[AI4V] FR - 312 — Transcript

Full transcript

1. 0:11Let’s start at the beginning: consider a brain
2. 0:14and imagine that it could also be the brain of a machine.
3. 0:18Incidentally, we talk more and more about artificial intelligence, and artificial neurons,
4. 0:24but strangely enough, we’re not yet talking about artificial brains.
5. 0:29As we saw earlier, our mental power derives
6. 0:32from the exchange of information between an environment and a brain.
7. 0:37To function, this system must be able to process incoming information.
8. 0:41This processing, our primary cognitive function,
9. 0:44is therefore essential.
10. 0:47Through our senses, we detect the different modulations
11. 0:51and variations of our environment through changes in colour,
12. 0:55sound, speed, and many other parameters.
13. 1:00When this processing is passive, we speak of ‘subliminal perception’
14. 1:04and when it is active, we speak of ‘observation’.
15. 1:10Here’s an example to illustrate the difference:
16. 1:13Subliminal perception
17. 1:15is when you’re working
18. 1:17on your computer and you subconsciously hear
19. 1:19the background music in a café, without actively paying attention.
20. 1:23Observation is when you listen carefully to that same music,
21. 1:28concentrating on the instruments, the rhythm and the lyrics.
22. 1:33But is a machine also capable of processing information from its environment?
23. 1:37The answer is ‘yes’!
24. 1:39Take the example of the robot vacuum cleaner:
25. 1:41it uses many sensors to map its environment.
26. 1:46It’s no coincidence that we talk about ‘computer vision’.
27. 1:50So, how does it work?
28. 1:52In its simplest form, when you submit a photo to ChatGPT
29. 1:56for analysis, it uses a model known as computer vision.
30. 2:02It sees what’s on the image.
31. 2:04Now, whether this is a subliminal perception or an observation,
32. 2:09I admit that I can’t answer that question.
33. 2:13Today, we talk more and more about multimodal
34. 2:15models: AI tools capable of analysing not only text,
35. 2:20but also images, sounds and videos.
36. 2:23Basically, these are models
37. 2:24that integrate several senses into a single system.
38. 2:28But doesn't that sound like something the brain would do?