After a couple more weeks of playing around with AI (Grok) for search, it’s now hard to go back to DDG or other conventional search engine.
What a godawful waste of time those things have become. The first page of search results for most queries is now almost entirely ads. Followed by results sometimes tangentially related in some way to the query terms, but usually based on “clearly you really meant x when you asked about y” substitutions that probably make someone money somewhere. None of it makes sense, and unlike the old electronic Reader’s Guide to Periodical Literature, sophisticated queries (booleans, requireds, etc.) don’t work reliably or at all anymore.
How the search engines can have steadily reduced the utility of their product over the past fifteen years baffles me, but they’ve all done it, so there must somehow be some financial benefit to doing it (ad revenues from skewing search results, etc.). But Amazon’s on-site search engine does exactly the same thing (perverting even simple queries so that your search always returns umpteen pages of unrelated products from unpronounceable Chinese Alibaba resellers).
Grok is not just more responsive, it’s more efficient, because it’s not returning websites but actual content. And not just content, but digested content, summarized to the degree you request. Along with options to expand on your queries, while using the prior queries and responses to inform suggestions for further exploration and interpretation.
I’m coming to enjoy our new robot overlords.
There is an issue I’ve noticed, however. Grok makes errors. Spectacular errors in some cases. While researching a suitable location for a story on Mars, it argued with me that the coordinates I had picked from a map were not in a crater, only to finally relent and admit that it was wrong, the location was in fact in a crater. Some searches went hilariously sideways until I got it to admit it was creatively embellishing things, at which point I learned to include “vfo” (verified facts only) in certain queries prone to these flights of fancy.
The funniest experiences so far have been with music video searches. “Careless Whisper” came up on a YouTube mix while I was working. Out of curiosity, I asked Grok where the video had been filmed (Miami, it turns out – it had looked like someplace actually exotic to me). It returned a list of where each scene had been filmed – only, the scenes described didn’t fit what was in the video. I asked it about some of the mystery scenes, and it provided detailed breakdowns of sequences that weren’t in the video at all. Even the scenes that were in the video were described entirely differently from how they actually appeared.
The best of these so far has come from asking it to summarize the symbolism of the video for Wardruna’s “Raido“. I was expecting something about the rider freeing himself and the horse from their bonds, growing from their subsequent journey together, finally finding independence and going their separate ways. I was not expecting a vivid description of the rider mounting the horse, journeying to a mead hall, slaughtering a horde of witches, and being carried off to Valhalla while the horse gets a Viking ship funeral. I’m sure that would have been an interesting adaptation, but it has nothing whatsoever to do with Einar Selvik and his horse trotting along a beach and through a meadow together.
I’m actually relieved to find that one can’t fully trust the AIs (at least at this point). It means that they remain an assistant for now, and not a substitute brain we can use to do all of our thinking. Unfortunately, that situation probably won’t last, but by the time they reach that point, Elon will probably have Neuralink advanced to the point where they can be spliced into your skull-meat…