Amazon's turning to artificial intelligence to get rid of store lines while YouTube is still tuning its algorithms to flag objectionsable content.
I must admit that this new government shutdown has me in a bad mood, not so much for me, but on behalf of my many federal friends. This weekend, I had dinner with one friend who told me that he would be allowed to work for a few hours on Monday, “in order to prepare for an orderly shutdown.” For him, that means figuring out a way to rescue the office fish, which will likely be accomplished by draining the tank down to just a few inches of water and then transporting them to his home. So, if you see some fish commuting around the beltway this week, they are victims of the recent government impasse too.
Another friend of mine who I saw this weekend is a federal IT contractor. He was told that he must come in during the shutdown, but that there is no guarantee that he will ever be paid for that work. He said he expects that the government will make good on what they owe, but you never know.
I doubt many feds are in the mood to read about advanced or unconventional technologies aimed at the federal government this week, so I wanted to keep things a bit on the light side. I promise that I will be back next time with reports of things like threat hunting as a service, and a new way to let computers and machines tap into machine learning to defend themselves from hackers.
For this week, everything will be on the light side, including a way to perhaps solve a problem that has been eluding humans for the past 100 years or so: long lines at store checkout counters. This comes from a potential new neighbor, Amazon, which winnowed down a list of new headquarters sites last week from over 200 to just 20, including three in the D.C. area. The three sites around the beltway include Washington, D.C., itself, Northern Virginia (in a spot somewhere between Fairfax and Loudoun counties), and Montgomery County, Maryland.
Living in Montgomery County, I was proud that it made the first cut, the only county on the list of mostly cities. But then again, the prospect of 50,000 more people flooding into an already crowded and overpriced area isn’t too thrilling. Perhaps they can get Amazon to widen I-270 first.
Traffic problems aside, Amazon has recently unveiled what it thinks is a way to completely eliminate lines at grocery store checkout counters. Calling the concept “Just Walk Out,” the new store design allows customers to do just that. You log in to the store using an app when you first arrive so it knows who you are, and more importantly, that you have arranged a way to pay. Thereafter, you just walk around and take whatever you want. Hundreds of cameras hidden in the ceiling record your every move, and take note of whatever you pick up and put in your cart or pocket. Supposedly, if you put something back, the store recognizes that too, and will take it off your potential bill. Whenever you are ready, you just walk out of the store and it will bill you for your merchandise.
Personally, I think Amazon might be over-complicating the issue a bit, piling cameras and artificial intelligence into a small store to track chewing gum purchases. It’s also limited by that technology. For example, similar products in the Amazon store must be the same size, and there are no single items like individual pieces of fruit. Foods like boxed salads or sandwiches must be the same weight and price, so the store only needs to track a box, not contents or choices. Also, right now the concept is in beta, and only available to Amazon employees. I wonder if they have tried tricking the system or hiding purchases, as will inevitably happen if it is ever opened to the public. And you could hardly charge someone with shoplifting if the store is designed to let them “just walk out,” and they happen to find a way to do it without paying.
Personally, I think the whole thing could be more easily accomplished with RFID chips, something I proposed in another column many years ago. You simply embed an RFID chip alongside a price tag, or anywhere on the packaging really. Given that RFID chips cost about a penny or two in volume, it would make it an inexpensive solution. People could pick up whatever they wanted, and then scan their credit card when leaving the store. Walking through an archway reader would confirm the purchase, recording all the RFID signals and printing out a bill, or simply sending it to an app. The store would require perhaps a single security guard to make sure nobody jumped the turnstile, plus all the usual shelf stockers armed with RFID-shooting price guns. And, unlike the Amazon store, you could have different size boxes of foods and types available, since they could all have different RFID tags, just like prices. But I suppose a more practical design like that isn’t sexy enough for Amazon.
Currently, there is only one cashier-free Amazon grocery store, in Seattle near their headquarters. But hey, if they expand into the D.C. area, then a new “just walk out” store might also come with it.
I picked on Amazon a little bit, so I should also do Google, or more specifically, their subsidiary, YouTube. The streaming video service is no stranger to controversy, but it recently got stung by Internet showman Logan Paul (I am being extremely nice calling him that) filming a dead body he and his crew found in Japan’s Suicide Forest. Whereas he could have handled it in a somber and professional manner, perhaps turning it into a teaching moment about suicide prevention, he instead went a different way, acting with all the discretion and finesse of a drunken frat party.
But here is the thing, YouTube’s censorship algorithms were totally fine with poking a body with a stick, for a channel aimed at teens and kids no less. The artificial intelligence didn’t flag it as anything bad, though the Internet community, and more specifically YouTube’s advertisers who suddenly saw their products sitting alongside that behavior, certainly did. So now YouTube is hiring humans and sidelining their AI.
Some might call this a failure of AI, or technology, but I would instead flag it as a failure to program a decent algorithm. One of my other loves outside of IT is running a site devoted to video games. We have a small but very active YouTube channel. It’s nothing amazing, but we have a lot of fun doing podcasts and let’s play videos of popular video games, and even board game parties. There is nothing too outlandish about our videos, and I even tell the other content providers to keep their language clean. Yet, every time we uploaded a new video over the past few months, it would get flagged as “inappropriate for most users” according to the AI guarding YouTube.
This included incredibly tame videos like one I did, following a request from a WWII vet, over Veteran’s Day of a cartoonish World War II bomber simulation. Go ahead, please watch that video, and tell me if you find anything even remotely offensive, because YouTube’s AI certainly did. The World War II vet, by the way, loved it.
The failure of the AI is at least creating jobs for humans, as YouTube shifts to having people watch the videos instead. This is not a bad thing, but it does have bad implications for many content providers, including me and my smaller games channel. Our partner status was unceremoniously revoked by YouTube, even though we never had any content or copyright violations in over five years of membership. Why? Because YouTube is only allowing the most popular content providers to apply for the program now, with nobody grandfathered in. And we don’t quite have the 1,000-subscriber minimum.
Humans may be accurate, but they are far slower than computers, and can’t watch nearly as many videos, so YouTube figures it must give a lot of folks the boot from their advanced programs and features, including removing monetization rights. Until that changes, might (and number of subs) equals right for YouTube videos, as unfair as that is to smaller, quality content providers.
So, there you go, a mini-smorgasbord of cool techie type things to chew on. Hopefully, I took your mind off the problems of government funding for a little while. And who knows, by the time you start reading this, or maybe by the time you finish, all the wonderful, dedicated federal workers will be back doing what they do best. Here’s hoping.
John Breeden II is an award-winning journalist and reviewer with over 20 years of experience covering technology. He is the CEO of the Tech Writers Bureau, a group that creates technological thought leadership content for organizations of all sizes. Twitter: @LabGuys