sourcegraph
March 29, 2024

For more than three months, Google executives have been eyeing projects from Microsoft and a San Francisco startup called OpenAI to capture the public’s imagination with the potential of artificial intelligence.

But on Tuesday, Google took a break from the sidelines, releasing a chatbot called Bard. Google executives said in interviews that the new AI chatbot will be available to a limited number of users in the US and UK, and will accommodate more users, countries and languages ​​over time.

The cautious launch is the company’s first public effort to address the recent chatbot craze fueled by OpenAI and Microsoft, and it’s meant to demonstrate that Google is capable of offering similar technology. But Google has taken a more cautious approach than rivals, which have faced criticism that they are promoting an unpredictable and sometimes untrustworthy technology.

Still, the release represents an important step toward avoiding a threat to Google’s most profitable business, its search engine. Many in the tech industry argue that Google has more to lose and gain from AI than any other big tech company, which could help a range of Google products become more useful, but also help other companies tap into Google’s vast Internet Search business. Chatbots can instantly give answers in full sentences without forcing people to scroll through lists of results, which search engines provide.

Google launched Bard as a standalone web page rather than an integral part of its search engine, beginning a tricky dance to adopt new artificial intelligence while retaining one of the tech industry’s most lucrative businesses.

“It’s important for Google to start getting into this space because that’s where the world is going,” said Adrian Aoun, Google’s former director of special projects. But the move to chatbots could help upend business models that rely on advertising, said Mr. Aoun, now chief executive of healthcare start-up Forward.

In late November, San Francisco startup OpenAI released ChatGPT, an online chatbot that can answer questions, write term papers and improvise on virtually any topic. Two months later, Microsoft, the company’s main investor and partner, added a similar chatbot to its Bing internet search engine, demonstrating how the technology could transform a market Google has dominated for more than two decades.

Since December, Google has been racing to ship AI products. It announced “Code Red” in response to the launch of ChatGPT, making AI a core priority for the company. And it prompted teams within the company, including researchers who specialize in AI safety, to collaborate to speed up a wave of new product approvals.

Industry experts wonder how quickly Google can develop new AI technologies, especially given the incredible pace at which OpenAI and Microsoft are releasing tools.

“We’re in a special moment,” said Chirag Dekate, an analyst at technology research firm Gartner. ChatGPT has inspired new startups, captured the public’s imagination, and spurred greater competition between Google and Microsoft, he said, adding, “Now that the market needs have changed, so has Google’s approach.” Variety.”

Last week, OpenAI tried to up the ante with a new technology called GPT-4, which would allow other businesses to build the kind of artificial intelligence that powers ChatGPT into a variety of products, including business software and e-commerce sites.

Google has been testing Bard’s underlying technology since 2015, but so far hasn’t released it beyond a small group of early testers because, like the chatbots offered by OpenAI and Microsoft, it doesn’t always work. Produce credible information and may exhibit bias against women and people of color.

“We’re very aware of the issues; we need to bring this to market responsibly,” said Eli Collins, Google’s vice president of research. “At the same time, we’re seeing excitement across the industry and everyone using generative AI”

Mr. Collins and Sissie Hsiao, Google’s vice president of product, said in interviews that the company has yet to determine how it will make money from Bard.

Google announced last week that artificial intelligence will be used in its productivity applications, such as Docs and Sheets, which businesses pay to use. The underlying technology will also be sold to companies and software developers looking to build their own chatbots or power new applications.

“The technology is still in its early stages,” Ms Xiao said. “We’re exploring how these experiences manifest across different products.”

The recent announcement is the start of Google’s plans to roll out more than 20 AI-powered products and features, including a feature called “Shopping and Trying on,” as well as creating custom backgrounds for YouTube videos and Pixel phones, according to The New York Times. image capabilities.

Instead of being integrated with its search engine, Bard is a standalone web page with a question box. At the bottom of the answer there is a “Google it” button, which takes the user to a new tab page containing the traditional Google search results page on the topic.

Google executives are pitching Bard as a creative tool, designed to draft emails and poems and offer guidance on how to get kids involved in new hobbies like fly fishing. Executives said the company is keen to understand how people use the technology and will further refine the chatbot based on usage and feedback. Unlike its search engine, though, Bard isn’t primarily designed to be a reliable source of information.

“We see Bard as a complement to Google Search,” Ms. Xiao said. “We want to be bold and innovative about how we use this technology to innovate and responsibly.”

Like similar chatbots, Bard is based on what’s known as a large language model, or LLM, an artificial intelligence technique that learns skills by analyzing vast amounts of data from the internet. That means chatbots often get facts wrong and sometimes make up information without warning — a phenomenon AI researchers call hallucinations. The company said it had worked hard to limit the practice, but acknowledged that its controls were not fully effective.

When executives demonstrated the chatbot on Monday, it declined to answer medical questions because doing so requires accurate and correct information. But the bot also mischaracterized the source of the answers it generated about the American Revolution.

Google posted a disclaimer below Bard’s query box, warning users of potential problems: “Bard may display inaccurate or objectionable information that does not represent the views of Google.” three response options and ask them to provide feedback on the usefulness of a particular answer.

Much like Microsoft’s Bing chatbot and similar bots from startups like You.com and Perplexity, the chatbot annotates its replies from time to time so people can see where it came from. And it dovetails with Google’s index of all sites, so it has instant access to the latest information posted on the Internet.

This might make the chatbot more accurate in some situations, but not all. Even with access to the most up-to-date online information, it can still misrepresent facts and produce misinformation.

“The LLM is tricky,” said Mr. Collins, Google’s vice president for research, referring to the technology that underpins today’s chatbots. “Bud is no exception.”



Source link

Leave a Reply

Your email address will not be published. Required fields are marked *