The Artificial Intelligence Company You Should Watch

The opening of WWDC starred a demo by a company named Anki, which is building AI for real-life toys and games. It’s part of a plan to launch a new generation of machine intelligence for much bigger applications. CEO Boris Sofman talks about the challenges of building intelligent things.

The Artificial Intelligence Company You Should Watch

Boris Sofman is the cofounder and CEO of Anki, a startup that had been in stealth for five years until it unveiled its first product with a bang as the headline demo at this year’s WWDC. Before founding Anki, Boris worked at iRobot and earned his sci-fi creds by getting a PhD from the Robotics Institute of Carnegie Mellon University. Here, he talks to Michael Grothaus about the future of AI.


Let’s get this one out of the way first, because it’s every developer’s dream. You were chosen to be the opening demo at this year’s WWDC. How did that feel?

It was such an exciting experience! And an honor as well. We’d been working on Anki for over five years without publicly announcing anything, so to finally be able to share with the world in such a way was simply incredible. We’d actually been working with Apple for a while already regarding selling Anki Drive in Apple stores. There really seemed to be an excitement at Apple not just around Anki Drive, but how we’re using their devices in ways no one has used them before.

What is AI exactly?


AI is the science of using data to allow systems to behave with intelligence and with a purpose. This can be purely in software (a video game character, a chess program, voice recognition), but the most exciting application for us is when you combine it with robotics to make physical things behave with a level of intelligence never before possible.

Will any computers ever have true intelligence that operates at a human level?

The core approach to any AI problem is to represent the structure and constraints of the problem in a way that is computable. This is harder for some problems than others, and as a result there are areas where AI is already superior to a human, but others where it will be a long time before AI can come close to a human’s capabilities. We’re nowhere close to a general-purpose AI that can do everything a human can, but we’re increasingly capable of addressing complex challenges and we’re knocking off problems that were previously untouchable.


And your company, Anki, is working on some of those challenges?

Yes. Anki is an artificial intelligence and robotics company that is taking these technologies out of the lab and into people’s everyday lives. With these technologies, we’re able to program physical objects to be intelligent, adapt and interact with their surroundings, and to surprise people with what is possible. Consumers have never before fully benefited from these technologies, which promise to make all aspects of our lives more fun, more useful, and more efficient. Anki Drive, which is a video game programmed for the real world, is the first step for us and demonstrates what’s possible with Anki technology.

You made some breakthroughs with Anki Drive. Tell me about them.


We solved three essential problems of AI and robotics in a way no one else has: positioning (knowing where you are and what’s around you), reasoning (using that knowledge to make intelligent decisions), and execution (making those things happen in the real world). Researchers in robotics have focused on these problems for decades, but our challenges was not only solve them, but to do so at a price point and reliability necessary for consumer products. By solving these problems using combinations of affordable hardware and very complicated algorithms and robotics technologies we are making it possible to bring the promise of AI and robotics for the first time into people’s living rooms.

By enabling physical objects to understand the real world and using that information to behave intelligently, we can treat these characters in the physical world as if they were just characters in a video game. And then we can take all of those elements that make video games so engaging and fun and literally program them onto physical characters to make an entertainment experience that has never been possible before.

But this is AI. Why work on consumer products?


We believe that in the future, AI and robotics will have a very real impact on people’s everyday lives. Until now, consumers have never been able to benefit from these technologies because they’ve been focused almost exclusively on space, defense, agriculture, and industrial applications. We saw a massive opportunity in consumer products to change the way people interact with the physical world around them. With Anki Drive, we saw an opportunity to create a new category of entertainment that brings these technologies to people in a familiar and fun way, but the same technologies that go into Anki Drive become building blocks for future products. The possibilities are endless, and ultimately, we are passionate about continually surprising people with applications of robotics and AI that they would never expect.

What hardware challenges did you encounter in the development of Anki Drive?

We have to replicate functionality that typically benefits from tens of thousands of dollars in sensors and computation with inexpensive components, mobile devices, and advanced software and algorithms. This required creative combinations of hardware and software that oftentimes uses sensors and other components in ways that they were never intended to be used. For example, we are using a camera (very similar to those found in many cell phones) in a special operating mode to look downward at the track and interpret what it sees 500 times per second in order to identify the car’s position and accurately track its motion.


The software side was the most challenging?

Most of the complexity of Anki Drive (and all our products) is pushed into the software side. Our algorithms have to deal with real-world uncertainties, constraints, and variations in environments and internal components. For example, each car in Anki Drive runs logic 500 times per second to sense its position, measure its motion, and adjust its motor commands. Meanwhile, our game designers and developers have to think about not just what makes a fun gaming experience, but how to optimize the rules, UI, and character behavior for the unique form of the game.

The challenges of such a deep tie between the hardware and software requires sophisticated algorithms at all levels of the Anki Drive system.

How could technology such as yours affect not only products but software development in the future?


We believe that software will increasingly impact physical objects and behaviors rather than just be focused on Internet and traditional software products. This brings with it unique challenges at the interfaces with hardware, wireless communications, and dealing with the difficult nature of real-world operation: understanding and reacting to uncertainty, physics, changing behavior of components, etc. We are building Anki’s core technology in a way that is modular and scaleable so that programming the physical world edges closer and closer to how we program virtual worlds.

You chose to develop Anki Drive on iOS? Why not Android?

There is a very clear overlap between the vision we have with AI and robotics and the role mobile devices play in that; iOS is a great platform to build on. In time, we’ll look at other platforms, but right now we want to make sure we have a great product right out of the gate.


In Anki Drive, the iPhone acts as the controller and “brain.” Would a product like Anki Drive have been possible without smartphones?

Yes, but it wouldn’t be nearly as compelling. When we started working on Anki in 2008, there was no concept of an app. As smartphones continued their incredible growth, however, it became obvious that this was a huge advantage for us. Mobile devices will continue to play a large role in our future products.

You say the “possibilities are endless” for AI. What will AI look like in five year’s time for consumers? How will AI make our lives better?


The most exciting part of robotics and AI is that the realm of what is feasible is constantly changing. Each year, we are surprised by the rate of improvement in the sensing and computation capabilities available at affordable prices, as well as the rapidly improving capabilities of mobile devices. Likewise, advances in algorithms and AI techniques are increasingly able to take advantage of these tools.

We see countless possibilities for these technologies, starting with entertainment but progressing to many other aspects of our daily lives. They will touch on everything from entertainment to our interactions with our homes to transportation, and over time these technologies and their applications will become more common and accessible. Almost anything we interact with in the physical world has the potential to act with autonomy and purpose, and the challenge is in identifying the truly high-impact opportunities at the right times.

Interested in Anki Drive? You can download the free app now to learn more about it. The Anki Drive system will be released in the fall.

[Image: Flickr user Martin Howard]