OpenAI recently announced its SearchGPT prototype (https://openai.com/index/searchgpt-prototype/), and the interweb was immediately abuzz. Of course, I immediately went to check it out only to find out that it's really not available. Yes, if you have a ChatGPT account, you can sign up for the waitlist, which I did. No love yet.
OpenAI does include a couple of video demos. After watching these I'm decidedly unimpressed. There's a lot of similarity between SearchGPT and Perplexity AI, which has been available for quite awhile. There are some differences, like being able to quickly access links, but that's just a small interface convenience. Don't get me wrong, I'm all for convenience, but it's hardly game changing. So, I don't see what there is to get excited about at this point.
There's also a pretty subtle problem in the demo. If you look carefully at the screenshot below, which is from the SearchGPT demo video, you'll see a problem. (It's circled in red.)
The second entry in the results is for a festival in Asheville, which is almost two hours away from Boone. Asheville is very nice, but it's not Boone. (By the way, I'm a proud Appalachian State alum, and I love Boone!) Sure, that's not a big deal, but it's kind of odd to see that big of a mistake in a rollout demo.
Comparing SearchGPT
For comparison, I put the same prompt "music festivals in boone nc" into Gemini and Perplexity. The results were not as extensive, but were fine. Gemini also included only events that were actually in the Boone area. Gemini gave some useful links as well. Perplexity gave more detailed results including details of the performers for the Jones House summer concerts, which seems like it would be useful.
Gemini's integration with Google Maps gives it an edge for some kinds of searches. My wife and I are taking a short vacation later this month. We're only going to be gone for a few days, and the town we're visiting seems to have quite a few good restaurants. I asked Gemini to recommend restaurants for breakfast, lunch, and dinner. It came back with a solid list based partially on ratings from Google Maps. So far, so good. What was really nice is that Gemini included a map showing the locations of all of the restaurants, which lets me focus on those that are near our hotel. Gemini also has a bit of an edge with its Google Docs integration. In two clicks, Gemini created a Google Doc with its results (without the map, though). You can also quickly share the results in a Gmail message. These are minor conveniences, but kind of handy. (Sorry to not show screenshots, but I want to maintain a bit of privacy.)
Minor improvements at best
From what we can see so far, the improvements offered by SearchGPT (if there are any, which is subjective) are incremental at best. It would be nice to have another search option, and there are many reasons to want to break the Google quasi-monopoly on search, but in terms of usefulness, SearchGPT is solidly "meh" in my view. But, keep in mind that this is all an initial take based on a demo, not real interaction. Which brings me to the biggest reason I'm not excited about SearchGPT.
For most of us, SearchGPT is vaporware; it doesn't really exist. I got very excited about the new voice capabilities of ChatGPT. These were announced in mid May and access is still pretty limited. I and millions of other users still don't have access. Those features aren't exactly vaporware since the do exist. But for most users, the new voice features are no better than vaporware. Google caught flack for showing off features that didn't really exist and OpenAI is edging into the same territory.
There's a bit of nuance here. I'm not saying that these products/features do not exist. They absolutely do. I'm saying that for most users they are effectively vaporware at the moment. My concern is whether these will scale. It's one thing for something to work well in a test environment. It's another thing altogether for it to work for millions of users. OpenAI is being smart here. They're testing and refining through a limited rollout prototype, which is a solid approach. But until there's large-scale availability, I'm not going to get worked up about a new paradigm for search. I will, however, keep up with developments so that I can keep you informed.
For all things AI Goes to College, including the podcast, go to https://www.aigoestocollege.com/