Cheating or Learning? Walking the AI tightrope in education | Erik Winerö | TEDxGöteborg
Summary
TLDRDans ce discours passionnant, l'orateur aborde la question de l'intelligence artificielle (IA) dans l'éducation en utilisant la métaphore d'un conte où la génération de l'IA pourrait être à la fois le dragon et le chevalier. En tant que professeur et chercheur en éducation, il explore comment l'IA générative, capable de créer des images, des sons et du texte, affecte le travail des enseignants et leur perception de l'apprentissage des étudiants. Il soulève les craintes des enseignants concernant le triche grâce à l'IA, tout en soulignant la complexité de définir ce qui constitue un triche. À travers des analogies sportives et des exemples concrets, il met en lumière les risques de l'automatisation des processus d'apprentissage, qui peuvent entraîner une illusion d'apprentissage. Il plaide pour une utilisation réfléchie de l'IA qui favorise le processus d'apprentissage plutôt que simplement la production de résultats, et encourage à utiliser l'IA comme un coach pour aider à surmonter les obstacles, plutôt que de prendre des raccourcis qui pourraient nuire à l'apprentissage véritable.
Takeaways
- 📖 L'introduction d'une histoire est un excellent moyen d'attirer l'attention du public, en commençant par une image forte qui représente un groupe ou une communauté.
- 🐉 Il est difficile de déterminer si l'IA est le dragon ou le chevalier dans l'histoire de l'humanité, car son rôle est encore en évolution.
- 👨🏫 L'IA générative, capable de créer des images, des sons et du texte, affecte la manière dont les enseignants perçoivent et évaluent l'apprentissage des étudiants.
- 🚨 Les enseignants craignent que l'IA ne soit utilisée pour tricher, mais la définition du triche est complexe et dépend du contexte.
- 🏆 L'anthropologue utilise la métaphore du saut à la perche pour illustrer comment le progrès technologique peut améliorer les performances, mais aussi les débats sur le triche.
- 🆕 L'adoption de nouvelles technologies, comme les perches flexibles, est initialement contestée mais finit par être acceptée car elle se concentre sur les résultats.
- 📈 Une enquête en Suède a révélé que 36% des jeunes de 15 à 24 ans ont utilisé l'IA générative pour leurs devoirs, avec 55% reconnaissant l'avoir utilisé pour tricher.
- 🛠️ L'IA peut être utilisée comme un outil pour aider les étudiants à surmonter les obstacles, mais il est important de ne pas perdre de vue la qualité de l'apprentissage.
- 🎮 Les jeux vidéo comme Super Mario illustrent comment l'amélioration des compétences des joueurs est plus importante que les améliorations des caractéristiques du jeu.
- 🤔 L'IA peut créer une illusion d'apprentissage, en permettant aux étudiants d'atteindre des objectifs sans le processus d'apprentissage approprié.
- 🧐 Il est crucial de réfléchir aux implications de l'automatisation des tâches, car cela peut priver les individus de possibilités d'apprentissage précieuses.
- 📈 L'IA peut être un outil de coaching pour aider les étudiants à trouver des solutions, plutôt que de simplement leur donner les réponses.
Q & A
Quelle est la meilleure façon de commencer un discours selon le script?
-Le script suggère que le meilleur moyen de commencer un discours est par une histoire, en commençant par décrire quelque chose qui nous représente, comme une petite ville ou un village, puis en introduisant un problème et enfin en incluant une solution.
Dans le contexte de l'IA et de l'éducation, quelles sont les préoccupations principales des enseignants?
-Les enseignants s'inquiètent de la manière dont le développement rapide de l'IA générative affecte leur travail, en particulier comment ils perçoivent et évaluent l'apprentissage et les connaissances de leurs élèves.
Quel est l'impact de l'utilisation des nouvelles technologies dans le saut à la perche sur l'évolution du record mondial?
-Le passage de la perche en bois à la perche souple a transformé le sport et a permis d'une amélioration du record mondial de plus de deux mètres, soit 48%, contre seulement 8% pour les 100 mètres masculins.
Comment le script aborde-t-il la question de la triche lorsqu'il s'agit d'utiliser l'IA générative dans l'éducation?
-Le script soulève la question de la triche en disant que 36% des jeunes de 15 à 24 ans en Suède ont utilisé l'IA générative pour leur travail scolaire, dont 55% l'ont utilisé pour tricher. Il suggère également que l'enseignement de l'utilisation appropriée de l'IA pourrait être plus bénéfique que de la considérer simplement comme un outil de triche.
Quels sont les risques potentiels de l'utilisation d'IA pour l'évaluation des devoirs scolaire?
-L'utilisation de l'IA pour évaluer automatiquement les devoirs pourrait priver les enseignants d'informations précieuses sur les erreurs et les difficultés des élèves, qui sont essentielles pour améliorer l'enseignement et l'apprentissage.
Comment le discours utilise-t-il la métaphore du Super Mario pour expliquer l'importance de l'apprentissage dans le processus?
-Le discours utilise la métaphore de Super Mario pour montrer que l'objectif de l'apprentissage est de se voir soi-même améliorer au fil du temps, plutôt que de dépendre d'outils externes pour progresser, ce qui pourrait créer une illusion d'apprentissage.
Quelle est la position de l'orateur sur l'utilisation de l'IA dans l'éducation?
-L'orateur n'est pas contre l'utilisation de l'IA dans l'éducation, mais souligne que cela est complexe et dépend du contexte. Il insiste sur l'importance de ne pas perdre de vue l'opportunité d'apprentissage lorsque l'on délégue une tâche à l'IA.
Comment le script suggère-t-il que l'IA devrait être utilisée dans l'éducation?
-Le script suggère que l'IA devrait être utilisée comme un coach ou un guide pour aider les étudiants à surmonter les obstacles, plutôt que de simplement leur donner la réponse, afin qu'ils puissent apprendre par leurs propres efforts.
Quels sont les avantages potentiels de l'automatisation de certaines tâches par l'IA pour les enseignants?
-L'automatisation peut être un temps sauveur pour les enseignants en leur permettant de gagner du temps et d'être plus efficaces, surtout dans des tâches répétitives et fastidieuses comme l'évaluation de devoirs.
Quelle est la conclusion finale du discours sur le rôle de l'IA dans l'éducation?
-La conclusion est que le rôle de l'IA dans l'éducation est entre les mains de ceux qui l'utilisent, et que cela dépend des choix qu'ils feront. Il est important d'être conscient de ces choix et de l'opportunité d'apprendre que l'on a.
Comment le discours aborde-t-il la question de l'apprentissage et de la technologie?
-Le discours aborde la question de l'apprentissage et de la technologie en utilisant des métaphores sportives et des exemples concrets pour expliquer que l'apprentissage repose sur la surmont des obstacles et que l'automatisation peut priver les étudiants d'opportunités d'apprentissage précieuses.
Quelle est la recommandation finale du discours pour les utilisateurs de l'IA?
-La recommandation finale est de ne pas se dérober à l'apprentissage en déléguant toutes les tâches à l'IA, mais plutôt de l'utiliser comme un outil pour aider à surmonter les obstacles et apprendre, en demandant à l'IA de fournir des commentaires et des orientations plutôt que des réponses prêtes à l'emploi.
Outlines
🤔 L'Intrigue de l'IA : Dragon ou Chevalier ?
Le premier paragraphe aborde la métaphore d'une histoire pour engager l'auditoire, en utilisant l'image d'une petite ville menacée par un dragon pour illustrer la situation actuelle avec l'IA. Le narrateur, un enseignant de longue date, exprime son intérêt pour l'impact de l'IA généative sur l'éducation et comment les enseignants perçoivent et évaluent l'apprentissage des étudiants. Il soulève les questions sur la définition du triche et utilise l'analogie du saut à la perche pour illustrer l'évolution des outils et des techniques, suggérant que l'IA pourrait être un outil pour surmonter les obstacles, tout comme les perches flexibles ont transformé le sport.
🎓 L'Éducation et l'IA : Outils ou Triche ?
Dans ce paragraphe, le narrateur discute des résultats d'une enquête suédoise montrant que de nombreux étudiants ont utilisé l'IA généative pour tricher. Il mentionne que certains enseignants pensent que l'IA devrait être enseignée comme un outil pour aider les étudiants, mais il inquiète que cela puisse causer des problèmes si l'objectif est confondu avec l'apprentissage réel. Il utilise la métaphore du sport et du jeu vidéo pour expliquer l'importance de l'effort personnel dans l'apprentissage et comment les améliorations peuvent créer une illusion d'apprentissage.
👨💼 L'Expérience de l'IA : Efficacité ou Dépréciation de l'Apprentissage ?
Le troisième paragraphe explore l'idée que l'IA peut être un outil efficace pour les personnes compétentes, mais que pour les débutants, l'automatisation de certaines tâches peut priver d'opportunités d'apprentissage précieuses. Le narrateur critique l'idée que l'automatisation des tâches humaines est toujours une bonne chose et souligne l'importance de l'apprentissage dans le processus. Il cite l'exemple de la correction d'essais par les enseignants, où l'automatisation pourrait entraîner la perte de connaissances précieuses sur les erreurs et les difficultés des étudiants.
🚀 L'AI comme Partenaire d'Apprentissage
Dans le dernier paragraphe, le narrateur conclut que l'apprentissage implique toujours de surmonter des obstacles et que l'IA peut être un coach ou un guide pour aider les étudiants à les surmonter par leurs propres efforts. Il mentionne des exemples comme Khan Academy qui intègre l'IA pour aider les étudiants à trouver des solutions, plutôt que de donner directement la réponse. Il encourage le public à utiliser l'IA pour obtenir des commentaires sur leur travail plutôt que de déléguer complètement la tâche à l'IA, afin de préserver l'opportunité d'apprentissage.
Mindmap
Keywords
💡Intelligence Artificielle (IA)
💡Éducation
💡Générateur d'IA
💡Tricherie
💡Pole Vault
💡Développement technologique
💡Illusion d'apprentissage
💡Automatisation
💡Choix
💡Super Mario
💡Coaching
Highlights
The best way to start a speech is by telling a story, painting a picture of something that represents us, introducing a problem, and including a solution.
AI's role in society is uncertain; it could be a threat (like a dragon) or a savior (like a knight).
The speaker, a teacher for two decades, is researching how teachers are experiencing the rapid development of generative AI.
Generative AI can generate images, sound, and text, leading to concerns about students cheating with this technology.
Defining cheating in the context of learning and technology is not straightforward, as illustrated by the pole vaulting analogy.
The technological advancements in the pole vaulting pole have drastically improved records, with accusations of cheating following new technology.
Generative AI's impact on learning and education is compared to the introduction of new pole types in pole vaulting.
A survey in Sweden showed that 36% of young people between 15 and 24 have used generative AI, with 55% admitting to cheating with it.
Some educators argue that instead of considering the use of generative AI as cheating, students should be taught to use it properly.
The importance of the learning process over the end goal is emphasized, using the metaphor of physical training and Super Mario.
Enhancements in modern video games may create an illusion of learning, as they allow players to progress without improving their skills.
The complexity of integrating AI in education depends on the context and the individual.
Greg Brockman demonstrated GPT-4's capabilities by transforming a hand-drawn sketch into usable website code within a minute.
Brockman emphasized the importance of understanding the code, highlighting the value of personal skill development over automation.
AI can perform tasks that are harmful or dangerous for humans, but it doesn't mean that humans should stop performing all tasks that machines can do.
Focusing too much on the end goal can lead to losing important aspects of the learning process.
Teachers also face the temptation to use AI for grading essays, but manual assessment provides valuable insights into student learning.
Learning involves overcoming obstacles, and the human mind is programmed to take shortcuts, but this can hinder long-term learning.
AI can be used as a coach or guide, helping individuals conquer challenges with their own effort.
Khan Academy's integration of GPT-4 is an example of AI being used to assist students in finding solutions, rather than providing answers directly.
Instead of asking ChatGPT to write a text, individuals can write their own text and then seek feedback, promoting personal learning.
The decision of whether AI is a dragon or a knight depends on individual choices and awareness of those choices.
Transcripts
Transcriber: Nhi Trần Reviewer: Hui Bin Go
They say that the best way to start a speech is simply by saying,
once upon a time, that is, you should tell a story.
And in this story, you should begin by painting
a picture of something that represents us like a small town or a village.
And then you should introduce a problem,
something that comes and threatens that town like a dragon.
And finally, of course, you should include a solution. A savior of some sort.
A knight to slay that dragon.
But when it comes to this story about AI and us,
it’s not entirely easy to determine what role this new technology is going to play
Is AI the Dragon? Or is AI, knight?
I'm a teacher, and I've been in the educational field
for about two decades now.
And when it comes to this story about us and AI,
it is those aspects concerning learning and education that interests me.
And in recent time, I have been given the opportunity to pursue a PhD where
I do research and on how teachers like myself are experiencing this rapid
development that we are seeing within generative AI,
how it affects teachers work,
especially how it affects the way they perceive
and assess their students learning and knowledge.
Generative AI in this context, being AI that can generate things like images,
sound, and in particular text.
And the most common initial reaction that teachers have or anyone,
is that they fear that their student might use this technology in
order to cheat.
But as you dive into these matters of learning and technology,
you quickly realize that it's not entirely easy to determine what
constitutes cheating and what doesn't.
And to explain, let me give you a well-used analogy, pole vaulting.
This is Armand Duplantis. He holds the current world record with 6.23m.
And this is Charles Hoff, a Norwegian who held that very same record
a hundred years ago. In 1923, Hoff cleared 4.21m
And this image that you see, it’s taken at that exact moment.
And back then, I actually think that Hope was quite happy that he didn’t jump
oversix metres, because
100 years ago, Pole vaulter landed on just thin layer of sand.
It was quite brutal.
But over 100 years the record has improved by over two metres.
Or in other words, by 48% .
That's quite a lot, especially when you compare it to other events.
For instance, in the men's 100m,
the corresponding development has only been a mere 8% .
So why has the record increase been so large within pole vault?
Well, the most obvious explanation lies in the technological development of
the pole. In 1923, Hoff used a pole made of hardwood.
And hardwood poles are not great at transferring vertical horizontal motion
into vertical height. When flexible poles came along, it transformed the sport.
What's worth noting is that we know that when new types of pole have been
introduced, it has been quite common for those using
the old type of poles to accuse the adopters of new poles of cheating.
But over time, everyone starts using the new polls
because people don't really care about what polls they use.
They care about how high they jump. What matters is the height of that bar.
The question, of course is, is generative AI in relation to learning
and education comparable to what new types of polls have been for pole vault?
Just a few weeks ago, a large survey was published here in Sweden,
where they had asked young people between the age of 15
and 24 if they had used generative AI in their schoolwork.
And 36% said that they had.
Among them, 55%
admitted that they had used it in order to cheat.
An additional 12% said that they had used it,
but they were unsure whether or not it had been cheating.
And in this context, it's quite common
for teachers who are tech nerds like myself to argue that, well,
we shouldn't consider using tools such as generative AI as cheating.
Instead, we should teach our students how to use it properly.
That if generative AI can be used as
a flexible pole to help our students overcome their higher obstacles,
then surely that's a good thing.
That what matters is the quality of the product that our students produce.
But that is exactly. That kind of arguments that might
lead us to mistake
the dragon for the knight.
You see.
The problem arise, when did you achieving a goal
as a valid expression of learning
Because in the context of learning
is actually within the journey towards the goal where the learning takes place.
To go on with that athletic metaphor
if your coach tells you to
run to a certain location to a particular spot
the purpose is probably for you to cover a certain distance
in order to train your physical fitness.
But if you then use the aids, such as a bicycle, or if
you take a shortcut you will reach
that destination but you will not get the intended training you are supposed to.
Let’s move from that real world metaphor to a digital one.
Raise your hand if you have ever played that classic 8 bit Nintendo game
Super Mario. You know, dee dee dee dee dee dee dee dee dee
Most of you have.
But have you ever reflected on the fact that Super Mario is
the basic same character on every level?
And in order to progress through the game, it's you,
the player that needs to get better.
Compare that to most common video games today.
When you pick up enhancements,
that enables you to go on through even harder levels.
You can even buy these enhancements sometimes.
Well, of course you might get better as you play,
but those enhancements might risk making you think that you’re getting way more
better than you actually are.
Enhancements can create an illusion of learning.
So what’s the point I’m trying to make here?
Am I saying that we shouldn't allow AI within education?
Am I saying that we shouldn't teach our students how to use tools such
as generative AI? No, that's not what I'm saying.
What I'm saying is that it is complicated because it's contextual. It depends.
It varies from person to person.
And to illustrate this, let me give you one more example. This is Greg Brockman.
He is a key figure at OpenAI, the company behind ChatGPT.
And in March this year, they unveiled their newest AI model, GPT four.
And it did this through a live demo broadcasted over YouTube,
where Brockman showcased some of the system's capabilities.
For instance, he took a picture of a hastily hand-drawn sketch of a website.
He uploaded that picture to GPT four,
asking it to analyze it and then write a code for such a website,
and within a minute, Brockman had usable code. Voila!
A new functioning website was born. It was quite impressive. It was awesome.
But there was this one thing that Brockman said during this presentation
that caught my attention. At one instance, when the system wrote code. Brockman said
with emphasis you should always look through the code to get a sense of what it
does never run untrusted code.
and it became so clear Brockman read that code just as easily as
I read my mother tongue Swedish.
And how do you think he has become so good at reading and writing code?
Well, probably because over the years Greg Brockman has read and written
a ton of code. And for someone with such a skill level,
having a system automating that process for you is a huge time saver.
It makes such a person more efficient and that's a good thing.
But for someone who does not possess such skills, such automation.
Might mean that they are being deprived of a valuable learning opportunity.
Now, some might say that, well, when AI can write code,
humans no longer need to. That when machines can perform tasks
human humans don’t know no longer need to perform them.
And that's certainly true in some cases.
If machines can do tasks that, for instance,
are harmful and dangerous for humans, they arguably should.
But we can't do the mistake of thinking that just because
a machine can do something, it automatically means that they ought to.
And this becomes especially clear when we focus too much on the end goal,
the quality of the product, or the height of that bar.
Because it's not only the learning that takes place during this journey.
Quite often there are other important aspects in that process as well.
Aspects that aren't always that obvious for us to see,
and we might risk losing them with automation.
Let me explain with one last example this time an example from education.
As mentioned, teachers fear that their student might use AI for cheating.
Well, let's turn the script a bit because. Sure. But teachers.
Can risk and be tempted to take shortcuts with AI as well.
Anyone who has been a teacher or anyone who has ever known a teacher,
can probably relate to the feeling of
a teacher sitting late at his desk Sunday evening with
a mountain of essays to grade that need to be handed out first thing Monday morning
If they’re in such cases where a magical button on your computer.
That would assess all those essays accurately and instantly.
It would be very tempting to press that button.
I'll admit I would have probably pressed it.
But should we? You see an essential part of being not just a good teacher,
but a great one, is knowledge about what mistakes our student makes.
What they struggle with, what they found difficult.
And such knowledge is to a large degree gained by doing manual assessment.
We should therefore be careful about automating such processes.
So what's the solution then? Well.
Firstly, we need to recognise that learning always involves overcoming
obstacles. If there is no resistance, no real learning is taking place.
The problem is that the human mind is programmed for taking shortcuts to look
for what is most efficient. It's evolutionary. We want to conserve energy.
The problem is that if we always take that shortcut,
we will risk losing valuable learnings that will benefit us long term.
You can see it as climbing a hill.
With today's advanced technology,
we can ask the technology to carry us all the way up to the top.
What I'm saying is that that might be a mistake. And there is another way.
Because a remarkable thing about today’s AI is that
we can ask it to coach us,
to guide us, helping us conquer that hill, but with our own effort.
And it is exciting to see that there are some solutions emerging within education
and technology that take this approach.
Khan Academy being one, for instance,
where they have integrated GPT four in such
a way that when students face problems within math, for instance,
the system does not just give away the answer,
but it helps the student finding the solution themselves.
But you can also do these things directly within ChatGPT or Bing or Google Board.
For you in the audience. It might be a small thing.
Like instead of asking ChatGPT to write a text for you,
write the text yourself and then turn to chatGPT, for its feedback.
Such a small shift might be all the difference.
Whenever we hand over a task to AI, we need to ask ourselves,
are we robbing ourselves of a valuable learning opportunity?
So whether AI plays the role of that dragon or that knight. It’s up to you.
It depends on your choices. It's your call.
But you can only have a choice if you are aware that you have one.
And now you are. Now you are.
(Applause)
Ver más vídeos relacionados
AI and Physics: new Math?
Et si l'IA était un nouvel instrument pour les musiciens ? | Benoit Carré | TEDxTours
IA : La science sans la fiction - Le HubIA Épisode 1
CSUnplugged - L'intelligence artificielle 2
Perdons-nous le contrôle sur l’IA ? | 42 - La réponse à presque tout | ARTE
How to Write a Book with AI in 2024 (2 Best Methods)
5.0 / 5 (0 votes)