Les forces des modèles de langage
8:05
Workshop PACEA-Flowers, Nov 15, 2022
4:51:49
PhD Defense - Sébastien Forestier
53:56
PhD defense - Benjamin Clement
46:13
5 жыл бұрын
PhD defense - William Schueller
56:15
Пікірлер
@mlefle
@mlefle 14 күн бұрын
Un de mes collègues a demandé à Chat-GPT4 de faire des exercices de grammaire sur les pronoms compléments. Seuls 20% des exercices ont été utilisables. Pour certains, le vocabulaire ou l'exercice était trop complexe pour le niveau demandé. Pour beaucoup d'entre eux, les phrases étaient mêmes erronées (non utilisation du pronom complément demandé ou phrase tout simplement incorrect). Et pour la très grande majorité, la consigne de l'exercice était alambiquée. Enfin si on demande un grand nombre d'exercices - par exemple 10 exercices de 10 phrases, on se retrouve rapidement avec les mêmes phrases qu'il répète ci et là dans les exercices. Finalement nous n'avons pas gagné de temps à utiliser une IA puisque j'ai dû tout revérifier, corriger et reformuler.
@lailadjoudi3806
@lailadjoudi3806 Ай бұрын
عطاولها 40 دقيقة ..حنا 10 دقايق كيفاش يقدر الواحد يبريزونتي ويشرح هكذا !
@philippwaag2173
@philippwaag2173 Ай бұрын
great content! despite hard to understand.... great content
@ibnaboubacarsana6375
@ibnaboubacarsana6375 Ай бұрын
Je suis d'accord avec vous. Sauf là on vous signifier que a^m×a^n =a^(m+n) est suicidaire. Selon la formule c'est une généralisation , une conjecture. Elle vient en aval dune explication détaillée.
@NoelKossonou-ko1dp
@NoelKossonou-ko1dp 2 ай бұрын
Je voudrais que vous m'aidiez à présenter ma soutenance de doctorat quel son les méthodes conseiller
@mariejoseecastelbasquin6604
@mariejoseecastelbasquin6604 3 ай бұрын
GOD FRUKTY WEST COAST AÉROPORT IMMIGRATION WORLD BANK PROGRAM PLANTES
@nourihsane8886
@nourihsane8886 7 ай бұрын
Bon travail et très bonne présentation
@damoncarr4305
@damoncarr4305 7 ай бұрын
🤷 "Promosm"
@speedmilesgood
@speedmilesgood 8 ай бұрын
Le problème avec les sérious games, c'est souvent le manque de suivit rigoureux scientifiques. J'en ai moi-même participé, une seule fois, en tant que designer pour la gestion et l'entreposage de matières dangereuses en entreprises dont ce n'est pas le but principal, et en tant que Game Designer, c'était à moi de me former à cette manière de procédé, mais l'expertise en moins. Je sais que ce n'est pas l'attrait principal de cette vidéo mais il fallait que je m'exprime sur ce sujet qui pourrait être mieux conçu.
@amadexi
@amadexi 8 ай бұрын
It is obvious that this video was made with GPT-3 lol
@alexandretl
@alexandretl 8 ай бұрын
GPT-3.5 was only used at 8:09 to demonstrate the problem of planification. The (simpler) example just before that was GPT-4. (you can tell the difference by the color of the GPT logo)
@amadexi
@amadexi 8 ай бұрын
@@alexandretl I mean the script of the video, even though I do not understand french
@alexandretl
@alexandretl 8 ай бұрын
@@amadexi so how could you say that ? the script wasn't made with gpt
@amadexi
@amadexi 8 ай бұрын
@@alexandretl There are tools on internet to know if texts has been generated by AI, and which AI is most likely to generate it.
@0sKiDo
@0sKiDo 8 ай бұрын
T'es démasqué !!@@alexandretl Depuis le début tu n'étais qu'un bot :o
@user-es3ep5zs7n
@user-es3ep5zs7n 8 ай бұрын
ربي ايبارك،ماشاء الله
@user-py7mm4jj9k
@user-py7mm4jj9k 10 ай бұрын
Très bon boulot
@boombalibertaire8653
@boombalibertaire8653 11 ай бұрын
Merci !
@khalidkhaer5813
@khalidkhaer5813 Жыл бұрын
Les fiches pdf dons vous avez parlé on les trouve où ?
@raflamansour815
@raflamansour815 2 жыл бұрын
pourriez vous svp me fournir vos slides sans contenus, merci
@abdelkarimel-lourqy9856
@abdelkarimel-lourqy9856 Жыл бұрын
Up
@womenandchildrenprotection2378
@womenandchildrenprotection2378 Жыл бұрын
Bonjour ! Merci beaucoup pour le partage de la vidéo. elle est chouette mais aidez-nous à le faire à notre niveau
@jmguiche9234
@jmguiche9234 2 жыл бұрын
La vidéo plante au bout de 6 minutes. Dommage.
@marionbertin1551
@marionbertin1551 2 жыл бұрын
j'aimerai savoir faire une table de matiere comme la votre pouvez vous m'aider
@hamidarahali2761
@hamidarahali2761 2 жыл бұрын
Tout d abord je vous félicite pour votre thèse ....Quel logiciel utilisez-vous pour la présentation?
@pabloop9891
@pabloop9891 3 жыл бұрын
very impressive!
@CheapDeath96
@CheapDeath96 3 жыл бұрын
WHERE IS THE BEST PART OF THE SONG ?????!?!???
@CheapDeath96
@CheapDeath96 3 жыл бұрын
CANT DO THIS TO ME
@pierre-yvesoudeyer2252
@pierre-yvesoudeyer2252 3 жыл бұрын
test
@paedrufernando2351
@paedrufernando2351 3 жыл бұрын
Dance baby..
@MrAlextorex
@MrAlextorex 4 жыл бұрын
Slides here: pages.isir.upmc.fr/~sigaud/flowers_deepRL.pdf
@alihammoud7539
@alihammoud7539 4 жыл бұрын
can you give an example with data ?
@CosmosMarinerDU
@CosmosMarinerDU 5 жыл бұрын
When the sound is this poor, why bother posting?
@florentweck
@florentweck 5 жыл бұрын
Les élèves font un peu de programmation et de la robotique en technologie collège avant d'arriver au lycée me semble-t-il... Cela même avant les programmes de 2016.
@ramisketcher2069
@ramisketcher2069 5 жыл бұрын
The slides aren't exist!
@Worralius
@Worralius 5 жыл бұрын
Congratulations! :)
@gorkemvids4839
@gorkemvids4839 6 жыл бұрын
Is it Q learnig? Is reward function is exploring itself?
@oscarmulin114
@oscarmulin114 5 жыл бұрын
No, judging by their paper they're using some sort of on-policy learning. I also don't think that Q-learning is appropriate for this task as the actions the agent takes are not discrete. See the paper yourself: arxiv.org/pdf/1708.02190.pdf It looks very interesting and promising.
@catelyn1569
@catelyn1569 6 жыл бұрын
CREEPY WHY ARE ROBOTS SO CREEPY OMG
@prof_shixo
@prof_shixo 6 жыл бұрын
That is interesting, however I can not understand the weird behavior (at 1:00 on the video) as after it had succeeded to touch and move the joystick, it made a noisy exploration (during the 2x speed)! instead of revisiting the interesting state again. It seems that this behavior took more time than needed, do you have any explanation?!
@InriaFlowers
@InriaFlowers 6 жыл бұрын
Hi, thanks for your interest. In fact, at each iteration the robot probabilistically chooses an object to move, based on its learning progress, e.g. 30% chance to move the hand, 20% to move the left joystick etc. and then chooses a random goal for this object, e.g. move the hand to the left etc. With this exploration mechanism, a new interesting trajectory of an object (e.g. moving the toy robot at 1:00) might not be reused directly after discovering it, but later, when the robot will choose a goal for this object. Also, the learning progress to move the joystick, estimated by the robot itself, will slowly increase after this discovery, but we might not see it behaviorally before a few dozens iterations. Sébastien
@prof_shixo
@prof_shixo 6 жыл бұрын
Thanks for the clarification.
@rodrigo7089
@rodrigo7089 6 жыл бұрын
Extraordinary!!!!, Excellent job!!!
@alirose1470
@alirose1470 6 жыл бұрын
The little shoes I can't
@RoyceBarber
@RoyceBarber 6 жыл бұрын
Superb. It's tragic almost no affordable robots in 2018 are bothering to be this stable.
@RoyceBarber
@RoyceBarber 6 жыл бұрын
That was impressive, why aren't you showing that kind of robot in newer videos? That was eight years ago.
@RoyceBarber
@RoyceBarber 6 жыл бұрын
Informative, but come on, let's see Poppy RUN! <3
@RoyceBarber
@RoyceBarber 6 жыл бұрын
I love Poppy, but I know it can move much faster than that, and it has legs ya know! Hook it up to some motion capture. Even if Poppy has to use an old ladys walker to get around, let's see something new in 2018. No more being careful to only say good things about Poppy. Let's get down and dirty, get some dirt on Poppy, make it faster and stronger than ever before. Stronger stepper motors, and a big battery on wheels behind it, so we can get some mobility without all that weight. Let's make 2018 when old man Poppy finally shows us he's becoming more humanoid and not just a rag doll!
@RoyceBarber
@RoyceBarber 6 жыл бұрын
very clever robot Inception LOL
@yashwanthreddy7654
@yashwanthreddy7654 6 жыл бұрын
What is the price u had spent making this
@chrisp4535
@chrisp4535 6 жыл бұрын
yashwanth bandi with all of the actuators and parts its nearly 6000$
@xuananhnguyen9117
@xuananhnguyen9117 6 жыл бұрын
Can you share the publication corresponded to this video?
@SixtySecondYoga
@SixtySecondYoga 6 жыл бұрын
I’m Poppy!
@did3d523
@did3d523 6 жыл бұрын
muppet
@MonaJalal
@MonaJalal 7 жыл бұрын
What is the person wearing on his right hand? Do you have a blog or paper on this you could share?
@xuananhnguyen9117
@xuananhnguyen9117 6 жыл бұрын
An optical tracking device: www.iotracker.com/indexdaed.html?q=optical_tracking
@catherinesimon4426
@catherinesimon4426 7 жыл бұрын
Could change our life if in good benevolent universal use. I am glad you chose Education.
@catherinesimon4426
@catherinesimon4426 7 жыл бұрын
Extraordinary !
@micoerfe571
@micoerfe571 7 жыл бұрын
OMG POPPY IS A ROBOT!
@rem3705
@rem3705 6 жыл бұрын
I KNEW IT!!!
@DanFrederiksen
@DanFrederiksen 7 жыл бұрын
I wonder how useful it is in practice and if it's a dead end approach.
@InriaFlowers
@InriaFlowers 7 жыл бұрын
Video description updated.
7 жыл бұрын
Brilliant :)
@Hasan...
@Hasan... 7 жыл бұрын
This is incredible! May I know the sensors used in this robot? Any more information will also help.. like: which microcontroller? which programming language? PID control? Anyway, Excellent demonstration!
@RoyceBarber
@RoyceBarber 6 жыл бұрын
Robots were better eight years ago lol.
@keghnfeem4154
@keghnfeem4154 7 жыл бұрын
groups.google.com/forum/#!topic/artificial-general-intelligence/UVUZ93Zep6Y groups.google.com/forum/#!topic/artificial-general-intelligence/jWdzPaxYHmU groups.google.com/forum/#!forum/artificial-general-intelligence