I want a "Red Teaming"
Michael Schneider
«Computer, make me a Darjeeling Tea with sugar», says Captain Kathryn Janeway, of the Federation starship Voyager. The computer makes a beep, Janeway walks over to the replicator and there’s her tea, expertly prepared by the computer.
Currently, this scene from Star Trek: Voyager is something that still science fiction. But only just about. All smartphones have some kind of personal assistant built in. In Apple devices, it’s called Siri and Google’s version is called Google Now. They use the phone’s microphone to listen to you give it commands. There’s only one problem: They’re stupid. They’re very, very stupid.
This is most likely going to change. The first contender for the throne of the convenient computer in your pocket you can talk to and it just does things is called Facebook M, which has the most un-googleable moniker since that town in Wales with the really long name. It is also a privacy nightmare, but let’s not get ahead of ourselves.
David Marcus, President of Messaging Products at Facebook, describes Facebook M’s competitive edge as follows:
Unlike other AI-based services in the market, M can actually complete tasks on your behalf. It can purchase items, get gifts delivered to your loved ones, book restaurants, travel arrangements, appointments and way more.
Therein lies the problem, or rather the stupidity of the assistants. While the computer on the Federation starship Voyager is able to think for itself. It learns and computes. Neither Siri nor Google Now do this yet. They do correlate data from your YouTube history and a variety of other sources, but they’re still very slow at actually learning something.
Now, I’m not going to deny that both Siri and Google Now add convenience to our lives. When driving, these personal assistants can be lifesavers as they just listen and perform basic tasks for the user. But the assistants are not convenient. Over the course of a few weeks, I’ve tried to do a variety of things with it. Mainly, I wanted to order pizza with it. So I sat down, overcame the initial awkwardness of talking to a phone and said: «I would like to order a pizza».
Either way, I was stuck with ordering my own pizza. So I did. While I was doing that, though, I thought that maybe these things can learn. The next time I was ordering pizza, I said: «I would like to order a pizza at $pizzeria» whereas $pizzeria is my favourite pizza joint.
I have been trying to teach both Siri and Google Now this since September now. They haven’t budged from their initial positions of showing me ads or the closest pizza places. Therefore, if they do have the ability to learn, it’s still very primitive bordering on the useless or they learn in other areas unrelated to my need for pizza.
However, what the assistant do right is understand me. They listened, figured out that I was talking about Pizza and would probably would like to eat one too. Subsequently, they delivered the best results that they could. I don’t doubt that preference based search results will be implemented as soon as there’s a price structure for companies that want to advertise. The Top Spot on Google does have its price.
Other tasks the assistants have mastered admirably, even if it took some semantics:
I’ve repeatedly tried this over the course of a few months, among other things. But those were my benchmark phrases. Both assistants, Siri and Google Now, have learned nothing or very little. Oftentimes, I thought I’d be faster if I just punched in my commands by hand. Both assistants also had significant lag between realizing I was done talking, uploading my voice to the cloud, analyzing it, downloading instructions and actually executing them.
Facebook M seeks to change that. This is what it might look like:
And now, people concerned with privacy have one reaction: «Wait… what!?»
Facebook M as well as every other assistant that implements the same functionality are a privacy nightmare. They actually are already. Because:
Let’s look at the example on the screen, buying baby shoes. To be able to buy them, Facebook needs the following:
In addition to that, Facebook M needs to aggregate both your preferences in shoes as well as reviews of other users that have bought them. I have explained how data aggregation and correlation works in an earlier Labs.
While we’re drawing up a privacy nightmare, let’s stick with drawing up dystopian scenarios. Is there anything else you notice about this screen? It takes away choice. And given Facebook’s history with promoted posts where companies and users could pay for their advertisement to be seen in prominent spots on users’ timelines, this does not seem to be too good for smaller businesses. Or medium sized ones at that.
Because let’s face it. Facebook news streams are – if we apply real world analogies – prime real estates. They’re personally crafted, curated and consumed spaces. Facebook is among the first things people check in the mornings and among the last things they check before they go to bed. In a way, it’s a very intimate spot. As an advertiser, this is where you want to be. As a company selling advertising space in that intimate spot, you can basically name your price.
Facebook M goes one step further. It exists where the most private parts of this intimate spot happens. It also has a massive potential for data correlation by means of having users’ entire profiles to back up targeted advertising schemes. Again, Facebook can pretty much name its price.
In the end, a user buys shoes from a company that paid Facebook an obscene amount of money to be the one option displayed. Smaller companies, not being able to afford to pay Facebook to be the one option, aren’t even being considered by either user or Facebook.
What, simplified, follows is this:
«But it’s only Facebook», you say. By numbers alone, 1.5 billion people on the planet have a Facebook profile. Currently, there are 7.3 billion people on Earth, more to come due to initiatives such as Internet.org that is also run by Facebook. Like it or not, Facebook is a societal juggernaut, setting trends, conversation topics and making business everywhere.
I will readily admit that personal assistants on smartphones are pretty awesome. Sure, they’re still cumbersome but nobody can deny that they have massive potential. If we one day can order our things the way Captain Kathryn Janeway of the Federation starship Voyager does, it will be fantastic. Looking at current technology, this isn’t too far off. All the assistants, be they Siri or Google Now, need is a bit more smarts and the ability to remember users.
However, despite all this, we should not forget that by using these assistants, we’re making our world smaller. We eliminate one of the things that made the Internet so great: choice. Instead, we’re being fed our only options in a take it or leave it fashion. This hurts smaller businesses and – let’s be honest – makes our lives just a tiny bit more boring. Because talking to our phones will become really old really quick.
Our experts will get in contact with you!
Michael Schneider
Marisa Tschopp
Michèle Trebo
Andrea Covello
Our experts will get in contact with you!