How ChatGPT can take Microsoft Bing into the future

Have you ever found yourself searching through endless pages of results on a search engine to find the answer to a complex question? Let’s say you want to know if a vegan diet is right for your dog. You might start your search journey by going to Google and typing “good vegan diet for dogs” into the search box and then having to understand the bunch of links that are generated. By the time you find an answer, you’ll have taken more time than you budgeted to go through articles and reports and their sources.

In the not too distant future, finding the answer to a complex question may not be such a daunting and mind-numbing process. Microsoft is said to be integrating the artificial intelligence technology behind ChatGPT into its Bing search engine in a move that could transform search as we know it. More specifically, Bing may have the potential to offer a search experience that beats Google, according to AI researchers, and potentially usurp the search giant’s decades-long dominance.

“ChatGPT is the first new technology in over a decade that has truly changed search and could, at least in principle, upend Google’s market dominance,” said Anton Korinek, an artificial intelligence researcher and professor of economics at the University of Virginia. “What the technology does is it allows consumers to interact with their computers in a much more natural and conversational way than with traditional search.”

Read more: Why ChatGPT will be everywhere in 2023

At this point, we don’t know exactly what Bing’s AI-powered search results might look like. Microsoft declined to comment for this story. However, AI researchers expect a meaningful departure from the status quo regarding how a search engine provides an answer and how users interact with it. After all, ChatGPT is not designed to browse the Internet for information (such as a search engine). Instead, the chatbot uses the information Lesson From large swaths of training data to generate a response.

“ChatGPT can respond to its users with one clear response compared to the countless links of traditional search engines. It also has capabilities far beyond traditional search engines, such as [the ability] “To create a new script, explain concepts, have a cross-talk between the user and the system, and so on,” Korinek said. People are still finding potential emerging that even the creators of ChatGPT weren’t aware the system had. ”

Microsoft announced its plans on January 23 to Invest more resources in OpenAI, the creator of ChatGPT, to the tune of $10 billion. The deal will help keep both companies at the forefront of what’s known as generative AI, a technology used in ChatGPT that can learn from copious amounts of data to create almost any content format (text, images, music, etc.) simply from a text prompt.

Search is just one of the suite of consumer-facing products in Microsoft’s stable that are likely to change beneficially for customers in the coming years. According to a report by The Information, the Seattle-based tech giant also has plans to integrate ChatGPT’s artificial intelligence technology into established products like Word, PowerPoint, and Outlook in an endeavor that could change how more than a billion people work and accomplish everyday tasks. For example, integrating it into Outlook could simply mean asking the email application to write a message on a specific topic.

“Microsoft will deploy OpenAI models across our consumer and enterprise products and introduce new classes of digital experiences built on OpenAI technology,” the company said in a press release announcing the expansion of the partnership.

Conversational search

For its part, Google and its cutting-edge subsidiary DeepMind have been working on similar systems for years. The search giant chose not to disclose it to the public, though, in part because of concerns about unethical behavior and how chat systems sometimes break social norms. For example, in 2016 Microsoft has created a chatbot called Tay that has been forced offline after he made a hate speech. Even ChatGPT, which has rules for creating positive and friendly content, can be manipulated to produce spam responses with the right prompts.

However, DeepMind CEO Demis Hassabis said in an interview with time This month his company is examining its own chatbot, called birdfor a “private beta” sometime in 2023. Google also plans to show off a version of its search engine with chatbot features early this year, according to New York Times report Citing sources familiar with the matter.

It’s no secret that Google search has generally become more conversational over the years. The company has made progress in this area with Google Assistant and with Knowledge boards In search, and for years, has been the talk of the future of search, showcasing LaMDA and MUM AI systems at the 2021 I/O Developers Conference.

Leveraging OpenAI’s artificial intelligence seems to be Microsoft’s way of trying to outsmart Google at its own game. In the wake of ChatGPT’s release, Google management issued a “red token,” according to New York times. The report said internal teams have been reset to start working on AI between now and the company’s expected conference in May.

Today, however, Google’s search engine remains the undisputed market leader as it has been for decades, commanding 84% of the global search market share, compared to Bing’s 9% (although it has grown in recent years) in 2022. , according to Statista.

Google did not immediately respond to a request for comment for this story.

Read more: New Microsoft tools use artificial intelligence to create any image you can imagine

How smart is ChatGPT?

By now you may have heard, ChatGPT is a complex chatbot that went viral after its consumer release in late November as a free online tool that anyone with an internet connection can access. The AI-powered chatbot has made headlines thanks in part to its ability to pull out delightful poetry, create meal plans and provide authoritative answers to complex questions within seconds after you’re prompted. The technology behind it isn’t entirely new, but no chatbot to date has been able to capture mainstream fascination the way ChatGPT has. This is largely because OpenAI has built a delightful user experience around the GPT-3.5 language model, and this is the phenomenon we know as ChatGPT.

GPT-3.5 is an improved version of GPT-3, which debuted in 2020 and learned from vast swathes of data and code to help it achieve its capabilities. According to researchers at Stanford University, GPT-3 is trained on 570 gigabytes of text and contains 175 billion parameters. (Meanwhile, it’s 45 terabytes of text data, said Dale Markowitz of Google.)Including almost all public websites. For comparison, its predecessor, GPT-2, was more than 100 times smaller, at 1.5 billion parameters.

This increase in size drastically changes the behavior of the model — GPT-3 can perform tasks that have not been explicitly trained, such as translating sentences from English into French, with few or no training examples. This behavior was mostly absent in GPT-2″ researchers from Stanfordthe Institute for Human-Centered Artificial Intelligence wrote in a 2021 publication.

“It is likely that the current version of ChatGPT already knows more about the world than any human individual, and is able to present this knowledge in an easily digestible form,” Korinek said.

Despite all the promises that ChatGPT holds, there are almost as many limitations. ChatGPT critics say it’s not always clear where the chatbot is pulling information from, making it difficult for people to trust the results. Skeptics also point out that ChatGPT will always be undermined by the incomplete nature of the data it was trained on, including biased or misleading information.

OpenAI has acknowledged the chatbot’s weaknesses in its current form. CEO Sam Altman said in a Twitter post in December that the product suffers from “strength and credibility” and that it would be “a mistake to rely on it for anything important right now.”

ChatGPT is incredibly limited, but good enough at some things to create a misleading impression of greatness.

It would be wrong to rely on her for anything important right now. It is a preview of progress. We have a lot of work to do on robustness and reliability.

– Sam Altman (@sama) December 11, 2022

But don’t look for the AI ​​bandwagon to slow it down.

“There will be a number of new systems such as ChatGPT that will enter the market in 2023, and the main reflection of the resulting competition is that consumers will have more choice and, hopefully, better products for consumers,” Korinek added.

GPT-4, which is under development, is You mentioned that it has 100 trillion parameters. But it’s not expected to be released unless OpenAI is “confident we can.” [release] Altman said in an interview strict VC In early January.

Altman also tried to manage the expectations of that fourth iteration of GPT, the complex language model that supports ChatGPT, saying “we don’t have AGI”. AGI stands for artificial general intelligence, or a technology that has its own emerging intelligence rather than relying on the deep learning models currently used by OpenAI. It’s the kind of wit that’s been epitomized in science fiction for over a century and made famous in recent years by the award-winning dystopian show Westworld.

“I suspect [AGI] is kind of what is expected of us,” Altman said in the same interview, adding that GPT-4 will “disappoint” people who hold out that hope.

Editors’ note: CNET uses an artificial intelligence engine to create some personal finance explanations that are edited and verified by our editors. For more see This post.