news

July 30th article in The Atlantic, titled: The AI ​​search war has begunEvery moment, people around the world enter tens of thousands of terms into search engines, with trillions of searches per year. Billions of people browse the Internet's search engine sites. Including Google, Microsoft andOpenAIMany powerful technology companies, includingGenerative AIThey are racing to seize the opportunity to use AI to reshape the field—the AI ​​search war is already underway.

The value of an AI search bar is obvious: instead of opening and reading multiple links, wouldn’t it be better to type a question into a chatbot’s page and get an immediate answer? However, in order for this approach to work, the AI ​​model must be able to scrape relevant information from the web.ChatGPTNearly two years after its launch, users are increasingly aware that many AI products are actually built on stolen information, and technology companies are trying to build relationships with media publishers who provide this content. Now, AI may "chew and regurgitate" first, and then decide what users see based on its opaque underlying algorithm. This also means that currently media organizations show ads and sell subscriptions to many users, but users will have fewer reasons to visit media publishers' websites.

AI models don’t capture the latest information beyond the training data, which is often months or years old. When I first spoke with Shevelienko, chief business officer of AI search company Perplexity, in June, he said, “One of the key factors for our long-term success is that media publishers continue to create factual news, because without accurate original material, AI can’t answer well.”

AI companies don’t seem to view human text, photos, and videos as works of art or products of labor; instead, they view content as an extractive mine of information. “People come to Perplexity not to consume news, but to consume facts,” Shevelenko told me. Perplexity and journalists don’t compete directly—the former answers questions, the latter publish breaking news or provide opinions and ideas. But Shevelenko acknowledged that AI searches will bring less traffic to media sites than traditional search engines because users have fewer reasons to click on links—the robots are providing answers directly. As a result, the deal between AI and media is increasingly a kind of blackmail.

But the media also has at least some ability to limit the ability of AI search engines to read their sites—refusing to sign or renegotiate agreements, or even suing AI companies for copyright infringement, as The New York Times did. AI companies appear to have their own ways to get around the media’s roadblocks, but it’s an ongoing arms race with no clear winners. Whether companies like OpenAI can win the AI ​​search war may not depend entirely on their software: media partners are also a big part of the equation.

The AI ​​search wars are trying to change the way people browse the Internet, which is a super network that organizes and disseminates knowledge in today's world. But the underlying foundation has not changed: knowledge, no matter what form it exists, is still the crystallization of human experience, wisdom and thinking, not the mechanical grasping of robots. (Author: Mateo Wang, translated by Chen Xin)