For all the talk about the transformative nature of the new technology called generative artificial intelligence, some of the business uses above can be much more mundane: formatting a PowerPoint slide, summarizing a call, or writing to-do lists.
Many of the first broad applications of generative AI broke into the realm of the consumer Internet, with open chat rooms and more sophisticated versions of Internet search. But this week’s announcements from Microsoft and Google about incorporating AI into the daily tools of knowledge workers and software developers show how mundane, yet highly profitable software for businesses can be the clearest money maker. .
“Looking ahead, we believe this next generation of AI will unlock a new wave of productivity growth,” Satya Nadella, Microsoft’s chief executive, said in announcing a suite of tools on Thursday. He added that the new features would “take the monotony out of our daily tasks and jobs.”
The new tools are more sober than visions of how generative AI could evolve or change Google’s search engine, used by billions of people, but they are a crucial part of Google and Microsoft’s strategies to capitalize on their investments in AI.
Microsoft has made a series of announcements outlining how it plans to bring AI to every corner of its business. has committed $13 billion in their partnership with start-up OpenAI, whose chatbot ChatGPT captured the public’s imagination when it was released at the end of november. A little over a month ago, Microsoft also integrated OpenAI models into its Bing search engine.
Thursday’s announcement strikes at the hearts of some of Microsoft’s biggest companies, in products like its suite of software that includes Word, Excel and Outlook. Office products and related cloud services produced $11.8 billion in revenue in Microsoft’s last quarter, while search and news advertising generated about $3.2 billion in sales.
Microsoft focused on integrating AI assistants, which it calls Copilots, into the software. It relies on data that business customers have already stored in company systems: chats in its Teams collaboration tool, documents stored in its cloud, and emails on its servers.
With Business Chat, a new feature for working with the tools, someone can request an update from the client and it will scan recent emails, meeting notes, and other information to generate a response.
The products are being tested by 20 commercial customers, and pricing and licensing details will be released in the coming weeks, Microsoft executive Jared Spataro said in an interview.
The wizards produce sample text, but Microsoft stressed that users should review and modify the results. When generating text, the Copilot may make mistakes or generate irrelevant information.
It can also suggest feelings or emotions. an executive presented how the Copilot in Word could write a personal speech celebrating his daughter’s high school graduation. “In summary, to say that we are proud of Tasha would be an understatement,” proposed the model.
When Mr. Spataro demonstrated how he could use the wizard to generate an email with his comments on a draft of a blog post, the AI tool generated an email saying that Mr. Spataro was “impressed” and had Made minor grammatical changes to the post. -although he had no way of knowing if he was “impressed” or if the changes were just grammatical.
“He doesn’t know anything,” Spataro said when asked about it. He said the email should be redacted, adding: “I mean, this is an example of why we call it Copilot..”
Last month, Microsoft removed some of the features of the new Bing after its chat produced inaccurate, strange, and sometimes creepy results. The new Bing had “millions of active users” in its first month, about a third of whom had never used Bing before, the company saying. The company has said that it will experiment with how to integrate ads into the results.
Microsoft has been competing with Google, which has said its chatbot, Bard, will launch in the “next few weeks” as an experimental demo. But Dan Taylor, Google’s vice president of global ads, said in an interview last month that the company had not yet discovered a way to earn money with the chatbot.
in a announcement on tuesdayGoogle outlined a similar path to profit from AI technology: by embedding it into software that companies pay for and selling the underlying AI to other organizations.
Google said it would incorporate AI into its email and word processing tools, Gmail and Docs, so it can compose emails, job descriptions and other types of documents from simple written prompts. With a few clicks, Google said, users can adjust the tone to make it more playful or professional, and have the AI crop or enlarge content. The features will first be available to what the company called trusted users.
Thomas Kurian, chief executive of Google Cloud, which sells software and services to other companies, said in a blog post that generative AI was a generational shift in technology, similar to the move from desktop to mobile. Powered by a system known as the big language model, the AI can generate text and other media when given brief prompts.
Just as software developers flocked to develop apps for the iPhone, Google expects many developers to want to build new AI apps and businesses. Mr. Kurian said the company would offer two new products, PaLM API and MakerSuite, to help his efforts.
Google also introduced the Generative AI App Builder, a tool to help businesses and governments rapidly develop their own chatbots. The company will also allow organizations to personalize AI with their own data through an existing product, Vertex AI.
Building large language models is an expensive undertaking that requires rare and specialized engineers, and purpose-built supercomputers to handle the processing demands. Most companies won’t have the resources to replicate the years of work Google, Microsoft or OpenAI have put into building these systems, so companies are racing to meet their demand.
Mr Kurian said he expected this generation of AI to have “a profound effect on all industries”.
cade metz contributed reporting.