Not too long ago I mentioned that the search engines will need to add ChatGPT-like functionality in order to stay relevant, that there's already a browser extension that does this for Google, and that Google has declared code red. Right on schedule, yesterday Microsoft announced that they're adding ChatGPT to Bing. (If you're not aware, Microsoft is a 10-figure investor in OpenAI, and OpenAI has granted an exclusive license to Microsoft, but let's not get into how "open" OpenAI is).
I heard about this via this HackerNews post and someone in the comments (can't find it now) was saying that this will kill original content as we know it because traffic won't go to people's websites anymore. After all, why click through to websites, all with different UIs and trackers and ads, when the chat bot can just give you the answers you're looking for as it's already scraped all that content. To be honest, if this were the case, I'm not so sure if it's such a bad thing. Let me explain!
First of all, have you seen the first page of Google these days? It's all listicles, content marketing, and SEO hacks. I was not surprised to hear that more and more people use TikTok as a search engine. I personally add "site:reddit.com" to my searches when I'm trying to compare products for example, to try and get some kind of real human opinions, but even that might not be viable soon. You just can't easily find what you need anymore these days without wading through ads and spam.
Monetising content through ads never really seemed like the correct approach to me (and I'm not just saying that as a consistent user of extensions that block ads and skip sponsored segments in YouTube videos). It reminds me a lot of The Fable of the Dragon-Tyrant. I recommend reading it as it's a useful metaphor, and here's why it reminds me (skip the rest of this paragraph if you don't want spoilers): there's a dragon that needs to be fed humans or it would kill everyone. Entire industries spring up around the efficient feeding of the dragon. When humans finally figured out how to kill it, there was huge resistance, as among other things, "[t]he dragon-administration provided many jobs that would be lost if the dragon was slaughtered".
I feel like content creators should not have to rely on ads in the first place in order to be able to create that content. I couldn't tell you what the ideal model is, but I really prefer the Patreon kind of model, which goes back to the ancient world through art patronage. While this doesn't make as much money as ads, I feel like there will come a point where creating content and expressing yourself is so much easier/cheaper/faster than it is today, that you won't have high costs to maintain it on average (just look at TikTok). From the other side, I feel like discovery will become so smooth and accurate, that all you need to do is create something genuinely in demand and it will be discovered on its own, without trying to employ growth hacks and shouting louder than others. I think this will have the effect that attention will not be such a fiery commodity. People will create art primarily for the sake of art, and not to make money. Companies will create good products, rather than try to market worthless cruft. At least that's my ideal world.
So how does ChatGPT as a search engine affect this? I would say that this should not affect any kinds of social communication. I don't just mean social media, but also a large subset of blogs and similar. I think people will continue to want to follow other people, even the Twitter influencer that posts business tips, rather than ask ChatGPT "give me the top 5 business tips". I believe this for one important reason: search and discovery are two different things. With search, there is intent: I know what I don't know, and I'm trying to find out. With discovery, there isn't: I don't know what I don't know, but I loiter in places where things I would find interesting might appear, and stumble upon them by chance.
Then there's the big question of having a "knowledge engine" skipping the sources. Let's ignore the problem of inaccurate information for now. I would say that disseminating knowledge at the moment is an unsolved problem, even through peer-reviewed, scientific journal papers and conference proceedings (this is a whole different topic that I might write about some day, but I don't think it's a controversial view that peer-review and scientific publishing is very, very broken).
I do not believe that the inability to trace the source of a certain bit of knowledge is necessarily the problem. I also don't believe that it's necessarily impossible, but lets pretend that it is. It would be very silly I think to cite ChatGPT for some fact. I would bet that you could actually get a list of references to any argument you like ("Hey ChatGPT, give me 10 journal citations that climate change is not man-made").
I think the biggest use cases of ChatGPT will be to search for narrowly defined information ("what is the
ffmpeg command to scale a video to 16:9?") and discover information and vocabulary on topics that you know little about in order to get a broad overview of a certain landscape.
However, I don't see ChatGPT-powered search killing informative articles written by humans. I see AI-generated articles killing articles generated by humans. "Killing" in the sense that they will be very difficult to find. And hey, if ChatGPT could actually do serious research, making novel contributions to the state-of-the-art, while citing prior work, then why shouldn't that work be of equal or greater value to the human equivalent?
In the case of AI-generated garbage drowning out good human articles just by sheer quantity though, what's the solution? I think there are a number of things that would help:
Overall I think that ChatGPT as the default means of finding information is a net positive thing and may kill business models that were flawed from the start, making way for something better.
I've had this problem with normal Google before (the information cards that try to answer your questions). For a long time (even after I reported it), if you searched something like "webrtc connection limit", you would get the wrong answer. Google got this answer from a StackOverflow answer that was a complete guess as far as I could tell. Fortunately, the person who asked the question eventually marked my answer as the correct one (it already had 3x more upvotes than the wrong one) although the new answer never showed up in a Google search card as far as I can tell. ↩︎