OpenAI downplays rumors of web search engine, GPT-5 • The Register

OpenAI, the maker of many chatbots and taker of much Microsoft money, denies it’s planning to unveil a web search engine on Monday.

“Despite reports, OpenAI isn’t launching a search product or GPT-5 on Monday,” a spokesperson told The Register in the past few hours regarding a product update coming early next week. CEO Sam Altman echoed that denial in a social media post.

For Microsoft, which has crammed OpenAI’s ChatGPT into its Bing search engine, that’s perhaps a bit of a relief. For Reuters, which this week published a report claiming the AI super-lab plans to announce a Google Search competitor, that’s either an invitation for soul searching about sourcing or a set-up for an “I told you so” moment.

We’re going to take OpenAI at its word on this one, even if the timing – one day before Google’s annual developer conference, Google I/O – would make for a particularly pointed product debut.

OpenAI is promising only to “demo some ChatGPT and GPT-4 updates.” Still, that’s rather a bland commitment, too much so to warrant social media messaging, an “alert the press” email, and a live streaming invitation to the wider world. It’s believed OpenAI may demo a voice-controlled assistant.

Altman’s tease about the Monday reveal – “we’ve been hard at work on some new stuff we think people will love! feels like magic to me” – sounds a bit like starry-eyed Apple marketing with fewer superlatives. So it’s probably not G-spotPT, or whatever OpenAI’s NSFW model ends up being called.

Given the talk of OpenAI pitching partnerships with publishers, the AI biz may be looking to show off how it can summarize current news content in its chatbot replies, which would be search-adjacent.

And that’s probably the way to think about how machine learning models will fit into the search ecosystem – they will complement it rather than replace it.

Keyword search works well for document retrieval and has become quite effective since the web emerged in the 1990s. Better still, it allows searchers to understand where documents came from and who created them.

AI models haven’t been designed to do that. They make predictions based on inputs but their results are abstractions of source material. And while there’s ongoing work to address that shortcoming, there are plenty of situations where authorship, trust, and accountability really matter, and a ChatGPT summary without citations won’t suffice.

More importantly, search relies on ad revenue from visitor traffic. When AI models provide answers within a particular company’s interface rather than directing people to other websites, the monetization options aren’t the same and the wealth may not be shared.

AI happens to be an expensive habit and it looks likely to become more so as data owners circle the wagons to prevent permissionless training and strike licensing deals. At the moment, OpenAI, Google, Microsoft, and others rely on subscription revenue and partnership deals. But there’s a limit to the number of subscriptions that organizations and individuals will tolerate. What’s needed is a system as inclusive as web advertising, flawed though that model may be.

If AI search aspires to rival web search, it needs to find a business model that isn’t just repackaging other people’s uncompensated content as summary blurbs, code snippets, or visual remixes, and selling the results with a markup.

OpenAI and its peers can’t expect that everyone creating digital content will want to have their work included in an AI model that enriches model makers but not anyone else. And those whose work has already been incorporated into existing models may have something to say on the matter too, if the law allows it.

We note the irony of OpenAI seemingly sending a copyright infringement takedown demand to the ChatGPT subreddit for using the lab’s logo. The moderators of that board have since said OpenAI has allowed that message board to use the branding after all. ®

Leave a Comment

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

Scroll to Top