Iris - Siri for Android proves Apple doesn't have an edge
Written by Harry Fairhead
Friday, 21 October 2011 00:00
Iris, a Siri lookalike, was put together in just eight hours and is now available in the Android App store. It raises a lot of questions, only some of which as easy to answer.
Siri, the natural languages understanding app that answers your questions, is supposed to be the advantage that the iPhone 4S has, not only over other models of the iPhone, but every other mobile device on the planet. To show you just how important consider the fact that Apple has produced a TV commercial featuring Siri:
It is clear that Siri's brand of humour (and even the mistakes and gaffs that it makes have so far been humorous) appeals to its audience. Users are finding that they simply can't get through to the servers and a "cannot connect" error message is annoying quite a few. So Siri does seem to be popular even if it could be just novelty value - but does it really give Apple an edge?
You might think so, given the amount of time and effort that has gone in creating Siri, which was originally a DARPA project. However, when you start to look a little more carefully you start to notice the flaws in the argument. To start with, speech recognition is becoming a fairly standard commodity. For example, Google provides a speech recognition system and Android has a native text-to-speech system - so input and output are solved.
Now we come to the question answering. The point being that Siri doesn't actually do the question answering but simply decides which of a number of online services it directs the question at. Notice that it is up to the online services to figure out how to respond to the question.
For example, general questions are passed on to Wolfram Alpha which is an "intelligent" question answering search engine. You can try it out at Alpha without needing Siri to act as an intermediate. Simply type in your questions such as:
"open the pod bay doors"
and you get the answer:
"I'm sorry, Dave. I'm afraid I can't do that"
You also get sensible answers to more sensible questions and there is an API which lets you make use of it within your own app.
What about Siri's supposed sense of humour?
If you want small talk then there is nothing better than a chatbot, and there are certainly open source chatbots that you could integrate with your own version of Siri. The only trick is working out when to pass something the user says on to which answering agent - and to get the right balance of usefulness and cuteness this would need some fine-tuning.
The point here is that Siri is almost an off-the-shelf, clip-together app - a mashup in the true sense of the term. As such it really isn't an edge for Apple.
As if to prove the point, a small group of adventurous programmers at Dextra decided to put together Iris (Siri spelled backwards) in a few hours, about 8 in total. The result is a Siri-like program that isn't quite up to scratch but what do you expect in 8 hours. You can talk to it and it answers your questions as best it can.
you can even try it out for yourself by downloading it from the Android Market. It clearly isn't currently well integrated with other Android apps, but it could be, and it is a proof in principle that Siri isn't anything special and certainly isn't another plus point for the iPhone.
If digital assistance, such as Siri or Iris, does become popular then there is a consequence that is difficult to predict. Search engines like Google are financed by advertising and Siri and Iris strip this out. Search engines such as Wolfram Alpha charge for commercial use beyond 2000 non-commercial API calls per month. Google similarly charges if you go beyond 100 queries per day. Does this mean that the end user will pay for Siri/Iris to answer their questions on a per-question basis? An interesting question?
Rubik's cube as computer game isn't a new idea, but in this case it is more a Rubik's cube that is a computer. The Rubik's Futuro Cube is a complete ARM-based machine with a form of touch input, wirel [ ... ]
A short time ago we reported that Google was using some sort of AI to find photos that contained named objects. It seemed a reasonable guess to say that it was using convolutional neural network but t [ ... ]