Siri co-founder talks to iMore about the future of technology, mobile interfaces, and implementing Siri

Dag Kittlaus, creator of Siri, recently spoke at an economic development event in his hometown of Michigan City, Indiana, and iMore had the chance to listen in and ask him some questions about the challenges they faced in implementing Siri and where he thinks the future of technology is headed. According to Kittlaus, interfaces such as Siri and talking to machines is something we can expect to see a lot more of in the very near future.
In 2007, after becoming frustrated with some management decisions at Motorola, Kittlaus connected with the Stanford Research Institute in Menlo Park, CA. He was hired to come look at their technology and figure out what possible products could be made to be commercialized. This is where he met Adam Cheyer and Tom Gruber, who co-founded Siri. Their first idea actually involved studying twins and DNA. Kittlaus joked about an idea he had to start a company specializing in DNA dating and molecular best matches, but didn't think people were ready for that. Yet...
Kittlaus also talked about what mobile phones were before the iPhone, and how interfaces were rapidly changing.
Mobile phones were just really hard, especially back before the iPhone came out. Phones were tough. It took about 30 clicks to find a ringtone and download it. But what if you could talk to a phone? What if you could just type in a few key words and it understood what you were trying to do and just did it for you? That's pretty powerful. The speech part of it, we didn't even start that part until about a year later. It just didn't work that well. About a year later, speech recognition got really good.
Kittlaus discussed what goes on when you interact with Siri.
First, when you speak it turns the sounds into words. It says "this is what you said". That's not what Siri does. We work with a 3rd party that does that. But what the world didn't have at this point was a machine that could understand what the words meant and do something about it. So that's what Siri does. Even after all the research and development, it took us three years to build it.
The next challenge they faced was getting Siri to interact with humans and getting people to understand what it was. It wasn't Google or a search engine. They didn't want it to take input and simply spit out 10 links. It needed to understand humans and be more human-like. Kittlaus joked about coming up with a tag line:
What are we going to call this thing and make people understand what it is? The first tag line I came up with was "Siri, practically human." Well, it's not really human yet and kind of sucks at this point. We need something better. So this version we'll call "Siri, periodically human." Then we said ok, at a certain point, the next version will be practically human. The version after that will be positively human. Then unfortunately, the version after that will have to be kill all humans.
Siri wasn't an overnight success nor an easy project to implement. During the talk, I had the opportunity to ask Kittlaus about the challenges they faced in implementing Siri.
The first time was -- How do you make all this technology work together and actually deliver this experience of talking to a machine? That's really hard. That's what we spent the first two years doing. After we got acquired by Apple, now we had to take this technology and make it available for 100 million users. That's a completely different type of challenge. Nobody had ever done it before. So the biggest challenge was getting it to work. Especially in the software business, you really need the best people. It's different than many other businesses because the impact of getting the right people means so much more than it does in many other industries. In sales, you can have two sales guys that do as many sales as one guy that's really good. But in software, it can be 50 times or 100 times different because if you don't have a good team, you're going to be writing bad software that all the good guys have to go and clean up later which really slows you down. You can't throw people at a software problem. You really need the best people.
Kittlaus went on to talk about the "law of accelerated returns" which basically states that each generation of a product exponentially doubles in power over the previous technology. Kittlaus pointed out that the iPhone has 10,000 times more computer power than the original Apollo program. If you apply the law of accelerated returns to the rate of technological growth today, we'll advance 20,000 years in the 21st century.
If I take 30 steps, I'm over the table. If I take exponential steps, I'm on the moon.
Towards the end of the session a member of the audience asked what the next version of Siri will be like. If I take 30 steps, I'm over the table. If I take exponential steps, I'm on the move. jokingly stated that he couldn't answer that question in fear of the attorneys at Apple, but was able to say that this is just the beginning and we've got a lot to look forward to.
A special thanks to Michigan City Economic Development for hosting such a great event and letting me take part in it!
iMore Newsletter
Get the best of iMore in your inbox, every day!
iMore senior editor from 2011 to 2015.
-
I only have limited use for Siri at this point, and "she" frustrates me as much as helps me; so I'm really looking forward to the things he says are in store for us!
-
Re: "Mobile phones were just really hard, especially back before the iPhone came out."
Exactly. Back in the day, being a "geek" meant you were smart enough (or obsessive - compulsive enough) to deal with complex interfaces and make them do what you wanted. Now, the definition of "geek" has been expanded to include people who say "I dig technology and stuff."
Also, back in the day, natural language recognition was one of the hardest things to do in the field of artificial intelligence. Not just simple keyword recognition, which is trivial in comparison. Real natural language recognition requires "world knowledge." And encoding world knowledge, the way humans do automatically and transparently, is extremely difficult and resource-intensive for a computer...
Anyway, great report Ally! Keep it up! -
Everybody is forgetting that when Siri was launch and put into their iPhone 4S that it was classified as "Beta."
-
I don't know, before the iPhone, Palm OS ruled smartphones. A very simple OS that I never found complicated or difficult in any way.
Obviously, he must have been talking about dumb phones, where you had to tap out stuff on the number pad. Siri could have added great functionality to those phones. On a smartphone, it's pretty rare that I find voice recognition useful. It's still not accurate enough to rely on, at least not for me. -
I love Siri so much I'm writing this response using her! For me the voice recognition piece is so good that when in the car and I need to send a text message I just speak it and I think it doesn't amazing job of translating my voice and accomplishing tasks that would otherwise be very difficult to do and pay attention to the road! I hope that Siri didn't make too many mistakes in translating what I was saying into this box but we'll see!
Manual entry: just checked the above and the voice recognition nailed it 100% again! The only thing that's awkward is saying "exclamation mark" out loud at the end if a sentence :-) -
No it didn't. Read it again. The word doesn't?
-
Yes, that would be handy, if dictating texts while driving was safe to do in the first place.
-
Had something happen today that was funny someone came into work she was lost looking for a street and Siri told her to come here. She started cursing at Siri I had to explane that the street comes thru here but you can't enter thru this gate so Siri was kind of right. This was private property it funny when she started yelling at Siri cursing her off calling her a bi$# lol