pages/txt/efficacy-deepfakes.txt (view raw)
1 11 May, 2020
2
3The efficacy of deepfakes
4
5Can we really write it off as "not a threat"?
6
7 A few days back, NPR put out an article discussing why deepfakes aren't
8 all that powerful in spreading disinformation. [1]Link to article.
9
10 According to the article:
11
12 "We've already passed the stage at which they would have been most
13 effective," said Keir Giles, a Russia specialist with the Conflict
14 Studies Research Centre in the United Kingdom. "They're the dog that
15 never barked."
16
17 I agree. This might be the case when it comes to Russian influence.
18 There are simpler, more cost-effective ways to conduct [2]active
19 measures, like memes. Besides, America already has the infrastructure
20 in place to combat influence ops, and have been doing so for a while
21 now.
22
23 However, there are certain demographics whose governments may not have
24 the capability to identify and perform damage control when a
25 disinformation campaign hits, let alone deepfakes. An example of this
26 demographic: India.
27
28the Indian landscape
29
30 The disinformation problem in India is way more sophisticated, and
31 harder to combat than in the West. There are a couple of reasons for
32 this:
33 * The infrastructure for fake news already exists: WhatsApp
34 * Fact checking media in 22 different languages is non-trivial
35
36 India has had a long-standing problem with misinformation. The 2019
37 elections, the recent CAA controversy and even more recently -- the
38 coronavirus. In some cases, it has even lead to [3]mob violence.
39
40 All of this shows that the populace is easily influenced, and deepfakes
41 are only going to simplify this. What's worse is explaining to a rural
42 crowd that something like a deepfake can exist -- comprehension and
43 adoption of technology has always been slow in India, and can be
44 attributed to socio-economic factors.
45
46 There also exists a majority of the population that's already been
47 influenced to a certain degree: the right wing. A deepfake of a Muslim
48 leader trashing Hinduism will be eaten up instantly. They are inclined
49 to believe it is true, by virtue of prior influence and given the
50 present circumstances.
51
52countering deepfakes
53
54 The thing about deepfakes is the tech to spot them already exists. In
55 fact, some can even be eyeballed. Deepfake imagery tends to have weird
56 artifacting, which can be noticed upon closer inspection. Deepfake
57 videos, of people specifically, blink / move weirdly. The problem at
58 hand, however, is the general public cannot be expected to notice these
59 at a quick glance, and the task of proving a fake is left to
60 researchers and fact checkers.
61
62 Further, India does not have the infrastructure to combat deepfakes at
63 scale. By the time a research group / think tank catches wind of it,
64 the damage is likely already done. Besides, disseminating contradictory
65 information, i.e. "this video is fake", is also a task of its own.
66 Public opinion has already been swayed, and the brain dislikes
67 contradictions.
68
69why haven't we seen it yet?
70
71 Creating a deepfake isn't trivial. Rather, creating a convincing one
72 isn't. I would also assume that most political propaganda outlets are
73 just large social media operations. They lack the technical prowess and
74 / or the funding to produce a deepfake. This doesn't mean they can't
75 ever.
76
77 It goes without saying, but this post isn't specific to India. I'd say
78 other countries with a similar socio-economic status are in a similar
79 predicament. Don't write off deepfakes as a non-issue just because
80 America did.
81
82References
83
84 1. https://www.npr.org/2020/05/07/851689645/why-fake-video-audio-may-not-be-as-powerful-in-spreading-disinformation-as-feare
85 2. https://en.wikipedia.org/wiki/Active_measures
86 3. https://www.npr.org/2018/07/18/629731693/fake-news-turns-deadly-in-india