PCR sits down with Evi’s CEO William Tunstall-Pedoe to talk about AI and the UK tech scene.
WHAT EXACTLY is Evi?
Evi is an artificial intelligence. She’s something you can talk to through your mobile phone – we have apps on iPhone and Android.
There’s a huge amount of technology there that understands language, understands the world, can answer questions and do things, so she’s a personal assistant that you can talk to through your phone and she’ll understand you and do things for you.
Is it similar to the ALICE artificial intelligence?
Well, ALICE is more of a chatbot whereas Evi genuinely understands and can answer questions.
The comparison that’s made most often is with Siri, which is Apple’s version on the iPhone. I assume you’ve seen the adverts on the TV? Well the comparison can be made but the technology is totally different and we’ve developed the product in a completely different way.
The idea is essentially that you can have access to all the information that’s online and, in due course, all the functionality that’s online just by asking an intelligent computer to do it for you. So instead of navigating to a website and navigating their interface, you simply ask in natural language for a computer to do it for you.
What was the initial inspiration for Evi?
Well initially it came out of imagining how the world would be and what the perfect experience and user interface for the world’s computer systems would be. I couldn’t believe that in ten or 15 years people would still be guessing keywords and browsing links to find things out – the most natural way to find something out is to ask a question, just like how people interact.
To solve that problem required solving a lot of other problems, which was a big investment, but Evi was the end result.
So what kind of technology did you have to develop on your own to get to Evi?
Well, one of the problems was that computers don’t understand documents. You see this in search engines; most of the world’s information is in web pages, which are written in natural language. Now, computers can’t really read and understand those documents, they can index the keywords but they can’t actually understand the knowledge that’s in the pages.
This is why when you use Google, it’ll give you a list of pages that may or may not be what you’re looking for and it’s up to you browse and identify where the right place to go is. So one of the first really big problems that we had to solve was how to represent knowledge in a way that a computer can understand and reason with.
So Evi has hundred of millions of ‘facts’ which are atomic bits of knowledge. She knows what those facts mean and she can combine them to answer questions she’s never seen before and to reason with the knowledge.
Another problem was how to take natural language and turn it in to something that can be told to the computer in a way that it can build something out of its knowledge.
And what differentiates Evi from Siri?
Well the technology is one very big differentiator. Siri still works quite statistically and basically is designed to figure out an external service to call. What Evi does is try to understand what the user is asking and to answer directly.
In terms of the software that’s being used it seems a little like what Autonomy’s come out with – the intuitive computing.
Well, Autonomy is statistical. It’s fundamentally in the Google-type, keyword indexing space rather than the understanding and reasoning space.
Another thing is that Autonomy is very much based in the enterprise space while we’re aimed at consumers. So if you look at keywords versus understanding… obviously Autonomy sell themselves as understanding but it is still just keywords and Bayesian statistics.
Okay. Well, I know you’ve already launched Evi, but how are you taking it to market?
We launched in early February. It’s been very successful and installed in well over a million devices. Feedback’s been extremely good and we’re continuing to improve the product.
Evi’s learning more every day and we’ve got a team of 25 people working to improve the product.
Just to touch on the team and the company, how did Evi get founded?
Well, I founded the business. We were originally called True Knowledge but we found having two
brands was a little confusing so we changed the name to Evi as well.
Again, the inspiration was as I said – how do you find information by just asking for it. The original mission was very much focussed on search and around 18 months ago we decided to develop our own mobile consumer product and that’s why Evi was born.
So we took all the software we’d developed and we packaged it together as Evi and launched it earlier this year.
Were there any real challenges in getting from inception to release?
A lot. As I said, what we’re doing is extremely difficult. We’re trying to make computer systems that are intelligent and that you can interact with in a natural way.
We’ve got all the core problems solved now. There were lots and lots of challenges but the product is being used, people love it, it’s on a million devices and we’re very happy.
So the real challenge was getting Evi working…
Getting her to a point where we could launch, yeah.
But now that she’s working, it’s starting to snowball, is that right?
Yes, we’ve got good reviews, a lot of downloads and we’ve got a team of people continually working to improve what she can do.
On a different note, you’re based in Cambridge. What are the benefits of being a UK-based software developer?
Cambridge is a very good place to have a high-tech business. There’s a strong high- tech community here and a very good university so there’s a lot of smart people and lots of appropriate skills.
You can get to central London in 48 minutes, which is faster than some people who live in London, so yeah it’s a great location.
What are your thoughts on Tech City?
I think it’s a good idea. Although we’re a little bit outrside of Tech City we are within commuting range of it. I think there’s a lot of value in having high-tech businesses in a cluster, you have synergies between the businesses, the entrepreneurs and the people who are there.
So what’s the end goal for Evi?
When you open up a product and allow users to interact with it using speech, then they can and will ask anything. So you instantly can’t define it in to some narrow vertical like you could if you had a limited set of buttons and interfaces.
Out goal and challenge and where the excitement comes from is in constantly improving Evi in ways that the users want. There’s lots of things she knows and can do and obviously there a lot of things that she doesn’t yet know and can’t yet do so we’re focussed on improving the impact of the product and improving her knowledge and getting her to do more things.
She’ll never be finished but she’s getting more and more useful and she knows more things every day.
Want to receive up-to-the-minute tech news straight to your inbox? Then click here to sign up for the completely free PCR Daily Digest and Newsflash email services. You can also follow PCR on Twitter and Facebook.