ABCDEFGHIJKLMNOPQRSTUVW
1
Interactive Workshop on Computational Measurements of Machine Creativity
June 20, 2021, CVPR'21
2
3
Group 1Group 2Group 3Group 4Group 5Group 6
4
5
Discussion 1Elements of CreativityElements - G1Elements - G2Elements - G3Elements - G4Elements - G5Elements - G6
6
With your group, create a list of the key elements of computational creativity. Summarize those elements (with their longer definitions) in the sheet for your group. (30 minutes)"expectation" and "entropy" -- meeting or violating expections; novelty and surprise. Market forces. Ethics.Novelty and surprise -- we're not sure how to quantify novelty without training data. How the user feels about the system (see creativity support index). What are user goals when interacting with system? Randomness -- how do we quantify/beat randomness?"criticality" and the importance of the audience to creativity. Metrics for novelty tend to find things that are unusal but not, perhaps, critical (== thought-provoking). Is computer-generated random or self-reflective. Creativity in relation to humanity.Shared experience between creative work and audience. Is creativity in the process or in the result? "ground" (context) of creavity vs "extension". Experience of art grows (positive ambuguity?)
7
Nominate a presenter to present results in main room (1 minute)
8
9
Discussion 2Evaluating ArtworksEvaluating - G1Evaluating - G2Evaluating - G3Evaluating - G4Evaluating - G5Evaluating - G6
10
Using the elements of creativity described in the previous section, assess examples of art made with creative AI. We have provided a list of artworks to choose from. (30 minutes)Hard to choose numbers. Though "violated expecation of creator" was possible to fill in, at least. Evaluating art w/o artist.Taking averages is hard -- very inconsistent/subjective scores. "depth" was an interesting axis.Many of our axes are audience-relative, so it's hard to come up with a [consistent] rating. Ranking was easier than rating. "system agency" was reasonable to rate.Hard to define/evaluate aspects separately ("novelty" vs "inventiveness", for example). Hard to interpret the artwork. Knowledge of how the work was created affects the novelty.
11
Nominate a presenter to present results in main room (1 minute)
12
13
Discussion 3Revise and ReflectRevise - G1Revise - G2Revise - G3Revise - G4Revise - G5Revise - G6
14
After hearing from other groups, revisit and revise your metrics. Reflect on which metrics were successful, which were not. Combine redundant metrics. Complete a final set of metrics with definitions, and then apply those metrics to a new set of artworks. (30 minutes)
15
Nominate a presenter to present results in main room (1 minute)
16
17
18
CVPR'21cmmc-cvpr21.comAhmed Elgamal, Hyeju Jang, Eunsu Kang, James McCann, Jean Oh, Devi Parikh, Peter Schaldenbrand,
Robert Twomey, Jun-Yan Zhu
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100