If AI Is So Great, Why Do I Get So Much Spanish- And Chinese-Language Spam?

Generations ago, IBM was the dominant player in tech. It was said that you couldn't be fired if you went with IBM, but within a few years of when I started, people definitely began to be fired for going with IBM, and I was never much of a fan. But when I started thinking about this post, I found this piece at IBM's website that tries to give an overview of artificial intelligence, or AI. It's as good a start as any.

I decided I needed to update my own understanding of AI, because the tech billionaires who've lined up with Trump want to influence AI policy. It seems to me that there are two main fields within AI, massive database search and machine learning. Massive database search involves things like searching huge numbers of phone calls or e-mails for keywords or other associations to identify targets. Machine learning involves teaching computers to underatand, translate, and generate natural language, produce original text or works of art, or self-drive automobiles.

Massive database search really isn't new. One of my first tech jobs in the 1970s was with a company that produced a newspaper publishing system. One of its features was keyword search of the AP feed. You could put in a word like "food", and you'd get every incoming AP story that mentioned food, which would make the food editor's work that much easier. But at the time, the company had a hush-hush contract with one of the three-letter agencies to do a much more comprehensive search, maybe of all phone calls all over the world, that mentioned two or more particular keywords in the same message.

(At one point, I believe my sister, a highly capable linguist, had a contract to review the output of such a system, but I know little beyond that.) But this was all being done in the 1970s, it's nothing new. But let's keep in mind that for generations, the three-letter agencies, as well as their Russian, Israeli, and other counterparts, have been running these searches. Find all the phone calls that have the words "infidel", "martyr", "bomb", and "airplane" in any of eight languages. Find all the e-mails that have "Trump", "schedule", "AR-15", and "scope". Sort by relevance and pass it on to the appropriate agency.

And hey, they were doing that in the 1970s, right? Think how good they must have gotten with it by now! Think of Moore's Law, the observation that the number of transistors in an integrated circuit will double every two years with minimal rise in cost. Computer power has long since gone through the roof! With these sorts of tools, we no longer are caught by surprise with sudden invasions, terrorist attacks. or assassination attempts! They're all a thing of the past!

As far as I can see, the main concerns the tech billionaires like Musk, Mark Andreesen, or David Sacks have with AI are related to this sort of massive database search. Keep in mind that in the 1970s, the three-letter agencies were already able to do mass searches of phone calls, and that was before text messages or e-mails. Computer power has grown to keep up with that, as well as bank and credit card tranactions. If you buy a gun, the agencies know it, and they can put you on a watch list at will.

They can track your movements by phone and GPS. They know when you check in and out of your hotel, and when you go in and out of your room. AI just makes these searches faater and easier. The frontiers at this point are facial recognition and DNA. At least theoretically, it's just a matter of time before we have the DNA profiles of everyone, we'll be able not just to track, but to predict, their facial features, and if you say the wrong thing in a phone call, they'll know who you are and where to round you up.

This is part of what concerns Musk and the others, although they're more worried that the government will decide who gets the contracts to do all this and who's frozen out of the market. There are also legitimate concerns about privacy, but I'm just not sure if these aren't overrated.

For instance, let's take the case of the so-far unidentified perp who assassinated United Health Care CEO Brian Thompson in Manhattan: New York police use facial recognition as they work to identify suspect in killing of UnitedHealthcare CEO. How's that working out? They've got a cell phone, bullet shells, a backpack -- maybe he fled on an electric bike (which proably has a GPS tracker). Maybe he's in Atlanta. Maybe a camera caught him in a taxi. So far, he's still on the lam. Maybe they know his name, maybe not.

For that matter, think how AI can track my computer use. It must certainly know I'm white, over 65, my income, my health status, my address, my e-mail, my interests (railroads, history, politics), I speak German pretty well, because I have German correspondents and Facebook friends, as well as small Latin and less Greek. In fact, my German Facebook friends are mostly AfD. The agencies know all about me, and I'm sure about them as well, as does the BND, as does Google, as does Meta

So why do I get so much Spanish- and Chinese-language spam, but no German? That's my puzzle. Why can't they locate the guy who shot the healthcare CEO? Why couldn't they predict the nut jobs who tried to shoot Trump?

Almost from the day I started working in tech, I thought the whole thing smelled of boondoggle. IBM struck me as a yuge boondoggle at the time. AI, at this point, strikes me as the same thing.