Présentation au printemps de la donnée 2025 à l'IGBMC - retour accueil
git clone git://bebou.netlib.re/prez-data-spring-igbmc
Log | Files | Refs |
prez.slides (3366B)
1
2
3
4
5
6
7
8
9
10 Environmental impacts of ICT :
11 Rarely discussed topics
12
13
14
15
16
17
18 › What topics ?
19
20 1. Can ICT contribute to reducing environmental impacts ?
21 2. How does Generative AI play into this ?
22 3. Beware of optimization
23
24
25 › 1. IT4Green
26
27 Every gram of co2e emitted by ICT enables other sectors to
28 avoid more than one gram of co2e.
29
30 - Studies from ICT lobbyists
31
32 › 1. IT4Green
33
34 Every gram of co2e emitted by ICT enables other sectors to
35 avoid more than one gram of co2e.
36
37 - Studies from ICT lobbyists
38
39 Multiple papers highlight the numerous weaknesses of those studies.
40
41 ----
42 Rasoldier et al., “How realistic are claims about the benefits of using digital
43 technologies for GHG emissions mitigation?”, Limits 22:2-4, 2022
44
45 Roussilhe et al., “A long road ahead: a review of the state of knowledge of the
46 environmental effects of digitization”, Current Opinion in Environmental
47 Sustainability, 62:101296, 2023.
48
49 › 1. IT4Green
50
51 * Perimeter (first and third order effects often ignored)
52 * High uncertainties
53 * Comparison scenarios
54 * Marginal effects of ICT in some contexts
55
56
57 › 1. IT4Green
58
59 * Perimeter (second and third order effects often ignored)
60 * High uncertainties
61 * Comparison scenarios
62 * Marginal effects of ICT in some contexts
63
64 "A digital solution cannot guarantee in itself net positive environmental
65 effects. Doing such an evaluation boils down to analysing a given context
66 and a wide array of factors in which a digital service is caught in to try
67 and infer plausible tendencies."
68
69 https://gauthierroussilhe.com/articles/les-emissions-evitees-de-la-numerisation
70
71 › 2. AI
72
73 * Impact of manufacturing of hardware by transistor goes up
74 * Energy consumption of hardware goes up
75 * Gen AI justifies building additional hardware of increasing complexity (and vice-versa)
76 -> In total contradiction with environmental goals we set for ourselves
77
78 › 2. AI
79
80 * Impact of manufacturing of hardware by transistor goes up
81 * Energy consumption of hardware goes up
82 * Gen AI justifies building additional hardware of increasing complexity (and vice-versa)
83 -> In total contradiction with environmental goals we set for ourselves
84
85 One part of "G Phase" (GenAI, GPU, GPT)
86
87 ----
88 https://gauthierroussilhe.com/articles/la-phase-g-les-gpu-et-les-ia-generatives-comme-nouvelle-phase-de-l-histoire-environnementale-de-la-numerisation-partie-1
89
90 › 3. Beware of optimization
91
92 The case study of the Unistra's datacenter
93
94 › 3. Beware of optimization
95
96 The case study of the Unistra's datacenter
97
98 * Geothermal power
99 * (partial and seasonal) reuse of heat
100 * 1.25 Power Usage Effectiveness (at best)
101 * Already artificialized land
102 * Local jobs
103
104 › 3. Beware of optimization
105
106 but
107
108 › 3. Beware of optimization
109
110 it is oversized
111
112 › 3. Beware of optimization
113
114 it is oversized
115
116 It is a political topic
117
118 In this context optimization :
119
120 * might be anecdotal (we don't really know)
121 * sucks the air out of the room and makes the questions of how much ICT we
122 want in our university and more importantly for what purpose less audible.
123
124 › Merci !
125
126 mail : arthur.pons@unistra.fr
127 site : http://arthur.bebou.netlib.re
128 slides : http://git.bebou.netlib.re/prez-data-spring-igbmc/log.html
129 or : ssh guest@bebou.netlib.re -p1459 prez-data-spring-igbmc
130