Microsoft uses ChatGPT to speed up Bing and beat Google

Bing’s new chat-style interface is a bigger departure from the traditional search box. During the demo, Microsoft VP of Search and Devices, Yousef Mehdi, asked the chatbot to create a five-day itinerary for a trip to Mexico City and then turn what he came up with into an email he could send to his family. The bot’s response lists its sources – a series of links to travel sites – at the bottom of the long response. “We take great care to return content to its creators,” Mehdi said. “We make it easy for people to navigate to these sites.”

Microsoft has also incorporated aspects of the underlying ChatGPT technology into the company’s new Edge browser sidebar. Users can prompt the tool to summarize a long and complex financial document or compare it with another. You can ask the chatbot to turn those ideas into an email, a list, or a social media post with a specific tone, like professional or funny. In the demo, Mehdi ordered the bot to create an “enthusiastic” update to post on his LinkedIn social media profile.

ChatGPT has been generating hype since the startup launched the chatbot in November, wowing and thrilling users with its fluid, clear responses to written prompts and questions. The bot is based on GPT-3, an OpenAI algorithm trained on piles of text from the web and other sources, which uses the templates it chooses to generate its own text. Some investors and entrepreneurs have hailed this technology as a revolution that could turn virtually any industry on its head.

Some artificial intelligence experts have called for caution, warning that the technology behind ChatGPT cannot distinguish fact from fiction and is prone to “hallucinations” – composing information in a detailed and sometimes convincing way. It has also been shown that the text generation technology is capable of reproducing the obnoxious expressions found in its training data.

Sarah Bird, Microsoft’s head of responsible artificial intelligence, said today that early tests showed the tool could, for example, help someone plan an attack on a school, but now the tool can “detect and defend” against the use of a chatbot. for such a malicious request. She said testers and OpenAI technology will work together to quickly test, analyze and improve the service.

Byrd also acknowledged that Microsoft hadn’t completely solved the problem of hallucinations. “We have improved it significantly since we started, but there is still a lot to be done,” she said.

OpenAI started out as a non-profit organization focused on making AI useful, but since 2019 it has become a for-profit enterprise with significant investment from Microsoft and recently received a new commitment from the tech giant. worth about $10 billion.

Microsoft has already released a version of the text generation technology inside ChatGPT in the form of Copilot, a tool that helps developers generate code. Microsoft says experiments show that Copilot can cut the time it takes to complete a coding task by up to 40 percent.

Additional report by Will Knight.