Of Turing Tests and Mechanical Turks

Posted: November 4, 2011 in Tech

In an internet chat room, a stranger pings you. You strike a conversation that seems to go on well at first. But after sometime, you realize that something’s amiss. Is the other guy too amiable? Or too omniscient? Are there patterns in his replies? Is there a standard response? Then it strikes you, the “guy” is a bot, pretending to be human…

If the bot is still undetected and you end the conversation thinking that you just spoke to a stranger from another part of the world, the bot just passed the Turing Test. However, if we talk to the bot again, will it remember the details of our previous conversation? What information will it extract about our profile? Can it learn the rules of human behavior through Machine Learning or will we have to hard-code the Rules?

Let’s consider a conversation:

  • Bot (Pretending to be human): Hi
  • Me: Hi
  • Bot: Wassup dude!!
  • Me: nm man, you say
  • Bot: time pass! Btw, I’m a UG student, wbu?
  • Me: Am also a student. So what do u study?
  • Bot: Some practical stuff… (Not a clear answer, so that it can change its path depending on my answer).
  • Me: Nice man.. i thought all students just learn theory bullshit!! Take my college for example.. am an engineer.. but all I learn is loads of crap! So what do you exactly study?
  • Bot: Am also an engineer.
  • …. The conversation continues and finishes…

Now, the bot has some information about me. It can create a profile and “paint” a picture about my behavior and typing patterns, expected responses. Thus, if a bot is truly “Intelligent”, we can have an endless conversation and the bot will perhaps get to know more about me than any human friend (that is of course only if it stores and analyses my responses).

So can a following bot pass the Turing test? It has a few characteristic features such as:

  • Creating a profile for each conversation and mining it for information about its user. Let’s assume the bot has access to “sufficient” memory, how ever much it may be.
  • Access to Wikipedia and general web and ability to get relevant responses for questions that may be asked during a conversation.
  • Ability to be consistent with what it says during a conversation. (e.g if it says “I am an engineer” during one conversation with user X, whenever it speaks to user X, it is always an engineer.)
  • Ability to be human (respond to some queries in a way any human may respond: example: “My pet died.. :(“ Bot: “Sorry to hear about that.. 😦 My condolences”). This can only be done by analyzing large volumes of chat logs and mining for specific patterns: e.g {died, condolence || sorry || sad}; {won, congrats|| m/ || victory}).

Let’s further assume that the chatting will take place one-on-one (no group chat).

Now, this bot evolves overtime. It can effectively serve as a Personal Assistant ( Siri anyone? 🙂 ) Also, this leads to another potentially unknown situation – a possible conversation between two bots that each pretends to be human. Will they be able to accurately gauge each other? Over here, even the style of conversation will matter. One bot may have to initiate new leads.

Such a technology will revolutionize Customer Relations, if the bots can be trained. Perhaps, each customer may receive his own customized bot for technical and miscellaneous support which can be double up as a PA like Siri.

Every bot can now be a Siri?

Every bot can potentially double up as a Siri-like PA

With Understanding, can come functionality. Perhaps bots can be commanded to perform certain functions through voice/ chat commands. (Maybe sitting in your office, you can ask your PA in the home to make you tea when you return. Of course, such functionality is available to many internet enabled household devices, but having a single bot manage them all makes it much easier for you to control)

There may be boundless applications for such a bot.. I would conclude by saying that we have reached a crossroads… Two centuries back, there was this awesome pretend machine called the “Mechanical Turk” which was supposed to be an automatic chess playing machine. But the mechanical Turk was just an elaborate sitting room for a good human chess player who controlled the machine from inside.

The Role Reversal - Machine Inside, human outside!

(The new age Mechanical Turk – A Reversal of roles – Machine inside, but pretending to be Human!)

How ironic that the roles have been reversed.. We will now have machines pretending to be human; and they are definitely improving at that!

Advertisements

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s