ChatGPT itself can also remember things about users that it can use later; Sometimes it does this automatically or you can ask it to remember something. These “long-term” memories influence how you respond to conversations. Fry says Search doesn’t have this yet, a new web search will start from scratch, but it should have this capability “in the next few quarters.” Once it achieves this, OpenAI says it will allow it to deliver much more personalized results based on what it knows.

“These can be enduring memories, like ‘I’m a vegetarian,’ or they can be contextual, like ‘I’m going to New York in the next few days,'” Fry says. “If you say, ‘I’m leaving for New York in four days,’ he can remember that fact and the nuance of that point,” he adds.

OpenAI says ChatGPT leverages its partnerships with news organizations like Reuters to help improve web search. Atlantic, Le Monde, Finance TimesAxel Springer, Condé Nast and Time. But its results contain information not only from these publishers, but also from other online sources that do not actively block the search browser.

Suzan Verberne, professor of natural language processing at Leiden University who studies information retrieval, says it’s a positive development that ChatGPT can now pull information from these reputable online sources and generate responses based on them. It also allows users to ask follow-up questions.

But despite its improved ability to search the web and cross-check sources, the tool is not immune to the persistent tendency of AI language models to make things up or get things wrong. When MIT Technology Review When he tested the new search function and asked it for vacation destination ideas, ChatGPT suggested “luxury European destinations” such as Japan, Dubai, the Caribbean islands, Bali, Seychelles and Thailand. Provided as a source article fromTimes, a British newspaper listed these places and those in Europe as luxury holiday options.

“When you specifically ask about facts that are not real or about events that never happened, the engine may still try to formulate a plausible answer that is not necessarily correct,” Verberne says. There’s also a risk of misinformation leaking into ChatGPT’s online responses if the company doesn’t filter its sources well enough, he adds.