The Miseducation of This Machine

While machines continue to learn with more sophisticated algorithms and larger amounts of data, humans need to understand how such learning works in order to take its results with the proper grain of salt. Let's make ML tangible and thus help you become a better machine teacher!

Transcript

1. The Miseducation Of this machine @lauralindal Studying all the things { }
3. “Can Machines Think?” Alan Turing 1950 - , -
4. “Can Machines imitate thinking?” Alan Turing 1950 - , -
5. Can Machines Learn? Me 2019 - , -
6. Can machines imitate Learning? Me 2019 - , -
7. Reproduce Remix Reflect → →
8. Reproduce Remix → → Reflect Content Warning
9. I Reproduction .
14. “Those who own past data colonize the future.” Hito Steyerl 2019 - , -
15. Data reproduces.
16. II Remixing .
17. Before we can continue this talk, I need to know whether you are human.
18. Please select all images with a bike.
23. You’re not a robot.
24. But I might be.
30. ... ... ... ... ... ...
31. Convolutional Neural Network [ ]
36. Algorithms reproduce. Algorithms remix.
37. III Reflection .
38. Please select all comments with hate speech.
39. nohate online .
40. How do machines identify Hate Speech?
41. TRAIN
42. Convolutional Neural Network [ ]
43. it s ’ that gorram computer again
44. it s ’ that gorram computer again
45. it s ’ that gorram computer again
46. it s ’ that gorram computer again
47. it s ’ that gorram computer again
48. it s ’ that gorram computer again
49. it s ’ that gorram computer again
50. it s ’ that gorram computer again
51. TRAIN TEST
52. TRAIN TEST
53. TRAIN TEST
54. Where do machines fail?
56. “Oh, I feel like such an asshole now. Sorry, bud.”
57. Hate Speech without swear words·Rhetorical questions·Metaphors and comparisons·Idiosycratic and rare words·Sarcasm and irony·Quotations or references
59. How do humans identify Hate Speech?
60. “Hate speech is understood as public communication that contains deliberately and or intentionally / discriminatory messages. Thus it is not a question of hate , nor a question of speech.” Liriam Sponholz 2018 - , -
61. Verbal Aggression Dehumanization threatening legitimizing violence / Negative Generalization
65. Content Warning for Racism and Sexual Violence On the next two slides
66. [translated from German] “They should be blessed by those highly specialized pieces of gold whom they once welcomed with teddy bears.”
67. “They should be blessed by those highly specialized pieces of gold whom they once welcomed with teddy bears.” legitimizing violence Dehumanization Negative Generalization Irony ( )
68. [what it actually means] “Women who helped those uneducated refugees when they arrived first in Germany should be raped by them.”
69. Content Warning OFF
72. Where do humans Fail?
73. Humans reproduce. Humans remix. Humans reflect.
74. “Consequence is no coincidence.” Lauryn Hill 1998 - , -
75. Be the human that reflects.
76. Image sources (in order of occurrence) https://www.flickr.com/photos/kygp/5549385252 https://ssrotterdam.com/photo-video/ https://www.flickr.com/photos/xiffy/3749629455/ https://www.flickr.com/photos/bertknot/45238148191/