The thing is, this doesn't matter anymore. You have very little control as Google tries to be smart. It's very hard if not impossible to find something older, obscure, things from other regions, languages, etc...
I think "Google-fu" just refers to being able to bend the search engine to your will. In the early days it was with operators and special keywords (inurl, etc) but today most of those are not as useful or actively harmful and so "Google-fu" has progressed. It's knowing which terms to drop from the error you are searching for, it's knowing how to phrase things correctly, it's knowing how to skim the results and separate the wheat from the chaff. Or at least that's what it means to me and how I use it.
I don't think they misunderstood and I think their point still stands. The irrelevance of Google's search results are becoming ever more unyielding to the user's intent. The portion of results that are SEO spam for every query is increasing. The amount of your query being dropped and ignored in your search is increasing. Google results are becoming increasingly irrelevant and is on a trajectory to a point of completely ignoring your query where the results are strictly a combination of spam and a random pick of websites.
"Google-fu" is not progressing. It's struggling to hang on by the decreasing number of threads before it's utterly ineffective.
Where we used to say “rome fall why”, you’d now write “why did the roman empire fall”. Because the AI likes that phrasing and produces better results.
Soon you’ll write a 300 word description of what exactly you’re looking for, like you would when asking a trusted expert, and Google will figure something out. The days of keyword searching are long gone.
Not to pick on your quick example, but actually testing it, these two terms [1] [2] have nearly identical results. Top result in both cases is history.com, followed by wikipedia, and the next 4-6 results are the same but in slightly different order.
I think this is actually an example of the benefit of their AI. Despite the big difference in the "style" of phrasing (simple english vs more formally naming the subject noun), both seem to map to a very similar representation in their embedding space. I've run into frustrations with this myself, but for basic questions like this it seems like the search works Pretty Good.
Consider that "Why did the Roman Empire fall?" consists of 1 word that describes the type of question being asked ("why"), 2 useless junk words ("did the"), and 3 "key words" ("Roman Empire Fall"), of which 2 should really be treated as a single word referring to a single concept/entity ("Roman Empire", for which "Rome" is a synonym in some cases).
Humans instinctively know this, so we are able to construct queries like "why rome fall".
But that "why rome fall" query, which we think of as a purely mechanical keyword search, already requires quite a bit of sophisticated processing in the search engine. The system has to recognize that "fall" is synonymous for "collapse" or "wane in power" and not synonymous for "autumn". It also has to recognize that "rome" means "the (Western) Roman Empire" and not the modern city of Rome in Italy or "the Holy Roman Empire" or the city of Rome, NY, USA. It furthermore needs to interpret "why" in such a way that it emphasizes results with "reasons" or "explanations", rather than something like a "timeline" or "summary".
Personally I find it really weird that Google is interested in pushing users more to interact with its digital librarian / AI assistant, instead of continuing to improve keyword search.
I have a few guesses as to why they are going this way:
1. It makes the user interface simpler from an engineering perspective (fewer user-facing buttons and options to implement and test).
2. There is strategic benefit to making search more of a black box. Maybe they are specifically trying to "educate" users to expect and be comfortable with such black boxes. Maybe the plan is to get people so accustomed to "AI assistant" search that they see keyword search as outdated, and thereby secure a competitive advantage for the next several years over other search engines, by having the biggest and best AI models.
3. They are trying to increase the amount of rich "natural language" user search inputs in their data. Making keyword search worse will encourage people to use queries that more closely resemble natural language. I assume that this has strategic benefit related to Guess 2 above.
My own google-fu tells me that both of those searches are likely to be pretty poor - using the word "fall" instead of "collapse" is likely to snap up quite a few weird results about autumn tourism in italy and including "empire" in the second query feels likely to get you a batch of other poor results (like, for instance, the collapse of Russia commonly known as the third roman empire).
Personally I'd suggest "collapse of rome" which does deliver you a rich embedded result specific to the fall of rome.
I agree that Google's search parsing peaked a while back though, it seems to be getting weaker and weaker and now partially relies on the fact that search term autocompletion on mobile devices will supplement it by helping present an array of options near what you might want.
That may work for common topics, but it does not appear to work for niche ones, at least in my experience.
If I want to find a particular user-run forum on some obscure bit of some hobby, "<hobby name> <forum topic>" brings it up. But if I type out "Forum for <hobbyists> discussing <topic>" I get... a random selection popular of fora where someone has mentioned <topic>, often in passing or with minimal information.
Except that this isn't objectively an improvement, even in a perfect world where AI is substantially better than it currently is.
> you’d now write “why did the roman empire fall”. Because the AI likes that phrasing and produces better results.
"the AI likes that phrasing" is exactly the problem here. How is anyone supposed to know what the AI "likes", other than painstaking trial-and-error in the unbounded and arbitrarily high-dimensional search space of human language?
Even the people who built the model probably don't know. Language models (and deep NNs in general) are extraordinarily complicated things, and there are problems with pretty much every technique that purports to provide visibility into their inner workings. There are just too many parameters and too many "information paths" in such a thing for regular people to wrap their heads around it. The ability to incorporate a high amount of complexity is a big part of why those models are so effective to begin with, but it also makes them really hard to reason about.
"AI" is currently in a weird spot where it's starting to kinda-sorta behave like an intelligent human in some limited settings, but in general is nowhere near as smart as a human. Most models still have a very shallow conceptual understanding of anything, even if they're becoming uncanny in their ability to match sophisticated patterns. It might not even be possible to teach some concepts to language models as they currently exist today, if only because there is only limited conceptual understanding available to be learned from corpora of text and images, even huge ones. Humans are still tremendously more effective than our best language models at understanding meaning and intent. Can an AI ever learn about love, regret, fear, or bliss, by reading millions of news articles and books and looking at millions of images?
Thus AI right now is in a kind of "worst of both worlds" situation, where it is complicated enough to be hard to reason about precisely, but still mostly unsophisticated and therefore highly sensitive to how inputs are crafted. Therefore it's hard to formulate inputs that provide useful outputs. It's still alpha-level technology at best, and there might be one or several conceptual innovations remaining between what we have today and something resembling general intelligence.
Consider also that "AI assistance" is complementary to keyword search, not a replacement for it. Google search AI is becoming something like a "digital librarian", a creature that can understand your queries and guide you to a starting place in the relevant literature. But much like in a real library, the digital librarian is going to be most useful as a starting point. At some point, if you already know what you're looking for, you still are going to want to search on "structured" criteria, as well as, yes, keywords embedded in text.
And finally, do you really want to type a 300-word description in order to get good search results? I was already getting good results with 3 keywords. I have already done the sophisticated pattern-matching and concept-graphing in my own brain, and now I know exactly what terms I want to look for. Why should I be forced to coach an AI on how to redo all that work for itself, instead of just letting me do a damn keyword search? Not to mention wasting my time and giving me carpal tunnel typing it all out.
You touched on the fact that the AI right now is primitive and unable to parse regular english well - I agree that it still has a ways to go in this regard but, while all of us complaining here might prefer the old google, we are the "in" crowd that actually put in the time to learn the old arbitrary rules. It isn't great to keep around arbitrary rules purely for the sake of consistency if those rules are bad. I think the fact that the search results are often unable to clearly distinguish different questions (and may return some autumn related results for "fall of rome") is a clearly bad thing - but the old format we're used to required a lot of learning and adaptation (the aforementioned "Google-fu") that shouldn't be a necessary skill for future generations.
I had a good laugh yesterday. I was setting up a password vault for my mother and had to get the login pages for various services and a few did like Etsy does - search for "Etsy Login" and you'll notice their SEO has managed to put /search?q=login above /signin in google's results. It amuses me whenever SEO is taken to such an extreme that it actually makes the results from your company less useful for people actually looking for your company.
It's the opposite now - you increasingly have to use "Google-fu" to make sure that your query is not creatively reinterpreted in ways that are virtually guaranteed to yield irrelevant results (but more of them) - e.g. substituting words with "synonyms" (which aren't), or removing the most important keyword from the query altogether.
The thing is, this doesn't matter anymore. You have very little control as Google tries to be smart. It's very hard if not impossible to find something older, obscure, things from other regions, languages, etc...