Why was it too late? They were still number one when pagerank was published, all they needed to do was put more resources into it. Google’s algorithms weren’t magic. The biggest problem was lack of vision, and willingness to bet on that vision. Investors in Google were swinging for the fences, that mentality would never have been at home at DEC.
Google’s algorithms were definitely “magic” by the standards of the time. Altavista never got much better than literal keyword search even when they introduced PageRank-style weighting. Google immediately had useful results from misspellings, questions, and related keywords, and only got better from there. Plus it was much faster; responses in 1-2 seconds versus 5-10 sometimes for AV. Maybe what you said is true, but they were deep in the hole as soon as Google launched.
I don't recall AltaVista's response time as slow (although I was impressed by Google's fast response time, which they proudly reported then). It was the page-load time, which was slow for AltaVista compared to the one of Google, which initially drove me away once that alternative was available.
Yeah, all portal sites were cluttered heavily. Google was only the logo, the search input field, and the submit button. That was one of the biggest advantages for me.
True, and Google is slowly being disrupted by the next generation of Internet search (looks like disruption happens every 20 years or so).
The age where "search" is meant to return pages will eventually end. I know it's early, but you can think of GPT-3 as the next generation of search engines. You consume the entire "Internet knowledge", and you answer questions, by merging information from multiple sources. Not just returning one existing page. That's sort of what GPT-3 does, you just don't think of it as a search engine yet.
Google felt "magic" in 2000. A search-oriented GPT-3 will feel "magic" soon.
GPT-3 will be remembered and used by no one in no time. Google is already answering questions to queries and also offering alternative queries. It's now in a hybrid state where it tries to answer your stuff but also provide links. Quite interesting and quite usable already. GPT-3 is a cute experiment, nothing more.
The lists of questions that Google returns as snippets can be seen as possible elements of a paragraph for a process that doesn't quite know what a paragraph is. Clicking on them acts as a survey response, a signal for that element of information's significance in the topic for which you searched.
> Google is slowly being disrupted by the next generation of Internet search
What is the name of the search service (or services?) that are better than Google? You make such big statements and provide zero examples.
Bing is inferior to Google; duckduckgo is even worse - the search is often unreliable (even after years of work); Tineye got hindered by GDPR and it was search via picture anyway. So what else have we got there?
I remember the time when Google came out very well. It was a strictly better product than Altavista. Once you started using Google you would never go back to Altavista - the difference was like night and day. Altavista was dead for you. There were no obstacles to switch either.
I assume that if a new product came today, that was better, even say 25% better, people would start switching via word of mouth.
Back in the day Altavista lost because their technology was inferior. It wasnt about website design (although Google's clean landing page helped), it was 100% technology: Google was strictly better. In fact at that time there were other search service too - they were even worse - and often provided nearly random results (e.g. search for William Shakespeare -> get random porn website...).
PageRank algorithm sounds easy once you know it, but it is easy once someone tells you about it. It is much more difficult to invent. Back in the day many other people worked on improving search (those were the times of catalogs and webrings) and Google were the first (?) to come up with something like that. PageRank was basically bleeding edge research sponsored by spy agencies. It only sounds easy with the benefit of hindsight.
Also not directly to you, but the opening poster basically writes that:
* they could make Altavista work - but the meddling management hidnered it (how? did they have their own PageRank equivalent? I doubt it)
* they could have made blockchain work - but the meddling management hidnered it
I see a certain Scooby doo pattern. And a senior developer/manager/architect (with 20++ years of experience under their belt) who claims that they could make blockchain work, while most people with such experience know that blockchain is an empty buzzword.
all of search isn't question/fact based. I'd love to see the real numbers, but I'd think that less than a third of all searches are suited for a factual question answer system.
Google search is still largely based on human feedback. what people link to, what terms they choose to use in relation to other terms. It will be difficult to disrupt that
does the Ad business model work with GPT-3 answer ?
Isn't the point of google to make it 'confusing' enough to the point that I will click on the ad because it's clearer ?
How GPT-3 search-oriented answer would help with that ?
Google was inferior to AltaVista for quite some time.
According to Wikipedia: "In 2000, AltaVista was used by 17.7% of Internet users while Google was only used by 7% of Internet users, according to Media Metrix."
So, two years after official founding, Google was still not ahead to AltaVista.
AltaVista even had a "visual clustering" thing that used Java (which would work great now with Javascript) that would allow you to refine your searches. I still cry that we don't have the equivalent of that 25 years later.
Then AltaVista got caught in the great DEC "hostile giveaway". Which left Google sitting in the right place with nobody to really compete against them.
One problem was that AltaVista was early--it basically predated common people using the web--and so it wasn't quite so clear how you monetized it. DEC effectively ran it as a goodwill "free service" to the internet. Even in 1999, on-line commerce wasn't very big. Remember, the big AOL/Time-Warner merger was in 2000.
The one thing that Google got right was timing. Google was in the right place when everybody switched from services like AOL to basically just accessing the web directly. And this let them put ads in the search which could be used for monetization.
(Also, an aside it's possible for it to be totally true too)
Early Google users definitely trended towards power users, and if the average early Google user made 3x as many queries as the average Altavista user, Google would have higher actual query volume.
The publishing of the PageRank algorithm was the wrong milestone; my bad. The important milestone was a less well-defined point when search engine users and competitors realized that Google was disruptive.
PageRank was the way forward but the true schlep, in the Paul Graham sense [1], was the brute force continuous hardware scaling required to keep up with the growth of the Internet. DEC needed a technical pivot first and foremost. I've seen no evidence that the AltaVista team understood the technical challenge but I could be easily convinced that bean counters and/or management stifled a promising response; absence of evidence is not evidence of absence.
My recollection from the time (I didn't work at either place but was in the general area) was that everyone assumed Google were delivering their performance by use of massive hardware resources (I heard 70K machines at the time). So perhaps AV folks just couldn't imagine getting the hardware budget to complete effectively?
It's also worthwhile I think to point out that at that time, it wasn't settled that "search engines" as we know them today (like Google) were _the_ way to use the Internet. There were alternatives such as Yahoo (curated directory), and browser-side catalogs (netscape.com home page) that were much more popular. So it is possible also that AV folks weren't exactly thinking "We have to light the afterburners to go after Google in this immensely important internet search space". They might have been thinking "odd, someone spent $xxxM on hardware to run a search engine, that'll never work out".
WRT to curated directories, in the late 90s when Google came (along with others like Northern Light), it was clear that both directories (Yahoo and Moz in particular) and the established search providers (Yahoo, AltaVista, Metacrawler, etc.) were being overwhelmed.
Yahoo (aka Inktomi) search frequently took multiple pages of to find anything of quality. Curated directories were missing quite a bit. Furthermore, if I remember correctly there was some sort of payola scandal around Moz/OpenDirectory editors.
The market was primed for a better solution, and PageRank provided it.
Inktomi was pretty good then. In fact they powered both Yahoo and Microsoft web search at the time. When they were acquired by Yahoo, they formed the basis of Yahoo search, that was well respected in the search research community for years until Yahoo started to implode.
Early Google used commodity hardware to the point of being comical. Somewhere there is a "historical" display of commodity motherboards spread out on cork trays in a rack. Maximizing density/minimizing cost.
The ability to scale in this way with commodity hardware was new and important.
When Google published PageRank in 1998, they didn't publish MapReduce, the framework required to do coordinate large distributed computations, until 2004. By then, AltaVista was dead.
If AltaVista had thrown resources at the problem using any of the then widely known techniques for scaling, they would have been constantly dead in the water.
