Does wisdomofchopra.com pass the Turing test?

wisdom of chopraSome time ago I created wisdomofchopra.com, a site that generates random Deepak Chopra-style quotes. In case you aren’t aware, Deepak Chopra is a new age guru famous for interspersing his sentences with the word “quantum”. I thought it would be fun to make a random phrase generator based on words from his Twitter feed, thus wisdomofchopra.com was born. The source code for the generator is available on GitHub.

One thing I get asked a lot (often jokingly, of course) via the Wisdom of Chopra Twitter account is “does the generator pass the Turing test?”, or in other words, are the quotes generated by Wisdom of Chopra indistinguishable from those of Deepak Chopra himself? Fortunately, I happen to have data that can answer this very question! If you look at the top menu bar on the wisdomofchopra.com site, you will see a link to a quiz. The premise of the quiz is very simple: you are presented with a quote, and you have to tell if it is either a randomly generated quote, or a genuine musing of Dr Chopra. I have stored (anonymously, don’t worry!) every answer, so I can see how many of the answers are right and wrong.

When I present the raw numbers you will reach a predictable yet somewhat disappointing conclusion, but first lets approach this scientifically. We have two hypotheses, the first being that the two sets of quotes (random and real) are distinguishable from each other. If this were the case, you would expect to see significantly more right quiz answers than wrong ones. The second hypothesis is that Wisdom of Chopra passes the Turing test, and therefore the two sets of quotes are indistinguishable to humans. If this hypothesis is correct, then the person undertaking the quiz would be forced to choose an answer randomly and the task becomes akin to flipping a coin, so the number of right and wrong answers would be statistically the same.

So, onto the big reveal! I can tell you that from June 22nd 2012 to the present day, there have been 73,999 correct quiz answers and 39,581 wrong ones. Now, just by glancing at those numbers you should be able to conclude that the ayes have it and the generator fails the Turing test, but if you wanted to, you could put them through a binomial test to make absolutely sure! Still, around 35% of the answers are wrong, so telling the quotes apart is far from a doddle!

So there you have it. Even though it would be great if it did, I’m afraid I have to report that wisdomofchopra.com does not pass the Turing test. Sorry!

4 Comments

  1. in fairness, Chopra himself also fails the Turing test, bring indistingushable from a toddler pulling random words out of a box of fridge magnet “teach your baby to read” tiles.

  2. Paul Durrant

    I’m not surprised that the generator can be distinguished from genuine quotations. If you have the data, it would be interesting to split the results into the results for genuine quotations and random quotations.

    How many genuine quotations were erroneously identified as random?
    How many random quotations were erroneously identified as genuine?

Add Comment Register



Leave a Reply

Your email address will not be published. Required fields are marked *