One of the latest artificial intelligence systems from MIT is as smart as a 4-year-old
When kids eat glue, they’re exhibiting a lack of common sense. Computers equipped with artificial intelligence, it turns out, suffer from a similar problem.
While computers can tell you the chemical composition of glue, most can’t tell you if it is a gross choice for a snack. They lack the common sense that is ingrained in adult humans. 
For the last decade, MIT researchers have been building a system called ConceptNet that can equip computers with common-sense associations. It can process that a person may desire a dessert such as cake, which has the quality of being sweet. The system is structured as a graph, with connections between related concepts and terms.
The University of Illinois-Chicago announced today that its researchers put ConceptNet to the test with an IQ assessment developed for young children. ConceptNet 4, the second-most recent iteration from MIT, earned a score equivalent to the average 4-year-old. It did well at vocabulary and recognizing similarities, but did poorly at answering “why” questions. Children would normally get similar scores in each of the categories.

One of the latest artificial intelligence systems from MIT is as smart as a 4-year-old

When kids eat glue, they’re exhibiting a lack of common sense. Computers equipped with artificial intelligence, it turns out, suffer from a similar problem.

While computers can tell you the chemical composition of glue, most can’t tell you if it is a gross choice for a snack. They lack the common sense that is ingrained in adult humans. 

For the last decade, MIT researchers have been building a system called ConceptNet that can equip computers with common-sense associations. It can process that a person may desire a dessert such as cake, which has the quality of being sweet. The system is structured as a graph, with connections between related concepts and terms.

The University of Illinois-Chicago announced today that its researchers put ConceptNet to the test with an IQ assessment developed for young children. ConceptNet 4, the second-most recent iteration from MIT, earned a score equivalent to the average 4-year-old. It did well at vocabulary and recognizing similarities, but did poorly at answering “why” questions. Children would normally get similar scores in each of the categories.

Notes

  1. toucha-mah-junka reblogged this from smarterplanet
  2. bozonkwark reblogged this from joshbyard
  3. thequietthingsthatno1knows reblogged this from smarterplanet
  4. aib0 reblogged this from consumedbychagrin
  5. jian198502 reblogged this from joshbyard
  6. melissajarquinmolina reblogged this from joshbyard
  7. hydrophobichacker reblogged this from smarterplanet
  8. wytchkraft-the-architekt-of-kaoz reblogged this from joshbyard and added:
    WOW.
  9. quisevadet reblogged this from johnasavoia
  10. johnasavoia reblogged this from therearepeoplewho and added:
    This is cool but that flow chart w/e is funny as hell
  11. danielsenshi reblogged this from joshbyard
  12. activenonsense reblogged this from joshbyard
  13. therearepeoplewho reblogged this from joshbyard
  14. joshbyard reblogged this from futurescope and added:
    MIT AI as Smart as a Four-Year Old (via smarterplanet)
  15. consumedbychagrin reblogged this from smarterplanet
  16. thatkidwithapassion reblogged this from erikangle
  17. elfboi reblogged this from bwansen
  18. ahardshell reblogged this from futurescope
  19. pizzahutdotcom reblogged this from futurescope
  20. kanagnostou reblogged this from futurescope

Recent comments

Blog comments powered by Disqus