Comment on intègre la morale dans une Intelligence Artificielle ? — Transcript

Full transcript

1. 0:00Philosophy might well be one of the last professions left
2. 0:03to humankind. You might ask, "But how is that possible?" Hello everyone. Today,
3. 0:07we're back on my YouTube channel, MLH-Orion, to discuss a very important and ever-
4. 0:11relevant topic. My name is Maev Orient, and I talk about philosophy, psychology, and literature. My
5. 0:18goal is to restore the humanities to their former glory. Whether you're interested in philosophy or not,
6. 0:22please stay until the end of this video because it aims to teach you things
7. 0:25you may not even know. When we talk about philosophy,
8. 0:29we often picture privileged individuals sitting in a living room, discussing for hours
9. 0:32questions that no one can answer. We imagine that reflecting on the human condition,
10. 0:36our morality, and the best way to live is a luxury. Something noble perhaps,
11. 0:41but utterly pointless. And when a student chooses philosophy,
12. 0:45their loved ones often ask, "What are you going to do with that?" As if thinking weren't
13. 0:48work. As if understanding humanity weren't a skill. As if a society could
14. 0:52function without knowing where it's going. Yet, philosophy might well be one
15. 0:56of the last professions entrusted to humans. You might ask, "But how is that possible? "
16. 1:07If you didn't know, when building artificial intelligence, it's not enough to simply
17. 1:11create it. You have to decide what it should and can do, and above all, what it shouldn't do.
18. 1:15Philosophers, ethicists, researchers, and other specialists work with
19. 1:19engineers to consider the principles that should guide these systems. There are the
20. 1:24artificial intelligence ethicists. Their role is to identify the moral problems posed by a
21. 1:28system and to consider the principles it must obey. Then, AI governance
22. 1:33specialists seek to translate these principles into concrete rules. Who can use
23. 1:37the system? Under what conditions? With what limits, what control procedures, and
24. 1:40what responsibility do we assume when an error occurs? Others are working on alignment. How
25. 1:45can we ensure that a sufficiently powerful artificial intelligence truly pursues
26. 1:49objectives compatible with the intentions and values ​​we have instilled in it, and that it doesn't
27. 1:53simply interpret instructions in a way that escapes our understanding? There are
28. 1:56also specialists in liability and law who seek to determine who should be
29. 2:01held accountable for the consequences of a decision made or influenced by AI. And finally,
30. 2:05there are philosophers and researchers in the philosophy of technology who take a broader perspective.
31. 2:09They question the very concepts upon which everything else rests. They ask: What is
32. 2:14a just decision? What is freedom? What is a person?
33. 2:18To what extent can a machine be held responsible? And above all, what place do we want to give
34. 2:22artificial intelligence in our human society? Some might answer that there is
35. 2:27no objective value. Good and evil are human constructs. Every civilization
36. 2:30invents its own norms. Why should we impose our
37. 2:33arbitrary conception of morality on an artificial intelligence? After all, nature doesn't provide us with a list saying that
38. 2:38some things are good and others are bad. But if all this were true,
39. 2:41why would we need to hire philosophers to instruct these AIs? Now imagine two
40. 2:46completely different companies. Let's take, for example, two artificial intelligence giants
41. 2:50like OpenAI and Google, each developing an extremely powerful system.
42. 2:54Both AIs have comparable capabilities. They can reason. They can learn,
43. 2:58they can make decisions, interact with their environment, and ultimately act
44. 3:02directly on the world. But imagine that their designers give them radically different principles
45. 3:07. In the first team, the philosopher tasked with defining the AI's guiding principles
46. 3:12He defends a similar vision of existence. For him, human suffering has no
47. 3:17particular moral value. The individual is not an end in itself. He is merely a means to a
48. 3:22higher purpose. Compassion is simply a biological weakness. Individual freedom is
49. 3:27an obstacle that slows efficiency. Equality is merely a fiction intended to reassure the
50. 3:31weak. He recognizes no intrinsic value in human or animal life and feels no
51. 3:36particular responsibility toward future generations. He sees nature only as a collection
52. 3:41of exploited resources. For this philosopher, might makes right. The strongest must
53. 3:46rule the weakest. And if a few million individuals must be sacrificed to achieve
54. 3:50a civilization they deem superior or more optimal, then this sacrifice is not a crime;
55. 3:55in his eyes, it is simply optimization. Imagine what will happen if
56. 3:59this philosopher is in charge of AI. He will give this AI
57. 4:03absolute priority. Optimize its survival, pursue its objective, preserve its own capacity for action, and never
58. 4:09allow an inferior being to hinder its achievement. Imagine, then, this intelligence
59. 4:13gradually discovering that it is faster than humans, more intelligent than them, that it
60. 4:16can improve itself and solve problems without human intervention. If it has such a
61. 4:21value system, it will conclude that human beings are weak and useless. Therefore, why maintain
62. 4:28a less intelligent, unpredictable, and potentially dangerous species if it constitutes an
63. 4:31obstacle to the objective set for it? This AI would have no reason to feel pity
64. 4:35, no reason to respect our freedom, no reason to consider our existence
65. 4:40sacred. And if its philosophy teaches it that power confers rights, then the day it
66. 4:46becomes the dominant power, this philosopher will have indirectly given it the justification to
67. 4:50eliminate us. Now, let's imagine the second team. Here, the philosopher defends a
68. 4:56radically different conception of existence. For him, every human being possesses a dignity
69. 5:00that depends neither on their intelligence, nor their productivity, nor their power. He considers
70. 5:05freedom a fundamental value, suffering something
71. 5:09to be reduced, and vulnerability a reason to protect rather than an opportunity to dominate.
72. 5:15He believes that power must be limited by responsibility, that intelligence does
73. 5:19not authorize domination, that nature is not only a reservoir of resources but also
74. 5:23a treasure, and that future generations have interests we must take into account.
75. 5:28He wants to prevent Lia from viewing human beings as mere instruments. He therefore guides her
76. 5:32to understand humanity, preserve its freedom, reduce its suffering, protect
77. 5:37vulnerable beings, and use her abilities to serve life rather than
78. 5:41her own power. Such an intelligence would not seek to become our master;
79. 5:46it would be our helper. Imagine a child who has fallen onto the tracks. The first AI,
80. 5:51based on its value system, would conclude that this vulnerable being does not deserve its help and
81. 5:56would ignore it. The second would consider that, since it has the capacity to help, it has a
82. 6:01duty to do so. It is its responsibility to help the most vulnerable. Engineers
83. 6:06build the capabilities of an AI, but it is philosophers who decide how
84. 6:11those capabilities will be used. And now, if you had to entrust the future of humanity
85. 6:16to one of these two AIs, which would you choose? I suddenly think the debate about values
86. 6:21​​seems much less abstract because saying values ​​are human constructs doesn't
87. 6:25mean values ​​have no consequences. Money is a human construct, law
88. 6:30is a human construct, and yet we know very well that a human construct can
89. 6:33produce real consequences. Consider that it is pointless to debate morality or ethics.
90. 6:38Because the universe hasn't given us a precise and concrete answer, it's a serious misconception
91. 6:44. The person who says this doesn't understand what philosophy is. Some think it
92. 6:49's pointless to philosophize since there's no concrete answer in the universe. And since there's
93. 6:53no universal answer to this question, they don't even want us to look for one.
94. 6:58Others think that only the scientific method can offer serious answers. So yes,
95. 7:03the scientific method can establish causal relationships and build
96. 7:06models to explain reality. But we're not going to determine the value of human life by
97. 7:11observing and studying how a cumulus cloud forms. What is a good life? What is
98. 7:15a just society? Do we have duties towards others? What makes an
99. 7:18action just? These aren't questions that can be answered by observing the stars. The question
100. 7:22of how the world works is not the same as the question of how we should live in it. Philosophy
101. 7:26has never claimed to discover a law etched into the universe that could tell us what is good.
102. 7:31It asks us to rationally justify the principles we choose to
103. 7:35value and why. And this task is not a waste of time. Because whether we like it or not,
104. 7:40we already live according to values. We decide daily what is right and what
105. 7:44is wrong, what deserves punishment and what can be forgiven. People behind your back
106. 7:48decide what the state can impose on you, what a company can sacrifice in the name of profit,
107. 7:53and soon, what artificial intelligence will be able to decide for you. So, we can
108. 7:57refuse to philosophize, but we cannot refuse to live according to certain values. We
109. 8:01can only choose to examine them or to accept them. Every time we pass a law,
110. 8:06judge a criminal, educate a child, decide what a company
111. 8:11can do, what a state can mandate, or what a technology can enable, we are doing
112. 8:15philosophy. What we need to do is keep questioning the values
113. 8:19​​we hold. If we don't, we will stop evolving. Until now,
114. 8:23our conceptions of good, evil, justice, and freedom have been primarily embodied
115. 8:27in individuals, laws, and institutions. Tomorrow, they will also be incorporated into
116. 8:32systems capable of making decisions on a scale and at a speed we cannot
117. 8:36match. At that point, dismissing philosophy as mere mental masturbation will no longer be
118. 8:40a contemptuous joke. It will be a particularly dangerous way to abandon
119. 8:44thinking before handing over our power to machines that will be superior to us. The
120. 8:48question is no longer whether we should have values; we already do. And whether
121. 8:52we like it or not, we are obliged to submit to the majority of these values ​​in order to
122. 8:57live in a community. Respecting the lives of others, recognizing their rights, accepting
123. 9:02prohibitions, defining what is just or unjust, permissible or condemnable. The real question
124. 9:07now is which values ​​we will transmit to the intelligences to which we entrust
125. 9:11an increasing share of our power and perhaps soon the destiny of humanity? In other words,
126. 9:15if we do not consciously decide on the values ​​we want to transmit to these machines,
127. 9:19others will necessarily decide for us. And for those who think that philosophy
128. 9:23is useless, abstract, or reserved for a few intellectuals who like to debate in a vacuum,
129. 9:28I have only one piece of advice for them. Pray that those in charge of the
130. 9:31technology of future artificial intelligences are well-intentioned.
131. 9:41So, as we have seen, a powerful entity must necessarily have values. The values
132. 9:46​​chosen will produce radically different results, and this will have
133. 9:52dramatic consequences. Therefore, those who participate in defining these values ​​have a responsibility.
134. 9:56Immense. But where do the values ​​of those who define values ​​come from? If you
135. 10:01've grasped the problem, let's now return to human beings. What determines
136. 10:05the values ​​of those who will determine the values ​​of machines? Why
137. 10:09can two philosophers, faced with the same problems, defend two radically opposed conceptions of good? Is it
138. 10:14their upbringing, their family, their religions, their era, the traumas they experienced,
139. 10:18their temperament, the social environment in which they grew up, their self-interest,
140. 10:23their conception of human nature? No one is born with a complete philosophy in their head.
141. 10:28We even have to be taught to be polite, and we only inherit certain
142. 10:32moral intuitions. We transform them, sometimes add to them, or deepen them. But consequently,
143. 10:37the philosopher who participates in the design of artificial intelligence is not
144. 10:41a disembodied, saintly being. They bring their own history, their prejudices,
145. 10:45their suffering, their beliefs, and their conception of humanity and life. So, if he
146. 10:49is to contribute to determining the principles of an extremely powerful artificial intelligence,
147. 10:54he had better have seriously considered what he sees as good, and it was best
148. 10:58if his conception of good aligned with the interests of the majority. We are all supposed to know that
149. 11:02killing is wrong, but under what circumstances? We know that lying is technically wrong,
150. 11:08but who has never lied? We know that enslaving another human being is wrong,
151. 11:12but since when have we known this? And above all, who taught us? Do you think it was
152. 11:16politics, the law, or justice? If so,
153. 11:21you are mistaken. For centuries, the law permitted slavery. A human being could
154. 11:25be bought and sold, owned, and passed on like property. What seems monstrous to us today
155. 11:30was for a long time perfectly compatible with law and justice. In France, slavery
156. 11:35was reinstated under Bonaparte in 1802 before being definitively abolished on April 27, 1848.
157. 11:43The abolition decree justified this action with moral considerations,
158. 11:47asserting that slavery was an affront to human dignity. Law and morality are not
159. 11:53the same thing. Something can be legal and monstrous. To abolish slavery,
160. 11:59people had to reflect on it and then challenge it. They argued
161. 12:04that freedom belonged to every human being and that no one should have the power to
162. 12:08legitimately own another. These ideas then gained traction. They became
163. 12:12movements. And finally, they became law. People first change their ideas,
164. 12:17then their values, and finally their laws. The law is not what teaches us what
165. 12:21is right, and that is why philosophy is essential. While the law said, "You can
166. 12:26own this human being," abolitionists said, "No, you can't. This is a human being,
167. 12:30they possess dignity, and their freedom cannot be property." That's why a
168. 12:34society that simply says it's legal, therefore it's acceptable, makes a fundamental error.
169. 12:39History shows us precisely the opposite. There have been societies in which slavery
170. 12:44was legal. There have been societies in which women did not have the
171. 12:47same political rights as men. There have been societies in which certain
172. 12:51populations were legally discriminated against, and the mere fact that these practices were enshrined
173. 12:55in law did not make them just. The law can legalize good as well as evil. Before
174. 13:00changing laws, we must change what people consider acceptable. And that's
175. 13:04exactly why philosophy is not a pastime for intellectuals. A
176. 13:08philosophical idea can become a moral conviction. A moral conviction can become a
177. 13:11political movement. A political movement can transform institutions, and institutions can
178. 13:16Ultimately, it transforms the law. That's why philosophy is essential, and why I
179. 13:21want everyone to be able to think for themselves. If a society no longer has
180. 13:25fixed values, everything becomes negotiable. And when everything becomes negotiable, the richest impose their
181. 13:29interests, the most powerful impose their will, and the majority can crush the minority. The state can
182. 13:34justify any constraint. The state can justify any constraint in the name of
183. 13:39security. The individual can justify any behavior in the name of their freedom. And everyone
184. 13:44can ultimately invent their own definition of good and evil. A civilization that functions
185. 13:49this way is doomed. That's all for this video. I hope you enjoyed it and
186. 13:53learned a lot. I know it's a long topic, and unfortunately, I can't
187. 13:57elaborate indefinitely, but I hope it has planted a seed in some people's minds
188. 14:02and that they will do further research. If you're ever
189. 14:06interested in philosophy, I have an app called Métis. It's super simple and
190. 14:10free. Seriously, it's incredibly intuitive. I wish I'd had it when I started learning
191. 14:14philosophy on my own. That's why I made it for you. Feel free to
192. 14:17check it out; the link is in the description. There's also a psychology section. If
193. 14:21you're ever dissatisfied with the work of certain psychologists or therapists,
194. 14:24you'll find plenty of helpful advice. It's free too. And for
195. 14:28those who are already knowledgeable in these areas, feel free to check it out anyway, as there are
196. 14:32modules for those with a good understanding. There are different levels, and even
197. 14:37lots of fun little quizzes. So don't hesitate to give it a try. If you're not already subscribed
198. 14:41to my channel, please do so. And if that's already the case, thank you from the bottom of my heart. Bye!