AI: What We're Gaining, and What We're Losing

Part 1, in which I explain myself and admit to my role in helping to erode digital privacy.

I’ve been working adjacent to big data, machine learning, and AI for a long time. One of my first jobs in publishing was for Miller Freeman Publishing, which published a magazine named AI Expert. The magazine was started during the 80s, during a period of excitement in AI research that came to an end in the early 1990s and was followed by what we now call the 2nd AI winter.

I joined the company shortly after AI Expert was shut down. My first job there was in the circulation department, working for Software Development Magazine, Unix Review, and several other software, database, and networking-related magazines.

The magazines were what’s known “controlled circulation” or “qualified circulation” publications. What this means is that if you filled out a “qual card” with information about yourself and you meet the profile the magazine is looking to reach, you’ll get the magazine mailed to you for free. Controlled circulation publications are attractive to advertisers because they provide more information about the people their ad will reach than an “uncontrolled circulation” magazine that anyone can buy or subscribe to. The downside, however, is that people who get magazines for free tend not to actually read them.

At the time (1995), the web was still in its infancy, and most qual cards were done on paper. One of my responsibilities was to sort through these cards and mail them to the fulfillment center that would determine if each person met the minimal qualifications and mail out the magazines if so. The job involved a lot of stapling, and I became known as “staple monkey.”

During my short time in magazine circulation, qual cards became web forms that were filled out and processed electronically. Magazines also started having web sites where users could create accounts that would allow them to interact in message boards, search job boards, subscribe to newsletters, register for conferences, and so forth. Each of these functions of the website gathered additional data about the users, but they were in separate databases and users often had to log in separately for different functions on the same website.

After I left Miller Freeman to start consulting and building websites, my company was asked to design a system for a magazine’s website that would connect all of its data sources and provide single sign on. The idea was that single sign-on would be convenient for the user and would also allow the company to connect all the different types of data gathered about each user into a profile that could be mined for data that would appeal to advertisers.

As we were working on the project, I felt that what I was doing was potentially dangerous and that it should be illegal to create such a detailed profile of someone without their knowing. But I also knew that the sales and marketing departments wouldn’t have the ability to analyze or use any of the data. There’s a big difference between collecting data and doing anything useful with it.

Long after I’d done my part to chip away at digital privacy, companies such as Google and Microsoft developed advertising platforms that did exactly what I feared would happen: they collected massive amounts of data about how people use the web and then used that data to target advertisements to them that are specific to their desires at the moment they see the ads. As highly targeted advertising got even more sophisticated, advertisers started using machine learning to figure out things about people’s needs and desires that they may not even be aware of themselves.

I wrote in a blog post in 2009 that the future of the web would be that it would only have one button called “Do what I want” and that the internet and your computer would know what that is without you having to give any further instructions. We’re almost to that point now.

We’ve had AI systems that are superior to humans at narrow tasks (such as playing chess) for a while now, but we may now be close (or even at) the point where we have AI that can surpass humans at generalized tasks (such as writing books, editing videos, creating software, and even having a conversation).

In this series of articles, I’m going to talk about how I think the world is changing, how that might be good, how it might be bad, and what I think we’re gaining and losing as humans. In the process, I hope to sort out my own thoughts on AI, improve my knowledge about what’s happening, and figure out how and whether I want to be involved in it.