Declarative vs. procedural knowledge (Turing test)


Imagine a mad scientist who is trying to create an artificial being indistinguishable from a human and early in the development of this machine, he asks you for help testing the mind. And so he confines the mind to a box. The only way it can interact with the world is by reading input messages and responding by printing output messages. And he presents you with the following challenge, he shows you two identical boxes: the other is connected to a human being typing the output messages. And you are allowed to ask the boxes any sequence of questions you can imagine. Do you think you could distinguish the machine from the human? What would you ask? A triumph of mathematical and mechanical skill, is this great new automatic calculator at Harvard University. Intricate problems in mathematics put through the machine in coded form on tape are accurately solved in a minute fraction of the time required for a human calculation. In the first half of the twentieth century general purpose computers were only just emerging. But speculations were swirling about what would or wouldn’t be possible in the near future with robots or intelligent machines. It was a time when many human tasks were quickly being swallowed up by machines. People feared powerful industrialists would soon be supported by great machines. Machines which could do anything. Resulting in a class of poor labourers who would simply toil away to power the machines. In nineteen-fifty, Alan Turing wanted to ground these speculations about machine intelligence using a scientific approach known as a Turing Test. Where a successful pass would mean unequivocally that whatever is in the box could be identified as intelligent. And to this day, no machine has been able to accurately mimic conversation over long periods of time. Historically, the concept of knowledge and answering questions have always been intertwined. Across many cultures, there is a deep rooted belief that the source of all knowledge comes from a higher power. And one way to access knowledge from a higher power is by posing questions and waiting for a sign. In ancient Greece, citizens could take their toughest questions: usually key decisions in social and political life to a medium who would enter a special location known as an oracle. For example the medium Pythia from Delphi would enter the Sanctuary of Apollo ask her question, fall into a trance-like state, and emerge with the answer. Turing referenced this history with his imaginary oracle machine. Inside his oracle machine would be access to an oracle database: a database which contained all answers to all questions. In one column there is a list of all possible sentences, think of this as our question column, and many of these sentences would be meaningless but every question you can think of is in this book. And in a second column: the corresponding answers. If a machine had this oracle database, it would be able to answer any question. But could Turing’s oracle machine exist in the physical world? Well the question column alone would contain every possible combinatoin of English words. Even if we limit ourselves to all possible twelve word sequences, that results in 10 to the power of 72 different sentences. Now even if you shrunk every letter down to the size of an atom, you would still end up with a book larger than the size of the visible universe. So the oracle database cannot exist. When we store knowledge this way, by simply memorizing facts, we call this declarative knowledge. For instance memorizing your multiplication tables is declarative knowledge. Music boxes can be said to ‘know’ a song declaratively. The music-roll declares each note. This works well with small numbers or short songs, but as the numbers get larger and songs get longer you, will eventually run out of memory. So if a machine only knows declaratively, meaning it memorizes everything, then there is a very real physical limitation: space. So if a machine is passing a Turing Test, we know it can’t simply be memorizing all possible answers. It must be taking advantage of another kind of knowledge. It must know how to know. Meaning it goes through a process to get an answer. For example knowing the rules of multiplication or knowing how to compose new music. And this kind of knowledge, known as procedural knowledge, is the focus of computer science. How do things become known? (Joe Fowler)

40 Comments

  1. Steven Prescott

    April 4, 2016 at 2:17 pm

    These videos are filled with an excellent amount of data and overall very good and insightful. In terms of explaining basic computer/cryptographic theory this is clearly one of the best channels on YouTube.

  2. Семен Лутай

    April 4, 2016 at 2:29 pm

    You guys doing a great job!

  3. SWAPAN JAIN

    April 4, 2016 at 3:02 pm

    you are back.

  4. Greg & Meryl Debrosse

    April 4, 2016 at 3:11 pm

    Turing tests made clear.

  5. Tinc

    April 4, 2016 at 4:43 pm

    I was disappointed with the end of the video until I realized it's only part 1. Now I'm more excited than before!

  6. Temo Lopez

    April 4, 2016 at 4:55 pm

    Freaking awesome.

  7. Happ MacDonald

    April 4, 2016 at 7:05 pm

    Now that you've had three to four different narrators, I have become curious how you keep the same pace and style of speech so consistently between them.
    Are y'all using a music sheet? Be honest, now. xD

  8. ZejithThemis

    April 4, 2016 at 8:13 pm

    The Turing test was managing to be convincing (indistinguishable from the human control) for 70% of 5 minutes of conversation.
    There are real humans today who would fail that, and I'm sure that a sufficient amount of computer power with a well tweaked NN chatterbot would pass.
    The Turing test is not a test of artificial intelligence, it's just an arbitrary idea of what that might look like, invented in an era when a C64 would have occupied a warehouse (hyperbole).
    It's a joke.

  9. Simon Björkman

    April 4, 2016 at 9:51 pm

    Excellent video

  10. WeBDaEMoN33

    April 4, 2016 at 10:26 pm

    Great video! Can't wait for more

  11. mark ptak

    April 4, 2016 at 11:24 pm

    Had to wait all day at work before watching this…….somehow I knew I'd want to savor it.

  12. mark ptak

    April 4, 2016 at 11:31 pm

    Yup….pretty awesome

  13. Noah Hounshel

    April 5, 2016 at 1:51 am

    You know I think you are one of the first people I've seen accept bitcoin on youtube.

  14. Jacob Pooler

    April 5, 2016 at 2:47 am

    Good Stuff!!!

  15. Sam C

    April 6, 2016 at 1:03 am

    great video as always

  16. Arlyn Swain

    April 6, 2016 at 1:48 pm

    New vid! Love it

  17. hellothere11

    April 6, 2016 at 7:47 pm

    Love it!

  18. Mackaber Witckin

    April 11, 2016 at 2:31 pm

    hehe, Oracle Database, does that count as "Product Placement"?

  19. BeefaloKing

    April 13, 2016 at 1:38 am

    How are these videos not more popular?

  20. Jake

    April 13, 2016 at 9:26 pm

    The black and white film is Metropolis by Fritz Lang for anyone interested.

  21. Jonathan Thomas

    April 14, 2016 at 5:47 pm

    Except that Oracle machine with all questions already exists. It is called the library of Babel. It started this last year. Any string of characters is already categorised and searchable.

  22. Julie Cruise

    April 15, 2016 at 10:08 pm

    more please

  23. Agustin Berasategui

    April 16, 2016 at 1:47 pm

    Another piece of art from Brit! Your videos are completely awesome!

  24. Shawn

    April 16, 2016 at 10:22 pm

    I'm going to finish my CS degree before this series is completed

  25. Marc Raynaud

    April 18, 2016 at 3:50 pm

    I built the machine invented by Alan Turing in his 1936 publication.
    If you want to see all the details are on the website http://www.machinedeturing.org

  26. John Batamica

    April 20, 2016 at 5:04 pm

    I was so confused in Systems and Signal class, but your videos on Language of Coins gave me better conceptual understanding than a textbook or a professor ever could.
    Thank you

  27. Wadson Fourrien

    October 25, 2016 at 12:21 am

    Wow! I love your videos. Your channel deserves more views. I'm a computer science student. More CS students need to watch this.

  28. Ghost Emblem

    November 3, 2016 at 4:28 am

    I would say procedural knowledge is the focus of machine learning computer science encompasses more than AI design.

  29. Eddie Bibby

    December 26, 2016 at 10:24 pm

    I do like this channel however the narration and script of this video was irritatingly slow and laboured.

  30. M T

    January 20, 2017 at 12:20 pm

    Assuming the list of sentences is in english, the number would not be the total number of combinations of words, but rather it would need to follow the rules of sentence construction. So, it would be much smaller than the number mentioned.

  31. Sudev Sen

    January 22, 2017 at 9:05 pm

    Metropolis is awesome

  32. itsVollx

    June 18, 2017 at 4:33 pm

    great video!

  33. WitherBossEntity

    October 6, 2017 at 1:46 pm

    Wait, so if I use an Oracle database, I know everything?

  34. Bidur Khanal

    January 13, 2018 at 5:57 pm

    this is not what i was searching for . . . . a thumbs up to this video though

  35. TheLurkingPanda

    March 12, 2018 at 3:58 am

    how does this channel have so few subs/views

  36. Yufei Zhan

    December 24, 2018 at 7:00 am

    I literally fall in love with these videos. They are sooooooo good.

  37. John W

    March 20, 2019 at 10:45 pm

    Your videos are a beautiful intersection of art and computer science. Thank you.

  38. Naureen Bokhari

    April 21, 2019 at 8:00 pm

    very representation… I don't comment much… but this deserved a positive comment. ^^

  39. Michael Hoefler

    June 21, 2019 at 4:11 am

    What’s 1 / 0 ?

Leave a Reply