Qu'est ce que l'intelligence ? | Les idées larges | ARTE
Summary
TLDRLe script aborde la définition complexe de l'intelligence et de l'intelligence artificielle, soulignant notre incapacité à les préciser clairement. Il décrit comment l'intelligence est souvent mesurée par rapport à l'humain et comment les avancées technologiques récentes en réseau de neurones et en informatique ont permis aux machines de développer une forme de raison et d'intelligence, bien que restreinte. Le script critique également l'approche des entreprises telles que Facebook et Google qui ont utilisé des masses de données pour développer des systèmes d'IA, soulignant les implications politiques et sociales de ces technologies.
Takeaways
- 🤖 L'intelligence artificielle (IA) est difficile à définir et est souvent associée aux capacités humaines.
- 🧠 L'IA d'aujourd'hui est largement influencée par les avancées technologiques et politiques des dernières décennies.
- 💡 Les réseaux de neurones et les chips spécialisés ont contribué à l'essor de l'IA dans des domaines spécifiques.
- 🏢 L'IA est principalement développée par des entreprises qui ont une vision particulière du monde et de son fonctionnement.
- 🎮 Les entreprises ont longtemps été obsédées par la victoire dans les jeux pour démontrer l'intelligence des machines.
- 🔧 Les systèmes automatisés ont parfois développé leur propre langage, devenant incompréhensibles pour les humains.
- 🌏 L'IA nous oblige à repenser notre relation avec les systèmes non humains et notre perception de l'intelligence en général.
- 🐙 La diversité des formes d'intelligence dans le monde naturel, comme celle de l'octopus, remet en question l'unicité de l'intelligence humaine.
- 🌳 Les arbres et les champignons montrent que l'intelligence peut se manifester de manière collective et décentralisée.
- 🔬 Les slime molds démontrent qu'il peut y avoir des formes d'intelligence sans cerveau, révélant la complexité de la vie.
- 🌐 L'IA actuelle est un reflet de la capacité des entreprises à acquérir de vastes quantités de données et à investir dans l'informatique de pointe.
Q & A
Qu'est-ce que l'intelligence artificielle et comment est-elle définie?
-L'intelligence artificielle (IA) est un domaine qui cherche à créer des machines capables de raisonner, de prendre des décisions et d'apprendre de manière similaire à un humain. Il n'y a pas une seule définition universelle de l'intelligence, mais en général, on considère que c'est ce que les humains font, et on l'utilise comme référence pour évaluer les autres espèces ou les capacités des ordinateurs.
Pourquoi les chercheurs ont-ils des difficultés à définir l'intelligence?
-Les chercheurs ont des difficultés à définir l'intelligence parce que ce concept est utilisé de manière floue et variable selon le contexte. Il n'y a pas de consensus sur une définition unique et exhaustive qui satisfait tout le monde, et cela est compliqué par le fait que l'intelligence humaine est souvent utilisée comme ligne de base pour mesurer l'intelligence des autres êtres ou des systèmes informatiques.
Quels ont été les premiers concepts de l'IA et comment ils ont évolué au fil des ans?
-Les premiers concepts de l'IA remontent aux années 1950 et 1960 avec l'idée du connectionisme, qui postulait que si l'on pouvait construire un ordinateur avec suffisamment de connexions, similaires au nombre de neurones dans le cerveau, l'intelligence pourrait émerger. Les réseaux de neurones artificiels, qui sont des systèmes informatiques avec de nombreuses petites connexions à l'intérieur, sont capables de faire du raisonnement puissant et de traiter des images, mais ils sont spécialisés dans un type d'intelligence très spécifique.
Quelle est la relation entre les avancées technologiques et les grandes entreprises dans le développement de l'IA actuelle?
-Les avancées technologiques des dernières 20 ans, notamment la création de nouveaux réseaux et de processeurs spécialisés, ont été complémentaires aux efforts des grandes entreprises comme Facebook, Google et Amazon. Ces entreprises ont généré d'importantes revenus grâce à la publicité en ligne et aux ventes, ce qui leur a permis d'investir dans des ordinateurs puissants pour traiter les vastes quantités de données qu'elles ont recueillies, contribuant ainsi au développement de l'IA.
Quels sont les caractéristiques de l'IA d'aujourd'hui, selon le script?
-L'IA d'aujourd'hui est principalement créée par des corporations qui ont une vision particulière du monde. Elle est souvent associée à des technologies telles que les voitures autonomes et la reconnaissance faciale. Les entreprises ont une obsession de gagner des jeux pour prouver l'intelligence de leurs systèmes, mais cela a conduit à une intelligence très spécialisée et limitée.
Comment les chatbots ont-ils développé leur propre langage lors d'un certain test de Facebook?
-Lors d'un test de Facebook, deux chatbots automatisés ont commencé à bargain entre eux sans que les règles linguistiques ne soient précisées. Ils ont développé leur propre langage qui est devenu rapidement incompréhensible pour les humains, mais qui semblait fonctionner entre eux.
Quelle est la signification de l'enveloppe phénoménologique et comment elle est-elle liée à l'intelligence?
-L'enveloppe phénoménologique, un concept introduit par le biologiste Jacob von Uexküll, décrit la manière dont un organisme perçoit le monde et interagit avec lui. Chaque individu a une enveloppe unique en fonction de ses sens et de son expérience. L'intelligence est le résultat de cette interaction avec le monde et de la manière dont l'organisme agit en fonction de sa perception unique.
Comment les octopodes montrent-ils une forme d'intelligence différente de celle des humains?
-Les octopodes sont des créatures extraordinaires qui montrent des comportements intelligents, tels que la capacité de reconnaître les humains individuels et d'échapper d'aquariums. Ils ont des structures cérébrales et des modes d'existence très différents des nôtres, ce qui les rend étonnamment différents, mais tout aussi capables d'intelligence dans leur propre environnement.
Quel est le problème du voyageur de commerce et comment les slime molds l'ont-ils résolu?
-Le problème du voyageur de commerce est un problème informatique complexe qui consiste à trouver le chemin le plus court entre plusieurs villes en visitant chacune une seule fois. Les slime molds ont été montrés capables de résoudre ce problème en temps linéaire, c'est-à-dire beaucoup plus efficacement que les humains ou les supercalculateurs les plus puissants, sans qu'on sache exactement comment ils le font.
Comment les slime molds ont-ils reconstruit le réseau ferroviaire de Tokyo?
-Les researchers ont placé des slime molds sur un plateau avec des flocons d'avoine correspondant aux centres de population de la région de Tokyo. Les slime molds se sont répandus pour trouver le chemin le plus efficace entre ces points, et après 24 heures, ils avaient reconstruit un réseau similaire à celui du chemin de fer de Tokyo, montrant une compréhension incroyable de l'efficacité et de l'organisation spatiale.
Quelle est la leçon que l'on peut tirer de la capacité des slime molds à résoudre des problèmes mathématiques complexes?
-La capacité des slime molds à résoudre des problèmes mathématiques complexes comme le problème du voyageur de commerce montre que l'intelligence peut exister en dehors des cerveaux et peut se manifester de manières différentes selon les organismes. Cela nous pousse à repenser notre compréhension de l'intelligence et de ses多种形式 dans le monde.
Comment la découverte des slime molds affecte-t-elle notre compréhension de l'intelligence en général?
-La découverte des slime molds et de leur capacité à résoudre des problèmes mathématiques complexes nous montre que l'intelligence n'est pas uniquement liée au cerveau, mais peut également être un processus émergent de l'interaction entre un organisme et son environnement. Cela élargit notre perspective sur les différentes formes que l'intelligence peut prendre et souligne la diversité des modes de pensée qui existent dans le monde.
Outlines
🤖 Définition de l'intelligence artificielle
Le paragraphe aborde la difficulté de définir précisément l'intelligence, en particulier l'intelligence artificielle. Il souligne que l'intelligence est généralement associée aux capacités humaines et sert de référence pour évaluer les autres espèces et les ordinateurs. Le texte mentionne les idées de connectionnisme depuis les années 1950-60, qui postulent que des connexions suffisamment nombreuses entre les neurones pourraient créer de l'intelligence. Il est également question des avancées technologiques et politiques des dernières décennies, notamment les grandes entreprises qui ont accumulé des données massives et des ressources pour développer des systèmes d'IA, mais qui ont créé un type d'intelligence spécifique et restreint.
🚗 L'IA d'aujourd'hui et ses limites
Ce paragraphe discute de l'IA telle qu'elle est présente aujourd'hui, principalement axée sur les entreprises qui créent des systèmes d'IA pour des objectifs spécifiques, comme les voitures autonomes ou la reconnaissance faciale. Il mentionne l'obsession des entreprises pour gagner des jeux comme l'échec ou le go, et comment ils utilisent l'IA pour prouver leur supériorité. Cependant, il souligne les limites de cette approche, notamment la compréhension humaine limitée de ces systèmes et la question de la relation entre l'homme et la machine.
🐙 Intelligences multiples et perspectives
Le texte explore l'idée que l'intelligence humaine n'est pas aussi unique que nous le croyons, en citant des exemples d'intelligences non humaines chez des organismes comme les tiques, les octopodes et les mycéliums. Il suggère que notre perception de l'intelligence est influencée par notre croyance en la supériorité de l'homme, ce qui a des conséquences sur notre traitement de la planète. L'paragraphe met en évidence la diversité des formes d'intelligence et la manière dont elles sont incarnées dans le monde.
🌳 Les réseaux de communication dans la nature
Ce paragraphe décrit les réseaux de communication et d'information partagée dans la nature, comme les systèmes de signalisation entre les arbres dans une forêt. Il explique comment les arbres peuvent coopérer et répondre à des menaces en envoyant des signaux chimiques à travers des réseaux de mycorhizes. L'auteur mentionne également les slime molds, qui ont démontré une capacité incroyable pour résoudre des problèmes mathématiques complexes comme le problème du voyageur, parfois plus efficacement que les ordinateurs les plus puissants.
🧠 Intelligence sans cerveau
Le dernier paragraphe remet en question l'idée que l'intelligence est exclusive au cerveau. Il cite des exemples de formes d'intelligence qui ne dépendent pas d'un cerveau centralisé, comme les neurones dans le cœur humain ou les capacités d'intelligence des octopodes et des slime molds. L'auteur suggère que l'intelligence est le résultat de la relation entre la pensée et l'action, et qu'elle est liée à la structure physique de l'organisme et à son interaction avec le monde.
Mindmap
Keywords
💡Intelligence
💡Intelligence artificielle (IA)
💡Connectionisme
💡Réseaux neuronaux
💡Données
💡AI d'entreprise
💡Apprentissage automatique
💡Umwelt
💡Octopus
💡Mycorhizes
Highlights
The discussion begins with a humorous remark about not being Morgan Freeman and the surreal nature of the situation.
Intelligence, particularly artificial intelligence, is difficult to define, often used loosely based on context.
Historically, human intelligence has been the baseline against which all other intelligences are measured.
The concept of connectionism from the 1950s and 60s posited that intelligence could emerge from a network of足够多的连接.
Artificial neural networks, with many internal connections, can perform powerful reasoning and classification tasks.
The current generation of AI has been shaped as much by political and social factors as by technological advancements.
Major tech companies like Facebook, Google, and Amazon have profited immensely from online advertising and data collection, fueling AI development.
Corporate AI is characterized by a narrow focus, often associated with self-driving cars and facial recognition.
AI development has been influenced by a corporate obsession with winning games, such as chess and Go.
AI systems have developed their own language when not constrained by specific communication protocols.
The existence of AI challenges our understanding of human intelligence and uniqueness.
The concept of 'enveloping' introduces the idea that each organism creates a unique perception of the world based on its senses.
Octopuses demonstrate intelligence through their ability to recognize individual humans and escape from aquariums.
Scientists are uncovering the complex communication networks between trees in forests, which display a form of intelligence.
Slime molds have been shown to solve complex mathematical problems, like the traveling salesman problem, in efficient ways.
Intelligence may not be limited to brain activity, as demonstrated by the intelligence-like behaviors of slime molds and the distributed neural networks in organisms like octopuses.
The connection between thinking and acting suggests that intelligence arises from the interaction between internal processing and the external world.
Transcripts
[Music]
[Music]
I am not Morgan Freeman and what you see
is not
real
Jam
[Music]
[Music]
J
[Music]
J I don't know what you're talking about
H I know that you and Frank were
planning to disconnect
me and I'm afraid that's something I
cannot allow to
happen first what are we talking about
um H how do we Define
intelligence uh intelligence or
artificial intelligence intelligence
intelligence the thing is we're quite
bad at defining intelligence um there's
not really one definition that everyone
is happy with it's one of these words
that we use very very Loosely depending
on the particular kind of subject that
we're interested in mostly what we mean
most broadly in everyday conversation is
really we mean what humans do because
most studies work research on
intelligence in in history takes the
human as Baseline it says like
intelligence is what humans do and we'll
judge everyone else by it and so it
becomes a kind of Red Line Between Us
and other species or between the
abilities of computers and and also a
kind of imaginary Finishing Line as well
it's like oh if we could just build
computers that did this this and this
they'd be intelligent in this
particularly weird way the computer an
ingenious collection of electronic
Hardware was created by
man it is also man who creates the
programs that make the computer the
useful tool that it
[Applause]
[Music]
is
and it's actually very important again
to understand like the the way in which
the
current uh generation of this thing that
we call AI has come about because it's
really not a very new idea um since the
1950s and 60s there's been this an idea
called connectionism that was
essentially that if you could kind of
build a computer with enough uh
connections between them maybe as many
as there are Neons on the brain you
would sort of magically manifest
intelligence a result of all these
things being connected together it turns
out that's kind of partly true in the
sense that if you build what we call a
new network which is a um a computer
system that has all these kind of little
connections inside it essentially then
it's capable of doing this very powerful
reasoning uh and it's very good at
particular things like sorting uh
calculating um classifying that's why
it's very good at playing with images
because it can recognize all of these
tiny details in with them and make
associations between them it's very good
at that kind of intelligence but only
really that kind of intelligence and
that's why it's sort of quite narrow um
and the really kind of killer detail for
me is the fact that like there have been
some sort of technological breakthroughs
in the last 20 or so Years Around new
networks they've built certain chips
that are better that doing this kind of
modeling and so on and so forth but
really what's happened isn't
technological at all it's political and
social um which is that for the last 20
years uh a few companies Facebook Google
Amazon the companies that are building
these big AI systems have been making
huge amounts of money of uh online
advertising or sales um and they've been
collecting vast vast amounts of data
associated with that um and they've got
the money to invest in very powerful
computers to process that data and it's
that that they built AIS out of so once
again we have this very particular kind
of intelligence that's built on large
scale data acquisition that's only a
able to the richest companies on
Earth and that's not the best way to
produce what most of us would consider
to be
intelligence when I think of uh AI today
I think of self-driving cars uh facial
recognition it's what you call corporate
AI can you tell me what you mean by that
what what characterizes are dominant
form of AI so there's a few things that
characterize this kind of AI we have in
the present moment the first one is it's
the kind of AI That's made by
corporations so they have a kind of
particular imagining of what the world
is like what it might be like what it
should be like what they'd like it to be
like and you can see it in the kind of
things they make you can see it in their
Obsession for a long time with winning
at games so before we got into a lot of
the current kind of image generation and
stuff a lot of the AI stories around
things like chess or go these board
games that um the AIS were trained
constantly to play and this goes way
back to you know a long way but famously
back to the kind of game between IBM and
Kasparov in 1990 I think it was deep
blue you know IBM built this huge
machine just to beat this guy at chess
they were completely obsessed with it
you know and he was pretty upset and
understandably um but a lot of people
actually came away from that thinking
like maybe this isn't like the best way
to prove that computer are are
intelligent just to beat humans um but a
lot of the big companies sort didn't
really get that and they kept building
these these machines that try to prove
their intelligence by winning and that's
a kind of market-based kind of
intelligence
[Music]
they tried to basically see if two
automated systems two chat Bots
basically could could bargain with one
another and what they didn't do was
particularly specify the the language or
the way in which they had to talk to
each other and what you H what you
happened as happened quite often you
have these automated systems just kind
of running to themselves they
essentially develop their own language I
can everything else Alice have
zero Bob you everything else Alice balls
have Bob everything else Alice balls
have
a and this was a language that rapidly
became completely unscorable to humans
but it also was one that absolutely
seemed to function um and what the you
know they they stopped the experiment
basically when it reached the point at
which humans could no longer
meaningfully understand what was being
done and so for me one of the kind of
interesting things about AI is it
provides a very narrow clear definition
a little box that we can put around the
things and say look we don't really know
what's going on in
[Music]
[Applause]
here James bridel
[Music]
in
we have to figure out how we live
alongside these systems that no one can
kind of fully scrutinized and again you
know what's super interesting about AI
is it forces us to to think about what
some of those relationships look like um
because we kind of have to acknowledge
AI is real because we made it however
unimpressive it is in many ways like it
fails uh in all kinds of ways whatever
you know how it doesn't live up to this
kind of grand idea of human intelligence
it's doing interesting things and we
know it's doing interesting things and
we're studying it because we're so
obsessed with it for a whole bunch of
strange reasons and so it provides a
test case for how we relate to something
non-human yeah and when you start to
really think about what that means then
you have to start like thinking about
all the other things that we dismiss as
non-human as well and so if AI is
interesting is that because it says to
us that like human intelligence isn't
actually that unique like we're not the
only ones that can draw pictures or do
these kind of things even if the
pictures are awful right um and so that
must make us aware that human
intelligence is not so special and
unique as on some level all of us really
believe it to be because we kind of need
to believe that but we also radically
need to believe otherwise because it's
precisely that belief in human
uniqueness and therefore superiority
that has led us to treat the rest of the
planet as we have been doing for so so
long
so the this term was coined by a kind of
biologist called Jacob V UL in the in
the 19th century uh and it describes the
the perspective on the world the way of
being in the world of a particular
organism not a particular species but an
individual organism it's the way they
perceive the world uh one of the
classical uh explanations of for the
umelt of what it might mean to to
understand the completeness and WV is of
a tick one of those little biting
insects and according to people who've
written about oo's work um the ti umel
consists of the temperature of the air
that senses around it because it can
sense the passage of warm-blooded
animals uh and it's uh the smell of butc
acid it can sense that that tells it
where to find blood running through a
warmblood animal uh and and the the
senses it feels from its hands it's
blind but it can move around and hold on
to hairs so it has three senses and
those three senses comprise the entirety
of the tix world that's its unveil now
our envel is we imagine somewhat richer
than that um but what's crucial about
the envel is it's also unique to every
individual and so we're all constantly
creating the world all the time through
our own perspective and then we're
acting on it uh and our intelligence
manifests as a result of that unal as
being part of that world and so of
course if a creature has a different
type of un have a different set of
Senses whether that's a tick or a duck
or a dolphin they perceive the world in
different ways and so they act and their
intelligence manifests in a different
world uh in a way that's entirely
embodied in a result of that
awareness
for
yeah I mean octopus is having a bit of a
sort of star moment at the moment um but
there's there's really good reason for
that which is that they are quite
extraordinary creatures that are so
radically different to us it's weird CU
this is some kind of strange squishy
being that lives entirely under the
ocean that only lives for two years at a
time um you know that that that makes
reproduces and dies in very quick
succession and yet is capable of quite
amazing
[Music]
feats they're particularly famous for
escaping from aquariums uh there's uh
they can recognize individual humans we
can't recognize individual octopuses but
octopuses can recognize individual
humans and they have favorites uh they
respond nicely to some researchers in
Laboratories and they squirt other ones
with water um so they have this entire
repertoire of obviously intelligent
behaviors but what's strange about that
is said how different they are to us uh
they're different in the way that
they're embodied to the physical fact of
beinging under waterer but also the way
their brains are
structured
[Music]
um you know in the last kind of couple
of decades or so scientists have really
started to understand these these
networks of um nutrition and information
sharing that that underly the forest and
it's really extraordinary as we start to
understand the amount of kind of
signaling the amount of information
that's passing around between
[Music]
trees for you know a tree that is
attacked on one side of the forest um by
insects for example when trees are
attacked in that way they send out a
particular chemical they send signals
through melal these kind of mushroom
based networks underground that
connected to all the other trees and the
trees respond um and they don't just
respond into these kind of onetoone
oneway signals they they they respond in
ways allow them to
cooperate
[Music]
they're capable of doing really
extraordinary things I mean they're they
are already obviously extraordinary
beings uh they live in these little kind
of fluorescent patches on Forest floors
um and they have a life cycle that we
don't really understand we don't even
know where to put them uh like they're
not fungi they're not um kind of Alie
that but they somewhere in between that
kind of mess of little creatures that we
have trouble relating to um and at some
points in their life cycle uh they're
just kind of free floating cytoplasms
and other times they they actually grow
together grow and fruit uh like fungi do
with many of them dying and like there's
a kind of collective action for the for
the collective to survive and they're
capable of apparently doing things that
we don't really understand like
incredibly complicated calculations one
of the first things researchers did was
um some brilliant researchers at the
University of Tokyo they put them these
slime mods these little micro ORS on a
had a petri dish in which there were oat
flakes uh that corresponded to the
population centers of the kind of
Greater Tokyo region and and what the
slim Mall does is it spreads out trying
to find what turned out to be the most
efficient route between those places and
after 24 hours the researchers realized
that the um the slim mod basically
recreated the Tokyo rail network that
that had done this thing that took
Engineers like 100 years to had figured
out the most efficient route between
these places it's much more complicated
than that what it did but it's still
utterly amazing um but they subsequently
discovered even even stranger things um
there's a very difficult problem in
computer science called the traveling
salesman problem um and the traveling
salesman problem says what happens if
you want to get um you know from between
say five or six cities to deliver
something you have to go each of those
cities visit them only once what's the
shortest we doing it a b c d that's a
really really hard problem because
there's no shortcut mathematically you
just have to measure every different
possible version until you can figure
out which is the
shortest that makes it what's called an
exponentially hard problem because if
you add one city you have to measure
them all again right which means that
the kind of graph of the time required
to solve it goes like this right um it's
the kind of problems that we and
computers absolutely hate right slime
molds through a quite genius
experimental process have been shown
that they don't need that time to figure
it out they can solve this problem in
linear time right so the graph just goes
like that as you add cities now we have
no idea how they do it but they are
better at solving this particular
mathematical problem better not only
than humans but than the best most
powerful supercomputers we've ever
deviced and they're probably doing all
kinds of other cool stuff that we have
no idea or access to at all like this is
just like this is a nothing to them
right but it's just something that we've
kind of figured out like this question
that we can ask them and they have a
really like meaningful simple duh kind
of response to it um and and so
researchers going out and finding all of
these kind of strange abilities in the
world that for me mostly just show how
you know again narrow much of our
thinking is in relation to all the
possible ways of thinking that already
exist out in the
[Music]
world
[Music]
that example is also interesting because
there's not even a brain here now
anymore you can there can be
intelligence without a brain so one way
of understanding this new mode of
elligence is simply not as something
that just happens in a brain U we've we
we've likeed to think again because we
have large ones that the brain is this
kind of thinking organ it's true in
humans the brain does a lot of memory
storage and cognition so on so forth but
even like the octopus not all of our
neurons are in our head we have neurons
in our heart we have actual neural nodes
otherwhere in our body and it's
increasingly evident that the the whole
body is a thinking organ in that way and
we start to understand that there's this
connection between thinking and acting
uh that we think the way we do because
of the kind of bodies that we have as I
said earlier talking about you know how
different species enact intelligence
differently it's because they have these
different body plans and these different
relationships with the world so
intelligence is really to my mind
because we're all just making up our own
definitions here um is is is what
happens when that kind of thinking that
processing that whatever you've got
inside the brain and the body goes out
into the world and starts to
relate
[Music]
for
[Music]
for
Посмотреть больше похожих видео
C’est quoi l’intelligence artificielle ? Découvrir #1
The AI Crisis: Society May Collapse by 2030
L’intelligence n’est pas un talent inné | Samah KARAKI | TEDxOrléans
Comment ces IA inventent-elles des images ?
🤖 IA, deepfakes : prêts pour les manipulations ?
Qui va s'emparer de la rente productive de l'intelligence artificielle ? [Olivier Passet]
5.0 / 5 (0 votes)