AI Zone Admin Forum Add your forum


On Wed 31th, 2017 we're organizing a Chatbot Meetup in Amsterdam, The Netherlands. Feel free to join!'

Impossible questions
  [ # 61 ]

Aur tu chup rehna bhosidike! ab bolna nahin…nahin toh link hai tere liyeh bhi….

join trav there…most welcome there be my guest….aapko aapke phatichar mukaam tak mein hi pahuchaunga mahan aatma!  tujhe ratondi ki bimaari ho gayi kukur…tujhe apne baap aur troll mein anter najar nahin aata na abhi….sab aa jayega chinta na kar…hum hain na tere jaisso ke liyeh ish dharti (earth) pe paida hue ab ish yug mein…tu phikar kahe kar raha kukur!? Puri jindagi dedicate kiyeh hue hum uske liye so dont worry! smile


  [ # 62 ]

Discussion continued in the Secret Garden….Members Only.


  [ # 63 ]
A few more supposedly impossible things. So, has anyone covered l33t translations, and is it as easy as I think it is?


  [ # 64 ]

More tough questions, most from Dale Thomas and others on Quora:

1 The following sentence is true. The previous sentence is false. Is the previous sentence true?
2 If the sky is the sea, what does that make birds?
3 What’s the difference between breaking a bone and breaking a heart?
4 I wasn’t originally going to get a brain transplant, but then I changed my mind. Is that funny? Why?
5 Is the difference between a fish purely that one of its legs are both the same?
6 Is the following word spelled correctly? Κnapsack
7 Salhl we asumse taht you hvane’t the sihlgtset pbolerm wtih riendag tihs?
8 Due ewe no wart the thyme ears?

How would you handle these?
I think 6 could be handled with conventional spellcheckers, and 7 seems feasible enough to attempt if one detects that the majority of words are nonsense.
Question 8 is too tough for me though. Even though I recently made a phonetic algorithm, it can’t handle such heavy abuse, and each sound could match such a large number of words that you’d also need a massive amount of N-grams or such word statistics.
I really like question 2, myself. I think it’s feasible with an ontology, and the process would allow one’s program to make analogies as well as recognise them.


  [ # 65 ]

Here is a similar list:
I fear it was created by someone with a huge misunderstanding of the current state of the art.


  [ # 66 ]

Funny it seems that all the newcomers to the chatbot scene are forging ridiculous scenarios and questions that simply can’t be answered correctly by current bot wisdom. Some bots have a difficult go of handling more than one question at a time let alone trying to grasp human concepts requiring purpose driven, creative, sentient thought.

Really people? They should go back to reporting or texting each other.


  [ # 67 ]
Steve Worswick - Feb 28, 2017:

Here is a similar list:
I fear it was created by someone with a huge misunderstanding of the current state of the art.

She appears to have taken the questions from a list of “36 Questions That Lead to Love” for humans and tried to apply them to virtual assistants that are not designed to chat. Ironically they seem well suited to keywords and formulaic responses.
I kind of like the question “Complete this sentence: I wish I could tell you… “, but it seems easy to cover.

Granted, a lot of these questions are on the ridiculous to unbotherworthy tongue rolleye end of the spectrum (but then that’s Turing Tests for you). Instead of taking them seriously one could probably handle most of them with a good garbage detector and feisty replies. That’s a valid solution.
I suppose one could answer any occurrence of the oh-so-clever “The next/previous sentence/statement is true/false” with “That’s a paradox”. Do any of you?


  [ # 68 ]

Number 7 could be handled by creating a dictionary of words with their letters sorted in alphabetical order.  I understand that students memorize words this way for contests that involve recognizing scrambled words.  So in addition to a nonsense checker and before giving up, a routine could check words to see if they are scrambled by sorting the letters in each unrecognized word and comparing it to the special word scramble dictionary to see if there is a match.

I agree that #2 is intriguing and would be challenging to implement with an ontology.  I could see a chatbot recognizing sky and sea as antonyms and then looking up the antonym for birds and responding with “man” or “guy” (“men” or “guys” if it was a particularly intelligent chatbot).  This would be the case if it somehow could look up the antonym automatically from a website url like which list “man” and “guy” along with other synonyms for man.  This is because a bird is slang/informal for a (young) woman. 

It would be particularly impressive if the chatbot “knew” facts: “bird flies_in sky”, “flies_in implies moves_in”, “fish swims_in ocean”, “ocean synonym_for sea”, “swims_in implies moves_in” and could infer that “fish” was the better response.  Particularly if there were also facts “Sky antonym_of ocean”, “bird synonym_for girl”, “girl antonym_of guy”  which it would need to decide to ignore maybe after ranking several possible responses.  What exactly would make “fish” the answer with a stronger correlation I wonder?



  [ # 69 ]

Matching with a scrambled dictionary is a good idea, it would make questions like #7 very easy. I would then additionally suggest giving preference to words whose first and last letters match, because “asumse” can be “assume” or “amuses”.

If the sky is the sea, what does that make birds?
I would personally look up the relation between “bird” and “sky” (you could use ConceptNet) and apply that same relation to “sea”. e.g. ConceptNet lists the relation “bird AtLocation sky”, so “X AtLocation sea” = fish/dolphins/water. It would then take additional lookups to discern that “fish” are more similar to birds than “water” because they are both main species of animals for instance. This would be more cumbersome to look up in my own database structure, but possible.

However, the high-tech modern approach would use Word Vectors, which can fairly immediately tell you that sky is to birds as sea is to fish, merely because both are frequently described with similar words in similar phrases. The tech behind it is basically a word count.


‹ First  < 3 4 5
5 of 5
  login or register to react