Is Machine Learning at Google Falling Apart? Google’s System Doesn’t Believe I‘m a Person.

By David Liedle

Thoughts here are my own, and are not related to my employer.

My hard-earned hat, given to me for completing 3 advanced code labs at Google’s booth at the Strata conference here in NYC.

I love Google! Several of my friends are Googlers (hi guys! Sorry about this…) They make great software, have great services, and their devices are some of my favorites. Lately, however, I have been having some shockingly, life-alteringly awful experiences as a user that lead me to believe the company as a whole is getting a bit ahead of itself with Machine Learning.

I’m not referring to Google Cloud’s services that allow you, the developer, to create, train, and implement Machine Learning solutions. I’m talking about my experience as a consumer, just trying to live a normal life in a Google world. At least, up until this evening when I found out Google’s automated systems think I’m an “Unknown person”.

I’m a paying customer for several of Google’s services, including Google One for our family’s storage needs, Google Play Music Family Plan for our family’s “Down by the bay, where the watermelon grow” needs (it’s a Toddler thing), Google Apps for business for my own website, and some seriously hardcore Enterprise services are being proudly and successfully implemented at my company, too. This is not a complaint about any of those wonderful services. We’re very happy with our Google Home devices, and all of our Android phones and tablets.

But there’s one aspect of Google’s services that have become so acutely painful and disruptive to my daily life that I have to call it out:

Machine Learning for consumer-facing products at Google is out of control.

Surprise notice that my whole life had suddenly changed…

I’m sure you’ve heard people rant about being banned from a service when they don’t think they should have been. This is not one of those rants. I don’t post spam. That’s really it, open and shut. There’s not even anything to defend here. The last thing I recall sharing to another group on Google+ was a video from a Google meetup where I was speaking about Google technologies AT GOOGLE’S NYC OFFICE for crying out loud, and I respectfully and carefully shared it with the groups I felt would be interested in the content, specifically related to Angular and the Dart programming language. Everything else I share is either general interest tech stuff from reputable sources, or links to content on my company’s website (a highly respected service provider used around the world by top brands). But get this: I only share these things on my own profile. I’m not dropping them on any groups or anything. You’d have to follow me or visit my profile to see these links.

What connection do I have with spam? I am one of the most proactive surfers you’ll ever meet when it comes to reporting spam, especially on Google+ where I actually care about the communities and the content people share. Did I go too far? Is there a certain number of blatant porn videos one can report on the Google Plus Nexus group that suddenly triggers you as the bad guy?

Just scroll through a wee bit of any Google+ group to see how infested they are with rampant, blatant, violent, pornographic spam. No really, go check it out. I’ll spare your eyeballs and just show you the outcry from real users here:

Post after post lamenting the porn/spam problem

Many of us dutifully report the spam posts in the groups we’re in to try to make our communities a better place. We like Google+ and want it to thrive. But when I myself get banned, I can only assume there’s a deeply broken automated system running amok.

That’s my best working theory at the moment. Somehow, somewhere there must be a pissed off little Google bot thinking “I’m tired of all these spam reports” and bringing down its automated hammer of judgement. I say it is my best working theory at the moment because I only just got banned today and discovered it this evening as I was trying to get some work done. At first I thought it would be no big deal — I’d rather have their spam system be operating so aggressively that it accidentally catches even the most pristine of users, right? I thought I’d just request a review and forget about it for a while, but no…

Think your Google+ profile doesn’t matter? So did I, until I realized that getting my profile suspended means I’m no longer a PERSON:

Angry face overlay to obfuscate my email address which isn’t that hard to guess if you need to reach me.

That’s Google Hangouts. That’s me. Or it was, until my identity was canceled. Now we have a problem, Google. Now it’s serious.

When I realized even my Gmail presence was affected…

Gmail is suggesting that I present myself professionally with my primary email account. Wouldn’t that be nice?

…I did what any rational person would do when they panic: I twote.

I said they lost their biggest fan, but really they kicked me out and in doing so took away my identity.

That’s right, the UI in Google+ warns you that you need to remove any spam posts before submitting for review. The problem? My profile doesn’t exist anymore. I can’t check it. I’m just… gone.

Won’t affect you though, right? You’re not a fan of Google+ and don’t really spend any time there, so it shouldn’t matter if you ever got falsely flagged.

Think again.

Google+ is so deeply wired into who you are on Google across the board that you would feel it as I have, whether you consider yourself to be a “user” of their social network features or not.

It’s inappropriate, it’s wrong, and I don’t believe a human reviewer at Google would have ever made such an error in judgement. I actually wondered for a moment if it was a malicious act by some other user; perhaps a jealous rival from a meetup that hates Dart, Angular, or… Geez, I dunno. I’m really stretching for an alternate scenario here.

Then suddenly, as if by force, another broken lightbulb appeared over my head and I recalled a connection between this situation and another, linked only (as far as I can tell) by one thing: Machine Learning.

Yep, Google Photos thinks my face and my wife’s face are the same person. Seriously.

Faces blurred here to avoid confusing Google any further than they already are…

That would be no problem if there were a way to tell it that the system has it wrong. Can you believe it? There’s not.

I can’t train Google Photos by tagging images of myself, my wife, and others. I can’t tell it “no, that guy with the beard is not a woman” or in any way correct this strange merge. I mean, I’m flattered that a machine thinks I look at all like my beautiful Katya, but come on. At a practical level there’s no way for me to say “This is me” (an option within Google Photos that is not available to me since there are no photos of “me”, only Katya).

No identity, no family, no friends… It’s getting lonely to be a Google customer.

Pilot error? No, I did my due diligence here. Support ticket number 0–9322000022536 if anyone from Google Photos / Google One is reading this. The best advice I could get (thanks for trying, Samantha — not your fault it works this way) was to turn off face grouping and hope the system forgot all of the hard work I put into labeling people I’ve met in New York, Pennsylvania, Massachusetts, New Jersey, Georgia, Florida, Arizona, California, Colorado, France, Germany, Hungary, Austria, Croatia, Italy, Spain, Russia, Canada, Mexico, seriously... I’m not going to remember everyone, so I had to screenshot page after page of labeled faces so these associations could be re-entered at a later time. The good news? Turning face grouping off and back on again didn’t even reset it. Leaving it off and choosing the same setting on my cell phone’s Photos app as well did result in a reset after several days, but now it just doesn’t see any faces at all.

That wouldn’t even bother me so much if it were a friend’s photos that were getting misidentified by the facial recognition system in Google Photos, but it’s not; it’s ME. My identity has just evaporated there. It’s like some kind of sci-fi nightmare.


Am I real?

I know that Machine Learning systems are still young, but with no way for me to function normally while my Google+ account is suspended and no way for me to override myself back into existence on Google Photos…

Google, please. You and I both know I’m a person. You’re a large group of exceptionally bright individuals yourself. Let’s not get ahead of ourselves, not when the user experience cost is this severe.

I’m doing my part. I’m using 2-factor authentication, a ridiculously strong password that I change frequently, reviewing authenticated services and devices on a regular basis, posting ethically, paying my bills for your services, and, well, being a known person — not an Unknown person as your system believes right now. Please do your part as well, and design your systems in such a way that humans can remain human, even when something goes wrong with the machines.



P.S. I was just about to hit ‘Publish’ on this article when I received a message from my wife asking “Who is ‘NULL’?”, because that’s what she sees instead of me on her phone now. If you work at Google please make sure the right people read this who can help me to regain my identity.