Hello and welcome to Eye on AI.
For today’s big story, we’re back to ChatGPT. OpenAI this past week announced its chatbot can now browse the internet in real-time. And while the update didn’t even warrant a blog post from the company (the news was announced on LinkedIn and X/Twitter), it’s significant both for how ChatGPT functions and the future of search.
The new capability—referred to as Browse with Bing—means ChatGPT isn’t limited to data from before September 2021 when its initial training data cut off. It’s currently only available to Pro and Enterprise customers, though OpenAI says it will expand to all users “soon.”
This is actually OpenAI’s second go at rolling out live browsing for ChatGPT. The company released the capability in June before quickly pulling it, citing that it was bypassing paywalls and privacy settings. There have been other ways for GPT users to tap the current knowledge of the web, such as by using third-party extensions or Bing Chat, which is powered by GPT-4. But the wide-scale integration of live browsing into ChatGPT marks a turning point for the tool as it continues to lead the growing parade of AI chatbots and assistants.
It also pushes the conversation about the role generative AI will play in search. Crucially, ChatGPT Browse with Bing will provide links to cite where it found the information it surfaces, but there’s still reason to be skeptical about factual accuracy. Generative AI’s “hallucination” problem, wherein these tools blatantly make up information and present it as fact, continues to run wild. And while citing sources sounds like the most sure-fire way to ensure accuracy (it is how human writers back up their work, after all), we’ve already seen that generative AI tools aren’t too great at this either. Just last week, for example, we wrote about New York Times columnist Kevin Roose’s failed experiments using Google’s Bard for email, where the generative AI tool repeatedly misattributed where it got information, in addition to both misconstruing and making up information. And that was just limited to what was in Roose’s own inbox, not even the whole web.
It’s one thing to put information into ChatGPT and ask it to summarize, create bullet points, or perform other tasks with the provided information. But when it comes to pulling in outside content, it’s clear users will still need to fact-check every piece of information and its source if they want to reach any level of certainty that what ChatGPT said is true. It sounds like a lot of work. And it begs the question: Is this truly a better way to search? Or if not now, will it be in the future?
These questions also apply to Bard, which can also access the web. But interestingly enough, they’re the same questions people have been increasingly asking about a totally different type of platform, TikTok. While the idea that TikTok is better for search than Google was initially framed as a GenZ preference (and indeed a survey suggests that 51% of Gen Z women prefer the app over Google for search), all sorts of users, marketers, and even tech editors are now seriously exploring the possibility. And it’s easy to see why: Google search pages today are filled with so many ads and made-for-SEO content that getting the information you want requires navigating through a seemingly endless thicket of extraneous matter.
TikTok, however, benefits from the clearly successful short-form video format, where any query is likely to be answered directly and by a friendly face delivering it with personality, possibly even with a hit song playing in the background. And ChatGPT offers a super simple interface and conversational experience, not to mention a host of additional capabilities.
But TikTok has the same problem as ChatGPT in that you have to circle back to Google to fact-check what you just learned. TikTok creators may deliver answers in more captivating ways, and there are reputable people sharing accurate information on the platform, but there’s also tons of misinformation. And in an interesting twist, TikTok recently confirmed it’s testing surfacing Google results in its search pages, showing how the platform is indeed leaning into search (and how Google might have a future plugging its search into other platforms).
All of this is to say that while platforms like TikTok, ChatGPT, and Bard have a ways to go in the accuracy and trust departments before they can dethrone classic Google, they’re clearly pushing the bounds of search. Google still holds over 90% of the search engine market (as Microsoft CEO Satya Nadella lamented in court on Monday), but with better interfaces starting to grab users’ attention, it’s becoming possible to imagine a world where Google as we know it becomes a plug-in or reference resource, rather than the global go-to for any and all inquiries.
And with that, here’s the rest of this week’s AI news.
But first...a reminder: Fortune is hosting an online event next month called "Capturing AI Benefits: How to Balance Risk and Opportunity."
In this virtual conversation, part of Fortune Brainstorm AI, we will discuss the risks and potential harms of AI, centering the conversation around how leaders can mitigate the potential negative effects of the technology, allowing them confidently to capture the benefits. The event will take place on Oct. 5 at 11 a.m. ET. Register for the discussion here.
Sage Lazzaro
sage.lazzaro@consultant.fortune.com
sagelazzaro.com
