[AI4V] FR - Interview Imed BEN NASR — Transcript

Full transcript

1. 0:11How is AI changing the world of work?
2. 0:14I think it’s happening right now.
3. 0:16We are clearly seeing the emergence
4. 0:19of certain uses.
5. 0:20We are witnessing this evolution, the emergence of jobs
6. 0:23and responsibilities linked to AI.
7. 0:27I think that most of the changes now concern
8. 0:31the management of AI-related projects,
9. 0:32because most companies haven’t necessarily implemented AI.
10. 0:35I’m not talking about ChatGPT, etc. I’m talking about the in-house
11. 0:39integration of an LLM linked to a company’s knowledge base.
12. 0:43Most large companies are doing it,
13. 0:45and I think there will be a similar trend for all companies,
14. 0:48for reasons of security, efficiency,
15. 0:52and data confidentiality.
16. 0:55This means that the integration of AI will take a little time
17. 1:00before it can become effective and add
18. 1:03real value to a company’s strategic direction.
19. 1:07In my opinion, companies should not rush into integrating AI,
20. 1:11since there are three main points to consider.
21. 1:14The first is getting the entire company to embrace AI,
22. 1:17i.e. ensuring employees are familiar with AI.
23. 1:21Creating a sense of familiarity and excitement around AI
24. 1:27helps to limit resistance
25. 1:30to the integration and adoption of AI within a company.
26. 1:33That’s point one.
27. 1:34The second is to identify the departments,
28. 1:37given the company’s strategic direction,
29. 1:40which are most likely to benefit from the contribution of AI.
30. 1:44This could be the marketing department,
31. 1:46for example, by implementing an AI-based chatbot,
32. 1:49if the company is very marketing-oriented.
33. 1:52It could be the finance department,
34. 1:53if the finance department is the core of a company’s added value.
35. 1:57Or it could even be supply chain or production.
36. 2:00These are generally strategic choices
37. 2:02made by senior management. And on the basis of
38. 2:06these strategic choices,
39. 2:07generally speaking, change management will follow,
40. 2:10as well as a thorough audit to identify,
41. 2:15given the technical aspects in place, how the integration of AI
42. 2:19could bring added value
43. 2:21without necessarily creating additional technical constraints.
44. 2:24And the third point is the technical side.
45. 2:28Taking into account the platform, hardware,
46. 2:30and systems already in place, you need to identify the technologies
47. 2:33that are most likely to capitalise on the company’s current strengths.
48. 2:37As things currently stand,
49. 2:39I think the future of AI lies in making it accountable.
50. 2:46There is a lot of talk these days about responsible AI.
51. 2:49What we are seeing is that
52. 2:52the rules of
53. 2:53security, non-discrimination
54. 2:56and a certain form of ethics and morality associated with AI
55. 2:59are not necessarily taken into account in current algorithms
56. 3:03and in the data on which the algorithms are trained.
57. 3:06That’s why many companies,
58. 3:09particularly those
59. 3:12offering AI solutions,
60. 3:15are looking at how to take all these aspects into account
61. 3:19when training algorithms,
62. 3:23so that the AI does not generate the same biases
63. 3:27that were present in the data on which it was trained.
64. 3:30For example, AI-based security solutions have been found
65. 3:33to recreate discrimination biases
66. 3:38because the data they are trained on
67. 3:40is biased in terms of social, racial, or other aspects.
68. 3:44One of the challenges today is to ensure that,
69. 3:46for example, if I have a company
70. 3:48with a customer database and I need to set up
71. 3:53a marketing automation system
72. 3:54to reach my customers, how can I ensure
73. 3:57that such biases, which are not observable from the outset,
74. 4:01do not impact the effectiveness of my marketing practices
75. 4:05and, above all, do not result in practices that are unethical?
76. 4:10In fact,
77. 4:11ensuring data quality from the outset,
78. 4:16and here we really stress the importance of quality data
79. 4:18for training AI, is a vector of efficiency.
80. 4:22Another point is how to respect consumer privacy
81. 4:26when we’re looking to integrate
82. 4:29data from an algorithm
83. 4:32but we don’t know how the data is going to be subsequently used.
84. 4:36That’s why we speak of a ‘Human Centric Approach’.
85. 4:39In other words, today,
86. 4:40we are trying to create AI that is user-oriented.
87. 4:45This means trying to establish certain criteria,
88. 4:51certain constraints, to ensure that
89. 4:55the processing of the data of an average person,
90. 4:59e.g. a consumer
91. 5:04is appropriately managed.
92. 5:06In other words, we must respect their private life
93. 5:09and sensitive aspects
94. 5:12such as race, age, gender, etc.
95. 5:15And to ensure that the way in which
96. 5:17this data is used for training AI
97. 5:21corresponds to certain moral and ethical values.
98. 5:25Because common cases arise, particularly for companies
99. 5:28that are committed to a CSR approach.
100. 5:32And this also means using responsible AI systems.
101. 5:35I think is that the philosophy
102. 5:39of an organsiation should be clear on all these different aspects,
103. 5:43including its approach to such technology.