The self-learning neural network 20Q.net has already asked a billion questions
Experimental artificial intelligence 20Q.net overcame the mark of 50,000,000 tests and came close to the mark of a billion questions. This neural network is built on the game "Twenty Questions". A person makes a word, and the system guesses it, taking into account his age, gender and nationality in the process of communicating with a person.
Artificial intelligence is trained in humans. Receiving answers from us, he adds to his knowledge base. A billion answers means that now he is able to guess almost every thought of a person.
To check, we guess the word "UFO" and see how quickly the system can guess it. As answers you can choose not only “yes” and “no”, but also fuzzy categories: maybe, perhaps, doubtfully, partially, etc., and also mark questions as irrelevant.
')
Question 1. Is it animal, plant, or mineral? Other.
Question 2. Do you keep it during use? Irrelevant question.
Question 3. Is it in the office? Not.
Question 4. Do you use it at night? Irrelevant question.
Question 4-2. Is it helpful? Unknown.
Question 5. Is it more microwave? Yes.
Question 6. Is it cold blood? Irrelevant question.
Question 7. Can I buy it? Not.
Question 8. There you can put something? Probably.
Question 9. Does he have a hard shell? Yes.
Question 10. Is it metallic? Yes.
Question 11. Does it provide protection? Yes.
Question 12. Is it direct? Not.
Question 13. Can you live in it? Yes.
Question 14. Does it glow? Sometimes.
Question 15. Can it be very hot? Probably.
Question 16. Can it be refilled? Irrelevant question.
Question 17. Is this a spaceship ? Close.
Question 18. Have you seen this in life? Not.
Question 19. Will it break if it falls? Not.
Question 20. Is this a space station ? Close.
Question 21. Can you control this? Not.
Question 22. Does it come in different colors? May be.
Question 23. Is it round? Yes.
Question 24. Can you hear it? Yes.
Is that a UFO?
Very good. Especially when you consider that, in fact, the 17th question was already correct. Artificial intelligence is clearly progressing. What if he soon starts guessing the words from the first question?
Artificial intelligence is trained in humans. Receiving answers from us, he adds to his knowledge base. A billion answers means that now he is able to guess almost every thought of a person.
To check, we guess the word "UFO" and see how quickly the system can guess it. As answers you can choose not only “yes” and “no”, but also fuzzy categories: maybe, perhaps, doubtfully, partially, etc., and also mark questions as irrelevant.
')
Question 1. Is it animal, plant, or mineral? Other.
Question 2. Do you keep it during use? Irrelevant question.
Question 3. Is it in the office? Not.
Question 4. Do you use it at night? Irrelevant question.
Question 4-2. Is it helpful? Unknown.
Question 5. Is it more microwave? Yes.
Question 6. Is it cold blood? Irrelevant question.
Question 7. Can I buy it? Not.
Question 8. There you can put something? Probably.
Question 9. Does he have a hard shell? Yes.
Question 10. Is it metallic? Yes.
Question 11. Does it provide protection? Yes.
Question 12. Is it direct? Not.
Question 13. Can you live in it? Yes.
Question 14. Does it glow? Sometimes.
Question 15. Can it be very hot? Probably.
Question 16. Can it be refilled? Irrelevant question.
Question 17. Is this a spaceship ? Close.
Question 18. Have you seen this in life? Not.
Question 19. Will it break if it falls? Not.
Question 20. Is this a space station ? Close.
Question 21. Can you control this? Not.
Question 22. Does it come in different colors? May be.
Question 23. Is it round? Yes.
Question 24. Can you hear it? Yes.
Is that a UFO?
Very good. Especially when you consider that, in fact, the 17th question was already correct. Artificial intelligence is clearly progressing. What if he soon starts guessing the words from the first question?
Source: https://habr.com/ru/post/7574/
All Articles