Document (#41451)

Author

Harnett, K.

Title

Machine learning confronts the elephant in the room : a visual prank exposes an Achilles' heel of computer vision systems: Unlike humans, they can't do a double take

Issue

[20.09.2018].

Source

https://www.quantamagazine.org/machine-learning-confronts-the-elephant-in-the-room-20180920/

Year

2018

Abstract

In a new study, computer scientists found that artificial intelligence systems fail a vision test a child could accomplish with ease. "It's a clever and important study that reminds us that 'deep learning' isn't really that deep," said Gary Marcus , a neuroscientist at New York University who was not affiliated with the work. The result takes place in the field of computer vision, where artificial intelligence systems attempt to detect and categorize objects. They might try to find all the pedestrians in a street scene, or just distinguish a bird from a bicycle (which is a notoriously difficult task). The stakes are high: As computers take over critical tasks like automated surveillance and autonomous driving, we'll want their visual processing to be at least as good as the human eyes they're replacing. It won't be easy. The new work accentuates the sophistication of human vision - and the challenge of building systems that mimic it. In the study, the researchers presented a computer vision system with a living room scene. The system processed it well. It correctly identified a chair, a person, books on a shelf. Then the researchers introduced an anomalous object into the scene - an image of elephant. The elephant's mere presence caused the system to forget itself: Suddenly it started calling a chair a couch and the elephant a chair, while turning completely blind to other objects it had previously seen. Researchers are still trying to understand exactly why computer vision systems get tripped up so easily, but they have a good guess. It has to do with an ability humans have that AI lacks: the ability to understand when a scene is confusing and thus go back for a second glance.

Theme

Information

Field

Informatik

Form

Bilder

Similar documents (content)

1. Tsotsos, J.K.: Introduction to the special volume on computer vision (1995) 0.26

   0.26164624 = sum of:
     0.26164624 = product of:
       1.3082311 = sum of:
         0.04283082 = weight(abstract_txt:study in 4607) [ClassicSimilarity], result of:
           0.04283082 = score(doc=4607,freq=1.0), product of:
             0.065813705 = queryWeight, product of:
               1.0548558 = boost
               3.4708736 = idf(docFreq=3611, maxDocs=42740)
               0.017975643 = queryNorm
             0.6507888 = fieldWeight in 4607, product of:
               1.0 = tf(freq=1.0), with freq of:
                 1.0 = termFreq=1.0
               3.4708736 = idf(docFreq=3611, maxDocs=42740)
               0.1875 = fieldNorm(doc=4607)
         0.14236848 = weight(abstract_txt:intelligence in 4607) [ClassicSimilarity], result of:
           0.14236848 = score(doc=4607,freq=1.0), product of:
             0.12805334 = queryWeight, product of:
               1.201392 = boost
               5.929549 = idf(docFreq=308, maxDocs=42740)
               0.017975643 = queryNorm
             1.1117904 = fieldWeight in 4607, product of:
               1.0 = tf(freq=1.0), with freq of:
                 1.0 = termFreq=1.0
               5.929549 = idf(docFreq=308, maxDocs=42740)
               0.1875 = fieldNorm(doc=4607)
         0.15880352 = weight(abstract_txt:artificial in 4607) [ClassicSimilarity], result of:
           0.15880352 = score(doc=4607,freq=1.0), product of:
             0.13772784 = queryWeight, product of:
               1.2459488 = boost
               6.1494617 = idf(docFreq=247, maxDocs=42740)
               0.017975643 = queryNorm
             1.1530241 = fieldWeight in 4607, product of:
               1.0 = tf(freq=1.0), with freq of:
                 1.0 = termFreq=1.0
               6.1494617 = idf(docFreq=247, maxDocs=42740)
               0.1875 = fieldNorm(doc=4607)
         0.19336218 = weight(abstract_txt:computer in 4607) [ClassicSimilarity], result of:
           0.19336218 = score(doc=4607,freq=2.0), product of:
             0.16917363 = queryWeight, product of:
               2.1833615 = boost
               4.3104496 = idf(docFreq=1559, maxDocs=42740)
               0.017975643 = queryNorm
             1.1429806 = fieldWeight in 4607, product of:
               1.4142135 = tf(freq=2.0), with freq of:
                 2.0 = termFreq=2.0
               4.3104496 = idf(docFreq=1559, maxDocs=42740)
               0.1875 = fieldNorm(doc=4607)
         0.7708661 = weight(abstract_txt:vision in 4607) [ClassicSimilarity], result of:
           0.7708661 = score(doc=4607,freq=2.0), product of:
             0.45199266 = queryWeight, product of:
               3.9094498 = boost
               6.431782 = idf(docFreq=186, maxDocs=42740)
               0.017975643 = queryNorm
             1.7054837 = fieldWeight in 4607, product of:
               1.4142135 = tf(freq=2.0), with freq of:
                 2.0 = termFreq=2.0
               6.431782 = idf(docFreq=186, maxDocs=42740)
               0.1875 = fieldNorm(doc=4607)
       0.2 = coord(5/25)
   

2. Garner, B.J.; Forrester, C.L.; Lukose, D.; O'Kane, M.J.; Kenne, P.E.: ¬A knowledge interface for library users (1992) 0.18

   0.18158117 = sum of:
     0.18158117 = product of:
       0.6485042 = sum of:
         0.012713563 = weight(abstract_txt:with in 170) [ClassicSimilarity], result of:
           0.012713563 = score(doc=170,freq=1.0), product of:
             0.04616987 = queryWeight, product of:
               1.0201961 = boost
               2.5176222 = idf(docFreq=9369, maxDocs=42740)
               0.017975643 = queryNorm
             0.27536494 = fieldWeight in 170, product of:
               1.0 = tf(freq=1.0), with freq of:
                 1.0 = termFreq=1.0
               2.5176222 = idf(docFreq=9369, maxDocs=42740)
               0.109375 = fieldNorm(doc=170)
         0.022733973 = weight(abstract_txt:system in 170) [ClassicSimilarity], result of:
           0.022733973 = score(doc=170,freq=1.0), product of:
             0.061799426 = queryWeight, product of:
               1.0221795 = boost
               3.3633559 = idf(docFreq=4021, maxDocs=42740)
               0.017975643 = queryNorm
             0.36786705 = fieldWeight in 170, product of:
               1.0 = tf(freq=1.0), with freq of:
                 1.0 = termFreq=1.0
               3.3633559 = idf(docFreq=4021, maxDocs=42740)
               0.109375 = fieldNorm(doc=170)
         0.08304828 = weight(abstract_txt:intelligence in 170) [ClassicSimilarity], result of:
           0.08304828 = score(doc=170,freq=1.0), product of:
             0.12805334 = queryWeight, product of:
               1.201392 = boost
               5.929549 = idf(docFreq=308, maxDocs=42740)
               0.017975643 = queryNorm
             0.64854443 = fieldWeight in 170, product of:
               1.0 = tf(freq=1.0), with freq of:
                 1.0 = termFreq=1.0
               5.929549 = idf(docFreq=308, maxDocs=42740)
               0.109375 = fieldNorm(doc=170)
         0.092635386 = weight(abstract_txt:artificial in 170) [ClassicSimilarity], result of:
           0.092635386 = score(doc=170,freq=1.0), product of:
             0.13772784 = queryWeight, product of:
               1.2459488 = boost
               6.1494617 = idf(docFreq=247, maxDocs=42740)
               0.017975643 = queryNorm
             0.6725974 = fieldWeight in 170, product of:
               1.0 = tf(freq=1.0), with freq of:
                 1.0 = termFreq=1.0
               6.1494617 = idf(docFreq=247, maxDocs=42740)
               0.109375 = fieldNorm(doc=170)
         0.039649148 = weight(abstract_txt:systems in 170) [ClassicSimilarity], result of:
           0.039649148 = score(doc=170,freq=1.0), product of:
             0.106162965 = queryWeight, product of:
               1.7296007 = boost
               3.414623 = idf(docFreq=3820, maxDocs=42740)
               0.017975643 = queryNorm
             0.3734744 = fieldWeight in 170, product of:
               1.0 = tf(freq=1.0), with freq of:
                 1.0 = termFreq=1.0
               3.414623 = idf(docFreq=3820, maxDocs=42740)
               0.109375 = fieldNorm(doc=170)
         0.079757825 = weight(abstract_txt:computer in 170) [ClassicSimilarity], result of:
           0.079757825 = score(doc=170,freq=1.0), product of:
             0.16917363 = queryWeight, product of:
               2.1833615 = boost
               4.3104496 = idf(docFreq=1559, maxDocs=42740)
               0.017975643 = queryNorm
             0.47145543 = fieldWeight in 170, product of:
               1.0 = tf(freq=1.0), with freq of:
                 1.0 = termFreq=1.0
               4.3104496 = idf(docFreq=1559, maxDocs=42740)
               0.109375 = fieldNorm(doc=170)
         0.31796604 = weight(abstract_txt:vision in 170) [ClassicSimilarity], result of:
           0.31796604 = score(doc=170,freq=1.0), product of:
             0.45199266 = queryWeight, product of:
               3.9094498 = boost
               6.431782 = idf(docFreq=186, maxDocs=42740)
               0.017975643 = queryNorm
             0.70347613 = fieldWeight in 170, product of:
               1.0 = tf(freq=1.0), with freq of:
                 1.0 = termFreq=1.0
               6.431782 = idf(docFreq=186, maxDocs=42740)
               0.109375 = fieldNorm(doc=170)
       0.28 = coord(7/25)
   

3. Heidorn, P.B.: Image retrieval as linguistic and nonlinguistic visual model matching (1999) 0.16

   0.15524994 = sum of:
     0.15524994 = product of:
       0.48515606 = sum of:
         0.007264893 = weight(abstract_txt:with in 1967) [ClassicSimilarity], result of:
           0.007264893 = score(doc=1967,freq=1.0), product of:
             0.04616987 = queryWeight, product of:
               1.0201961 = boost
               2.5176222 = idf(docFreq=9369, maxDocs=42740)
               0.017975643 = queryNorm
             0.15735139 = fieldWeight in 1967, product of:
               1.0 = tf(freq=1.0), with freq of:
                 1.0 = termFreq=1.0
               2.5176222 = idf(docFreq=9369, maxDocs=42740)
               0.0625 = fieldNorm(doc=1967)
         0.04061696 = weight(abstract_txt:visual in 1967) [ClassicSimilarity], result of:
           0.04061696 = score(doc=1967,freq=1.0), product of:
             0.115434095 = queryWeight, product of:
               1.1406604 = boost
               5.629804 = idf(docFreq=416, maxDocs=42740)
               0.017975643 = queryNorm
             0.35186276 = fieldWeight in 1967, product of:
               1.0 = tf(freq=1.0), with freq of:
                 1.0 = termFreq=1.0
               5.629804 = idf(docFreq=416, maxDocs=42740)
               0.0625 = fieldNorm(doc=1967)
         0.08036646 = weight(abstract_txt:humans in 1967) [ClassicSimilarity], result of:
           0.08036646 = score(doc=1967,freq=1.0), product of:
             0.18193339 = queryWeight, product of:
               1.4320083 = boost
               7.0677705 = idf(docFreq=98, maxDocs=42740)
               0.017975643 = queryNorm
             0.44173566 = fieldWeight in 1967, product of:
               1.0 = tf(freq=1.0), with freq of:
                 1.0 = termFreq=1.0
               7.0677705 = idf(docFreq=98, maxDocs=42740)
               0.0625 = fieldNorm(doc=1967)
         0.016242027 = weight(abstract_txt:that in 1967) [ClassicSimilarity], result of:
           0.016242027 = score(doc=1967,freq=3.0), product of:
             0.06265504 = queryWeight, product of:
               1.4555527 = boost
               2.3946586 = idf(docFreq=10595, maxDocs=42740)
               0.017975643 = queryNorm
             0.2592294 = fieldWeight in 1967, product of:
               1.7320508 = tf(freq=3.0), with freq of:
                 3.0 = termFreq=3.0
               2.3946586 = idf(docFreq=10595, maxDocs=42740)
               0.0625 = fieldNorm(doc=1967)
         0.039019525 = weight(abstract_txt:researchers in 1967) [ClassicSimilarity], result of:
           0.039019525 = score(doc=1967,freq=1.0), product of:
             0.12865132 = queryWeight, product of:
               1.4748304 = boost
               4.852748 = idf(docFreq=906, maxDocs=42740)
               0.017975643 = queryNorm
             0.30329674 = fieldWeight in 1967, product of:
               1.0 = tf(freq=1.0), with freq of:
                 1.0 = termFreq=1.0
               4.852748 = idf(docFreq=906, maxDocs=42740)
               0.0625 = fieldNorm(doc=1967)
         0.055497248 = weight(abstract_txt:systems in 1967) [ClassicSimilarity], result of:
           0.055497248 = score(doc=1967,freq=6.0), product of:
             0.106162965 = queryWeight, product of:
               1.7296007 = boost
               3.414623 = idf(docFreq=3820, maxDocs=42740)
               0.017975643 = queryNorm
             0.52275527 = fieldWeight in 1967, product of:
               2.4494898 = tf(freq=6.0), with freq of:
                 6.0 = termFreq=6.0
               3.414623 = idf(docFreq=3820, maxDocs=42740)
               0.0625 = fieldNorm(doc=1967)
         0.06445406 = weight(abstract_txt:computer in 1967) [ClassicSimilarity], result of:
           0.06445406 = score(doc=1967,freq=2.0), product of:
             0.16917363 = queryWeight, product of:
               2.1833615 = boost
               4.3104496 = idf(docFreq=1559, maxDocs=42740)
               0.017975643 = queryNorm
             0.38099352 = fieldWeight in 1967, product of:
               1.4142135 = tf(freq=2.0), with freq of:
                 2.0 = termFreq=2.0
               4.3104496 = idf(docFreq=1559, maxDocs=42740)
               0.0625 = fieldNorm(doc=1967)
         0.18169488 = weight(abstract_txt:vision in 1967) [ClassicSimilarity], result of:
           0.18169488 = score(doc=1967,freq=1.0), product of:
             0.45199266 = queryWeight, product of:
               3.9094498 = boost
               6.431782 = idf(docFreq=186, maxDocs=42740)
               0.017975643 = queryNorm
             0.40198636 = fieldWeight in 1967, product of:
               1.0 = tf(freq=1.0), with freq of:
                 1.0 = termFreq=1.0
               6.431782 = idf(docFreq=186, maxDocs=42740)
               0.0625 = fieldNorm(doc=1967)
       0.32 = coord(8/25)
   

4. Fewtrell, B.: audio description : Accessing visual media (1997) 0.15

   0.15111549 = sum of:
     0.15111549 = product of:
       0.75557745 = sum of:
         0.01089734 = weight(abstract_txt:with in 404) [ClassicSimilarity], result of:
           0.01089734 = score(doc=404,freq=1.0), product of:
             0.04616987 = queryWeight, product of:
               1.0201961 = boost
               2.5176222 = idf(docFreq=9369, maxDocs=42740)
               0.017975643 = queryNorm
             0.23602709 = fieldWeight in 404, product of:
               1.0 = tf(freq=1.0), with freq of:
                 1.0 = termFreq=1.0
               2.5176222 = idf(docFreq=9369, maxDocs=42740)
               0.09375 = fieldNorm(doc=404)
         0.086161576 = weight(abstract_txt:visual in 404) [ClassicSimilarity], result of:
           0.086161576 = score(doc=404,freq=2.0), product of:
             0.115434095 = queryWeight, product of:
               1.1406604 = boost
               5.629804 = idf(docFreq=416, maxDocs=42740)
               0.017975643 = queryNorm
             0.7464136 = fieldWeight in 404, product of:
               1.4142135 = tf(freq=2.0), with freq of:
                 2.0 = termFreq=2.0
               5.629804 = idf(docFreq=416, maxDocs=42740)
               0.09375 = fieldNorm(doc=404)
         0.014066009 = weight(abstract_txt:that in 404) [ClassicSimilarity], result of:
           0.014066009 = score(doc=404,freq=1.0), product of:
             0.06265504 = queryWeight, product of:
               1.4555527 = boost
               2.3946586 = idf(docFreq=10595, maxDocs=42740)
               0.017975643 = queryNorm
             0.22449924 = fieldWeight in 404, product of:
               1.0 = tf(freq=1.0), with freq of:
                 1.0 = termFreq=1.0
               2.3946586 = idf(docFreq=10595, maxDocs=42740)
               0.09375 = fieldNorm(doc=404)
         0.37191015 = weight(abstract_txt:scene in 404) [ClassicSimilarity], result of:
           0.37191015 = score(doc=404,freq=1.0), product of:
             0.48577705 = queryWeight, product of:
               3.3091984 = boost
               8.166383 = idf(docFreq=32, maxDocs=42740)
               0.017975643 = queryNorm
             0.7655984 = fieldWeight in 404, product of:
               1.0 = tf(freq=1.0), with freq of:
                 1.0 = termFreq=1.0
               8.166383 = idf(docFreq=32, maxDocs=42740)
               0.09375 = fieldNorm(doc=404)
         0.27254233 = weight(abstract_txt:vision in 404) [ClassicSimilarity], result of:
           0.27254233 = score(doc=404,freq=1.0), product of:
             0.45199266 = queryWeight, product of:
               3.9094498 = boost
               6.431782 = idf(docFreq=186, maxDocs=42740)
               0.017975643 = queryNorm
             0.60297954 = fieldWeight in 404, product of:
               1.0 = tf(freq=1.0), with freq of:
                 1.0 = termFreq=1.0
               6.431782 = idf(docFreq=186, maxDocs=42740)
               0.09375 = fieldNorm(doc=404)
       0.2 = coord(5/25)
   

5. Kanade, T.: Immersion into visual meida : new applications of image understanding (1996) 0.15

   0.14758304 = sum of:
     0.14758304 = product of:
       0.73791516 = sum of:
         0.10052184 = weight(abstract_txt:visual in 4822) [ClassicSimilarity], result of:
           0.10052184 = score(doc=4822,freq=2.0), product of:
             0.115434095 = queryWeight, product of:
               1.1406604 = boost
               5.629804 = idf(docFreq=416, maxDocs=42740)
               0.017975643 = queryNorm
             0.8708159 = fieldWeight in 4822, product of:
               1.4142135 = tf(freq=2.0), with freq of:
                 2.0 = termFreq=2.0
               5.629804 = idf(docFreq=416, maxDocs=42740)
               0.109375 = fieldNorm(doc=4822)
         0.1406413 = weight(abstract_txt:humans in 4822) [ClassicSimilarity], result of:
           0.1406413 = score(doc=4822,freq=1.0), product of:
             0.18193339 = queryWeight, product of:
               1.4320083 = boost
               7.0677705 = idf(docFreq=98, maxDocs=42740)
               0.017975643 = queryNorm
             0.7730374 = fieldWeight in 4822, product of:
               1.0 = tf(freq=1.0), with freq of:
                 1.0 = termFreq=1.0
               7.0677705 = idf(docFreq=98, maxDocs=42740)
               0.109375 = fieldNorm(doc=4822)
         0.02320773 = weight(abstract_txt:that in 4822) [ClassicSimilarity], result of:
           0.02320773 = score(doc=4822,freq=2.0), product of:
             0.06265504 = queryWeight, product of:
               1.4555527 = boost
               2.3946586 = idf(docFreq=10595, maxDocs=42740)
               0.017975643 = queryNorm
             0.37040484 = fieldWeight in 4822, product of:
               1.4142135 = tf(freq=2.0), with freq of:
                 2.0 = termFreq=2.0
               2.3946586 = idf(docFreq=10595, maxDocs=42740)
               0.109375 = fieldNorm(doc=4822)
         0.039649148 = weight(abstract_txt:systems in 4822) [ClassicSimilarity], result of:
           0.039649148 = score(doc=4822,freq=1.0), product of:
             0.106162965 = queryWeight, product of:
               1.7296007 = boost
               3.414623 = idf(docFreq=3820, maxDocs=42740)
               0.017975643 = queryNorm
             0.3734744 = fieldWeight in 4822, product of:
               1.0 = tf(freq=1.0), with freq of:
                 1.0 = termFreq=1.0
               3.414623 = idf(docFreq=3820, maxDocs=42740)
               0.109375 = fieldNorm(doc=4822)
         0.43389514 = weight(abstract_txt:scene in 4822) [ClassicSimilarity], result of:
           0.43389514 = score(doc=4822,freq=1.0), product of:
             0.48577705 = queryWeight, product of:
               3.3091984 = boost
               8.166383 = idf(docFreq=32, maxDocs=42740)
               0.017975643 = queryNorm
             0.89319813 = fieldWeight in 4822, product of:
               1.0 = tf(freq=1.0), with freq of:
                 1.0 = termFreq=1.0
               8.166383 = idf(docFreq=32, maxDocs=42740)
               0.109375 = fieldNorm(doc=4822)
       0.2 = coord(5/25)