[MW Weekend] Can you use Siri to game HQ Trivia?

HQ, a trivia app show, has become a sensation. Even at the Momentum Works office, folks will gather each morning to try their luck, before the CEO rushes around to get everyone back to work.

Copied in China within days

Of course, as you would expect, there are already dozens of HQ copycats in China, many of them with millions of active users.

Wang Sicong, son of the richest tycoon in China and founder/funder of multiple successful startups, proclaimed that he would splash CNY100k (US$15.4k) for the winner of a particular trivia.

If US$15.4k draws half a million users, then the user acquisition cost is less than US$0.015 – definitely a very very good deal, given how expensive user acquisition in China has otherwise become.

Of course, as everything else in China – people would immediately think about how to game the system 😉

So we thought, can you game HQ with the help of, well, Siri?

Is Siri smart enough to crack HQ?

So the ideal flow here is – we wake Siri up, Siri listens to the host, and tells you the correct answer, and you put in the answer into the HQ app, all within 10 seconds!

Yes a stable (and fast) connection is crucial here as well!

So we tested that with an HQ question:

Which poem format contains five lines – Haiku, Limerick or Sonnet?

For your information, the correct answer is Limerick.

But we tested with Siri.

Well, first, waking Siri up and letting it listen to exactly (not before not after) what the presenter says is a challenge.

Also, Siri can’t process multiple choice questions, hence a bit useless to do trivia by its own.

So we tried another way to process it – three iPhones with three people asking separate questions:

“Siri, how many lines does a haiku have?”

“Siri, how many lines does a limerick have?

“Siri, how many lines does a sonnet have?”

Our office sounded like a boiler room.

Siri was apparently taken aback. For a few attempts it tried to correct the human voice, thinking we were actually talking about something else.

OK – not ideal to repeat a few times when you only have 10 seconds for everything.

Turned out – Siri is not good enough. We still need to overcome a lot of potential missteps to reach to the end, and if we do we will probably be happier than Lauren May (https://www.youtube.com/watch?v=dyynfOA1S6Y).

How about Google Assistant

We then thought – Siri is not good enough. How about Google – at the end of the day, Siri is more for performing personal tasks rather than information search, while the mighty Google survive on knowledge (or rather, how to pull it out quickly).

We indeed tried Google Assistant. It turned out – not much better. Besides the Android phones we have in the office come in different models, but all overloaded with apps and unnecessary junk – and not fast enough.  

Trivia AI can be done, though – at a cost

AI companies of course can build a smart bot to first shorten this flow, and second, more importantly, have a dedicated ‘intelligence’ to process just trivia.

Of course, if you are doing it just for a PR story (in a way AlphaGo is) – it is a bit too much effort; besides, in an age where AI talent is highly sought after (and expensive), you would better put them for some other use;)

Of course, if you still want to develop it,  one possibility is selling this as a service to HQ users – one dollar for each question answered correctly 😉

Then it is easier for Trivia to sell “extra lives” – currently extra lives are used for people to invite their friends, a smart way of acquiring new users at zero cost.

Conclusion

In any case, the above was a good team building exercise. FYI –  We have some gems in our office – some of our team members have dozens of pub trivia trophies at home, and one of them even identified five different types of rodents by their photos (see below):

Challenge to your readers: Write to [email protected] identifying the rodents (top left to bottom right). If you can identify these correctly, you win a chance to have lunch with the MW team at our office!  Challenge is up until 20-Jan-2018.