AI is About Access, Not Interface

And how IFTTT is solving this

Linden Tibbets
Published in
5 min readOct 26, 2016

--

In the Understanding AI series we’re exploring how AI is transforming the way we work and live — from sales and marketing to productivity and art.

For many people, the first name that comes to mind when you say “Artificial intelligence” still isn’t Siri, Alexa or Google. It’s Hal 9000, the Terminator, or another character from a dystopian, not-so-distant future. They all talk like us, or look like us. We just can’t resist defining AI, an abstract and potentially revolutionary concept, in terms of ourselves. For better or for worse, we’re incredibly self-centered.

All we can see and hear is the interface — the way we interact with AI is what makes it intelligent to us. Not what it knows, what it can do or the choices it makes. It’s this misunderstanding that causes most of us, even some of the world’s most advanced technology companies, to miss the point: AI doesn’t need more human-like interfaces to be useful. It needs access.

The power of data and abilities

By access, I mean two different things:

  1. AI needs access to data so it can learn. Wikipedia, your favorite movies, billions of conversations in Spanish…etc.
  2. AI needs access to abilities in order to turn that learning into useful action, from unlocking your front door to creating a calendar invite on your behalf.

The technology behind self-driving cars is just AI with the right access: millions of hours of training data and real-time sensors coupled with the ability to completely control the car from the brakes to the air conditioning.

Some other great examples

To become a Jeopardy champ, Watson needed access to four terabytes of data, and the ability to hit that buzzer wand thing.

With the ability to manipulate the game controls and a bit of feedback on the success of those button presses, MarI/O can destroy your Super Mario high score.

AI can even be an artist: with access to Picasso’s images, Kubrick’s films, and the ability to record new videos, it can remix 2001: A Space Odyssey.

There are many other examples where access to the right combination of data and abilities unlocks AI’s potential. So why is our consumer technology lagging so far behind?

Your digital assistant isn’t personal yet

With the initial success of Alexa, AI-powered personal assistants from the world’s leading technology companies seem to be climbing the far side of the uncanny valley. But while interfaces are becoming more human-like in their mannerisms, access and usefulness hasn’t followed suite. These personal assistants are far from personal. They promise something that sounds like a mind-reading butler, but in practice is more like…Google search with your voice. They can tell us the height of the Empire State building in inches — but they can’t do our laundry, order our favorite food, or book our next haircut.

They have promise, but they still only have one piece of the puzzle: interface, without access. And getting it for them isn’t going to be easy. The data and abilities that are the most personal and the most useful to us are scattered across a constellation of service silos. As more and more of what we do plugs into the internet, the number of silos grow and our accessible context becomes increasingly fractured. Large technology companies are taking the dominate portal approach, and building out an entire set of third-party integrations. Like AOL and iOS before them, they’re trying to build the biggest and best portal, and hoping developers will come. But there’s a better way to get that access. And it gives anyone, not just AI, the power to do so much more with it.

We need controls and permissions we can understand

Instead of waiting around for one AI to win, giving them all of the access, and watching Skynet boot up, we need to think about this problem in a new way. And it’s a way we’re already familiar with in the physical world. Here’s an example: You’re going out of town for a couple weeks and need someone to mow your lawn and water the plants inside your house. Mowing is hard work and your adult friends won’t do it for just a few extra bucks. For this task, you turn to your next door neighbor’s teenage son, Tim. You give him the keys to the garage so he can access your lawn mower. But you don’t give him the keys to your front door. You trust him to mow your lawn, but you also know he’s a teenager — and teenagers like to party in empty houses. Instead, you ask your co-worker Jessica to water your plants, and give her the keys to your house.

You could have given both sets of keys to either of these people, but you made a rational decision based on trust and capability. We need to be able to make the same decisions when it comes to AI and digital access, but if we allow one AI to control all of the access we remove our choice from the equation. It would be like having a giant set of keys, that only Tim can use. The data AI needs to access is our data. The abilities we need them to have impact our lives. We need to be confident and in control of how they are accessed.

Where IFTTT comes in

As the CEO of IFTTT, I talk to plenty of partners who are concerned about the future of AI. How can they get or give the access they need to be successful? I also hear from users who are wary of all of the services competing for their access — for their trust, for their data, for the keys to their homes (figuratively and literally, in some cases).

The reason I get to hear about this issue from both sides is the same reason I think we’re uniquely positioned to solve this problem: we’re building a platform that grants companies the access they need while giving users the ability to see, understand, and manage that access. Lately, I’ve found myself describing the future of IFTTT as “PayPal for access” — we may not be a sexy, human-like AI interface, but we’re working behind the scenes to facilitate the types of access that will make AI successful. We’re not breaking down the network of silos where your personal data lives. We’re building drawbridges that you can raise and lower as needed. We’re also committed to growing that network and making it easy for new companies to join.

We’re looking beyond the AI interface, to unlock its true potential. Join us, and let’s build something cool together!

In the Understanding AI series we’re exploring how AI is transforming the way we work and live — from sales and marketing to productivity.

--

--