Siri Hacking...

I am slightly worried about this hole Apple Siri thing...

Now I am not normally the kind of guy that gets all paranoid about a company or government entity having detailed information about me online, however the human voice is something very unique to each individual.

If you have been living under a rock recently, Apple has just released its newest iPhone creation - the iPhone 4S. Essentially the 4S is a souped up version of the iPhone 4 with a better processor and camera. The most talked about feature however is a new piece of proprietary voice recognition and communication software called Siri. Now, it has been widely published that Siri was initially developed by the US government's top secret technology arm known as Darpa. This was then spun off into the Stanford Research Institutes hands and then became a company of its own before being swallowed by Apple a few years ago. That background alone could put some enthusiastic conspiracy theorists into a state of pandemonium.

What the software purports to do is to learn how the user talks - picking up on accents, phrases, and general grammar and language usage. To me this sounds amazing and truly is a remarkable step... The fineprint that got my attention though was the disclaimer on Apple's website that says Siri does this by recording the voice of the user and using this to 'learn' obviously through some kind of historical comparison algorithm.  The processing power is immense and it is assumed that a lot of this would be performed in the iCloud, a collection of mass servers that house immense amounts of data for access by users around the globe. Essentially your hard drive would be on the web and this hard drive would contain clearly recorded phrases that you have input to Siri.

I have two issues with this:

1. Siri as a software over time would establish a huge mass of recordings of the user's voice. If the software proves to be as popular and as good as what it sounds, the tool could become as common place as 'googling'. Therefore the software could theoretically come up with a library of words you have said that covers most of the dictionary. Now, where I am going with this is similar to those movies where they take a tapped phone call, rearrange the words and make it sound like the person is saying something completely different. This in reality is hard to do because we enunciate differently for words that can have alternative meanings and that can be positioned in different locations of a sentence. However, Siri should have theoretically learned how you would say a certain word in a certain context. So it is not to unreasonable to expect that the software after some time of use could string a coherent sentence together using proper grammar, word placement and verbal control such that it sounds exactly like you said it, even though you may never have said that sentence in the real world. In addition, if the software is good enough - even if Siri has not heard you say a certain word - it should be able to construct words (with the user's perfect voice and verbal control) by implementing algorithms that compare the user's implementation of certain syllables, vowels and consonants to that of a default recording of a word (eg. from other online recordings or voice softwares). It could theoretically construct sentences that you have never uttered! And all this in the cloud!

This could be useful in some ways I believe... For example, f it learns how to talk like the user when the user is sick... The user could tell the phone to dial his or her boss and tell them (in a sickly, but distinctively correct voice) that they are unable to come into work, thus leaving the user free to go play golf or do whatever...

But what if this fell into the wrong hands? What if a file became available that enabled anyone on the planet to be able to emulate the users voice over a voice only communication connection? There either could be a hell of alot of pranks going on or there could be a lot of issues with identity theft.

2. Apple has a tendency to patent anything... I am not so sure how they might go on this one, but I know that recordings of any kind, when original, are protected by copyright. Siri, for many millions if not billions, would be the first and only place where that voice is recorded. And the owner of the copyright? I have a funny feeling that Apple would argue that there software / cloud made the recording for the benefit of improving the software alone. The user did not have the intention of making a voice recording when asking 'where is the nearest pizza place?'. Consequently, for the next generation of iPhone users, Apple could theoretically own the copyright to pretty much any recording of every word you have input to the program. This could have massive complications!!!! Phone companies could be in breach of copyright in the future for recording a voicemail you leave, if you are ever interviewed by a news program they would be in breach of copyright and heaven forbid if you so desire to take on a role that requires you to speak publicly - everyone would have to pay Apple royalties just to hear your voice!

Hmmmm. I don't know if I want to buy an iPhone now... Mainly because I do not want to have to pay royalties for talking when I become President of the World. :p