Sunday, 12 May 2013

Person of Interest: The Machine (from a tech perspective)



This is a tech version of an earlier article. I should start by saying this: if you do not watch Person of Interest, then this article is not for you. It will make little or no sense to you. However, if you are one of the fortunate people that watch this show but are not a techie, then please read on. I theorized that Finch could still control the machine.  My primary concern here is The Machine’s command set/command structure.

Traditionally, issuing a command to a computer involves clicking or tapping on onscreen elements (icons and buttons) via a GUI (Graphical User Interface). Commands can also be issued by typing in specific textual syntax in a console. The name of this console and its appropriate command syntax varies between platforms/OSes. In UNIX based systems like Linux, Android, OS X and iOS, the console is known as the terminal. In Microsoft’s PC OSes, it is known as the Command Prompt. 

In recent years, a third method has emerged: Voice Control. The most common application of voice control is in automobiles where the driver or passenger can use voice commands to activate and control specific aspects of the vehicle’s functionality. This is usually restricted to infotainment and climate control. But like the aforementioned consoles, advanced voice control requires the use of a very specific syntax/command structure. Non-techies might know this command structure as the list of commands found in the car’s help system.

As is obvious, the state of affairs regarding voice control is not yet perfect. El Dorado will be the use of natural language/natural speech in issuing commands. Some progress has been made on this front. The most common example is the Wolfram Alpha Computational Engine. Don’t be scared by the big name. Many of you that have used Siri, Apple’s virtual assistant, have unwittingly used Wolfram Alpha. This is because Siri is powered by Wolfram’s servers. But even Wolfram is not completely natural speech in the sense that it can only answer fact-based questions. 

This brings me back to Person of Interest. In the very last scene of the 1st Season’s finale (Root Cause), we see Reese speaking to the Machine via a street/traffic camera that he needs to find Finch. And the phone rings. What I wish to point out here is that Reese issued a command in very natural speech and the Machine understood and complied. This was a form of control. In other words, user input. This method of input was further shown in the first episode of the 2nd season (The Contingency) where through flashbacks we see Finch testing and debugging the Machine by speaking to it. Later on in the episode, Reese again speaks to the Machine by threatening to down tools if it didn’t help him with finding Finch. This particular scene is very important. In literal terms, what happened here was a hack. What Reese inadvertently did was force the Machine to perform a function it had specifically been programmed not to do. And all this was done via natural speech. This natural speech command set was well demonstrated in the Season 2 finale. But The Machine in PoI is much more powerful than Wolfram Alpha. Its abilities go way beyond answering fact-based queries. It is able to understand and generate real-time natural speech in everyday parlance. 

So does such a system exist? It is very possible that in some secret lab somewhere lies something much powerful than Wolfram. But as far as the public is concerned and by what is publicly available, Wolfram is the best we’ve got. But just remember: You are being watched.

No comments:

Post a Comment