Many years ago I ran into a Silicon Valley investor who had invested in a couple companies I knew but also was an early investor in Google. He said he was investing in “real time data mining” pretty much exclusively. Naively I asked him if that was legal. He laughed and said if it wasn’t it would be made legal because there was too much money to be made.
Google famously has a long history of duping consumers and vendors into giving up data about themselves which Google then repurposes as part of its surveillance capitalism business model. There’s nothing hidden or surreptitious about Google’s methods–they do it right in your face. It’s so obvious, Purloined Letter style, that we often miss the enormity of the assault on humanity that is the Google hallmark. Easy to miss if your mind tends to filter out the debased and naively assume it’s so evil it must be illegal.
Google Books is a prime example of Google’s in-your-face grifting. As I discussed in Part I of Why is There a Google Books?, Google wraps their Google Books project in a shroud of effective altruism (“we’re building the digital library of Alexandria“). That effective altruism tends to disguise the fact that there is a pretty obvious customer for a product that can translate conversations in multiple languages and do it at scale. I refer to these customers generically as Customer X. In case you missed it, Customer X is not all that altruistic, at least not to the typical Googler.
So what is suspicious about Google Books? Coupling the Google Books translation engine with speech recognition would also assist Customer X in translating voice recordings in a variety of languages at scale including 61 dialects of Arabic. Want to bet Customer X could provide a use-case for a speech-to-text translation engine with 61 dialects of Arabic?
The way this worked in the early days of Android was that Google found another product that would allow them to trick users into giving up the data that Google needed for a vastly different purpose than the one customers were told. As Marissa Mayer told Infoworld in a 2009 interview:
One of the future elements of what’s likely to happen in search is around speech recognition.
You may have heard about our [directory assistance] 1-800-GOOG-411 service. Whether or not free-411 is a profitable business unto itself is yet to be seen [GOOG-411 was shuttered in 2010, the next year]. I myself am somewhat skeptical. The reason we really did it is because we need to build a great speech-to-text model … that we can use for all kinds of different things, including video search.
The speech recognition experts that we have say: If you want us to build a really robust speech model, we need a lot of phonemes, which is a syllable as spoken by a particular voice with a particular intonation. So we need a lot of people talking, saying things so that we can ultimately train off of that. … So 1-800-GOOG-411 is about that: Getting a bunch of different speech samples so that when you call up or we’re trying to get the voice out of video, we can do it with high accuracy.
Speech recognition is a natural outgrowth of text recognition. If you fast forward 13 years from Marissa Mayer’s disclosures (an eternity in tech), it’s easy to see that many if not all of Google’s efforts in consumer products are directed at perfecting a product that was not the one they told the world (including many judges and librarians) that they were interested in. A cynic might think of this kind of thing as consumer fraud on a truly debased scale.
Google is not alone in developing “large language models”. Common Crawl and other developers of large language models seem to be in the business of crawling and copying all the pages on the Internet (kind of like the Wayback Machine and the Internet Archive’s mass digitization effort under the guise of “emergency”) and repurposing that data into a large language model that can be uses for a variety of artificial intelligence purposes. Notably, one recent application of LLM is Meta/Facebook’s Galactica AI tool that purports to “write” scientific papers. You can read all about it in Gary Marcus’s Substack post, A Few Words About Bullshit. But just because Zuck flopped again doesn’t mean that someone else won’t develop an LLM machine with a cool sci-fi name that will make bullshit sound real which is kind of like AI versions of medical illustrations. Not so great.
Make no mistake, this started when Google tried to deflect attention away from digitizing all the world’s culture for fun and profit by calling it the digital Library of Alexandria. An odd choice because we know what happened to the real one which was controlled by various kings and was hardly “public” as we would use the term today. History teaches that the real Library of Alexandria was more like The Citadel in Game of Thrones. Which strongly suggests that the attempt at analogy is simply inapt and confirms that Google Books is merely another grift designed to distract like a game of thimblerig.
While Marissa Mayer tells us of Google’s true motives in the 2010s, it’s now all getting merged into Google’s artificial intelligence and robotics defense contracting. The total of mankind’s cultural treasure has become a mishmash and as my investor friend said, if it was ever illegal it would be made legal because there’s too much money in it.
When it comes to AI, no one is leaving it to chance. No lawsuits this time, no appeals, no class actions. No, sir.