Following Mark Zuckerberg the last few days testifying here in D.C. has been entertaining to say the least. A lot of it is just political theater, but there have been moments that make you think this is all the plot of some bad, bizarre science fiction novel.
Take for example, these three situations where Zuckerberg struggles to give any sort of a direct answer (which generally means “I know the answer but I don’t want to tell you”):
- When a user deletes their account, does the account actually get deleted and completely wiped off the server? Twitter users described Zuckerberg’s response as similar to the way Chester Cheetah stutters in answering questions in commercials. I’m going to take that lack of a direct answer as a “no, the info does not get completely wiped off the server.”
- When a person logs off Facebook, does this mean Facebook is no longer connected to a user’s browser? Zuckerberg’s tapdancing on this was really interesting because if a program I have terminated still stays in my browser and looks around, it’s not a program. It’s a virus. And since it’s Facebook, all users have already allowed it to come through the firewall and anti-virus protection. There’s nothing to stop it. Don’t think it happens? Search for something when you’re logged off. Then notice how Facebook miraculously shows you an ad for the same thing the next time you log on.
- Does Facebook accumulate information on people who have not even signed up for a Facebook account? Zuckerberg gave an answer that is basically “yes,” saying they need to do so for various reasons. As an example, if you allow Facebook to access your contacts, they’ll build what Zuckerberg calls a “profile” of each whether they have a Facebook account or not. Click on a link to a story on Facebook and you don’t have an account? They’ll grab info on your IP address, computer, phone, etc. and eventually match it up to other information they’ve gotten. Why? Because it’s what they do: gather and sell personal information.
But the part that really made me think of Zuckerberg as more like “Dave” in 2001: A Space Odyssey, was his insistence on Artificial Intelligence saving Facebook and the world. Indeed, there is actually a headline in The Washington Post this morning that says “Zuckerberg says AI will solve Facebook’s problems.”
Which is kind of frightening.
Artificial Intelligence is nothing new. It can be looked at much like you’d perceive an actuarial table on steroids. An AI program gathers as many facts as possible about a particular subject, then with the help of an algorithm, uses all the past factors as a way to predict a present or future situation. Us old-timers call it “experience,” but AI systems don’t sleep, eat French fries, or have senior moments.
In non-subjective areas, I can see how it would be invaluable. But in subjective areas, it can have limitations. And quickly yesterday, someone by the name of Jon Stewart Mill on Twitter presented a thread (he also links to this article to back up his thoughts) showing just how wrong all of that could go:
- Here's a serious worry about AI. When AI is allowed to do its thing based on standard statistics, social justice activists see the results as being "biased," when in fact they just represent reality (as indicated by the data, at least).
- How do you "correct" this fake "bias"? Introduce real bias. Make the AI begin with a warped picture of reality or a messed up value function. Make the AI actually racist, sexist, etc. That will be really messed up. Could it happen?
- SJWs now have serious power in tech, as we know. When Big Tech brings in "ethicists" or "ethical consultants," you can bet they'll be SJWs. The programmers themselves, who may not be SJWs won't have a say on this.
- The SJWs will make the programmers design the relevant AI's so that it gets them the "non-biased" results they want. Disaster and injustice will ensue.
So what he’s saying is if you analyze all the facts and they don’t say what you want it to say, some might change the formula until it does. Is that possible? Certainly. Could the AI Zuckerberg develops possibly have a particular bias so a preferred message shows up on everyone’s timelines and ones he disagrees with don’t? I don’t think there’s any doubt about that. Could other media outlets use the same program if it were offered? Who knows?
All of which makes me sad. I was a journalist for a decade out of college and strongly believe in it. I’m not a fan of The Washington Post slowly abandoning what used to be impartial journalism in favor of more advocacy of a particularly school of thought, but I am still a 7-day-a-week subscriber. The Loudoun Times-Mirror is delivered for free into my driveway every week, but if they charged, I’d pay for that too. Society needs newspapers and journalism that investigate stories that are swept under the rug, and whether it’s online or a dead-tree product, it needs to be supported.
But if publications or products try to control the message too much, it won’t change people’s minds; it will merely make many doubt the institution. Then when an editorial product does find the answer to something totally unbelievable, it will be the worst of all worlds; the truth was revealed, but nobody believed it.
So, Mr. Zuckerberg, put that AI product on the back burner. Focus instead on regaining your company’s credibility, because there are actually a lot of us who enjoy seeing pictures of our friends and reading about their lives. Without having the details of those lives sold on the open market.
Doing that wouldn’t be artificial. It would be a sign of REAL intelligence.