Computers are plain stupid
(but that’s just common sense)
Nienke Eckhardt & Pim Nauts
Gezond Verstand 2008
T...
Today
. Computers vs. humans
. Common sense?
. been there, done that: previous research
. MIT’s OMCS
. A kid’s OMCS (‘ours...
Computers
. unmatched computational power
. fast, flawless & accurate reasoning
. virtually endless stamina
. all of which...
Humans
. limited working memory
. slow, often wrong reasoning
. need a recharge daily
. no hope for the better (as specie...
However, still
humans are ‘smart’, computers
aren’t.
ner!
Win
Howcome?
Howcome?
. Humans see & know about ‘The World’
. Can add meaning (not just compute)
a & experience.
Howcome?
Common sense!
“...what people in common would agree on; that
which they sense (in common) as their ...
Common sense?
. Humans have/learn lots of apparent obvious information
. Enables us to derive meaning from any given situa...
So...
. Enable computers to perform
human-like reasoning and derive
meaning
. Then they’ll understand (?) !
been there, done that?
. Earlier attempts have been made
. Most viable: Cyc by Doug Lenat (1984)
Cyc ...
New attempt: MIT’s OMCS
. Open Mind Common Sense (by Push Singh)
OMCS
. ‘Human Computation’ (Luis von Ahn)
. Uses power o...
OMCS
Problems
. Adults often unaware of their CS-knowledge
. High- vs. low level facts (ex. water)
. Still a tedious job....
so?
so?
Are children able to add reliable, useful common
sense knowledge to the OMCS database?
Experiment
Basis
. Current interface not appealing to kids
. Add a game element
Participants
. Participants from ‘Kinder...
Experiment
Data collection
Player #1 (describer) Player #2 (guesser)
Given concept Sees relations
...
Experiment
Experiment
Furthermore
. Each participant plays both roles
. 3 rounds played, divided into levels of difficulty
Interface...
Experiment
(possible) Research questions
. Do children prefer ‘easier’ words or particular
relationships?
.Which relatio...
Experiment
Current Issues
. No data at start.
. Game: FLEX/Flash to AMMP-arch
Next
. Working game
. Gather participants
...
Questions?
I do: Anyone any idea what future implications could be?
http://commons.media.mit.edu/nl/
...
Further reading
Marvin Minsky - Society of Mind / The Emotion Machine
(especially ...
of 23

Computers are plain stupid (but that's just common sense).

I was invented to do a guest lecture on my thesis topic for fellow students in June (2008). Presentation was done way before the actual game went online.
Published on: Mar 4, 2016
Source: www.slideshare.net


Transcripts - Computers are plain stupid (but that's just common sense).

  • 1. Computers are plain stupid (but that’s just common sense) Nienke Eckhardt & Pim Nauts Gezond Verstand 2008 Tilburg University
  • 2. Today . Computers vs. humans . Common sense? . been there, done that: previous research . MIT’s OMCS . A kid’s OMCS (‘ours’)
  • 3. Computers . unmatched computational power . fast, flawless & accurate reasoning . virtually endless stamina . all of which are only increasing . human-programmed
  • 4. Humans . limited working memory . slow, often wrong reasoning . need a recharge daily . no hope for the better (as species) . human-programmed...
  • 5. However, still humans are ‘smart’, computers aren’t. ner! Win
  • 6. Howcome?
  • 7. Howcome? . Humans see & know about ‘The World’ . Can add meaning (not just compute) a & experience.
  • 8. Howcome? Common sense! “...what people in common would agree on; that which they sense (in common) as their common natural understanding.” “Common sense is the collection of prejudices acquired by age eighteen.”
  • 9. Common sense? . Humans have/learn lots of apparent obvious information . Enables us to derive meaning from any given situation (hence context) . Cows are mammals . Caffeine keeps you awake . Mammals need oxygen
  • 10. So... . Enable computers to perform human-like reasoning and derive meaning . Then they’ll understand (?) !
  • 11. been there, done that? . Earlier attempts have been made . Most viable: Cyc by Doug Lenat (1984) Cyc Cows are mammals . 3mil CS-facts put in db. Mammals need oxygen + . System creates inferences cows need oxygen! = However . Millions of facts still not enough . Terribly expensive ($20mil) . Completely irrelevant assertions (“Albania produces crude petroleum products”)
  • 12. New attempt: MIT’s OMCS . Open Mind Common Sense (by Push Singh) OMCS . ‘Human Computation’ (Luis von Ahn) . Uses power of the web . Anyone can add knowledge (by connecting concepts)
  • 13. OMCS Problems . Adults often unaware of their CS-knowledge . High- vs. low level facts (ex. water) . Still a tedious job... However . Children love learning facts . Children are more conscious of these CS-facts
  • 14. so?
  • 15. so? Are children able to add reliable, useful common sense knowledge to the OMCS database?
  • 16. Experiment Basis . Current interface not appealing to kids . Add a game element Participants . Participants from ‘Kinderunversiteit’ (UvT, TU/e) . Aged 10 - 12 . Stimuli presented independently
  • 17. Experiment Data collection Player #1 (describer) Player #2 (guesser) Given concept Sees relations Milk IS A liquid Describes: AT LOCATION cornflakes IS A liquid AT LOCATION cornflakes Player #2 guesses: Milk Result . Correct guess = reliable description!
  • 18. Experiment
  • 19. Experiment Furthermore . Each participant plays both roles . 3 rounds played, divided into levels of difficulty Interface . Webbased . Simple, easy-to-use, attractive to children
  • 20. Experiment (possible) Research questions . Do children prefer ‘easier’ words or particular relationships? .Which relations contain more errors? .Is there a correlation between age and amount of difficult words correctly guessed? . ...etc
  • 21. Experiment Current Issues . No data at start. . Game: FLEX/Flash to AMMP-arch Next . Working game . Gather participants . Gather results ... write thesis!
  • 22. Questions? I do: Anyone any idea what future implications could be? http://commons.media.mit.edu/nl/ (pimnauts@xs4all.nl)
  • 23. Further reading Marvin Minsky - Society of Mind / The Emotion Machine (especially chapter 6) Push Singh - The Open Mind Common Sense Project (2002) http://www.kurzweilai.net/meme/frame.html?main=/articles/art0371.html Luis von Ahn - Human Computation (PhD.thesis) Carnegie-Mellon University, 2005 - Human Computation (Google TechTalk) http://video.google.com/videoplay?docid=-8246463980976635143 (2006)

Related Documents