able to access private sets of data if companies choose to use them in order to fine-tune their own AI software.
Google also unveiled a host of AI-powered features for its Workspace productivity programs. Generative AI is coming to key products including Gmail, where people will be able to use the technology to compose and summarize emails, and videoconferencing platform Google Meet, which will incorporate AI to identify key actions stemming from meetings and to send follow-up emails.
The Mountain View, California-based company has stressed the need for caution when incorporating generative AI into its flagship search engine. Large language models – AI systems that ingest enormous volumes of digital text to train software that generates text on its own when prompted – have been known to provide inaccurate information. The consequences of misleading users who are seeking sensitive information could be severe.
Yet the risks may be more manageable in the enterprise context. Customers using Google’s tools to draft, say, marketing copy can work with the AI system to fine-tune the material until it is just right, Kurian said. Companies using Google Cloud technology for customer service can choose between different levels of automation, opting for more human interaction for sensitive topics such as finance.
“We are providing these building blocks because we recognize some of the issues that people have faced when using large language models,” Kurian said.