Pratique du Shadowing: Large Language Models explained briefly - Apprendre l'anglais à l'oral avec la vidéo
Choisissez un mode d'entraînement
Chaque mode ouvre une salle d'entraînement en plein écran.
81 phrases
Si les phrases sont trop courtes ou trop longues, cliquez sur Edit pour les ajuster.
1
Imagine you happen across a short movie script that describes a scene between a person and their AI assistant.
Imaginez que vous tombiez par hasard sur un scénario de court métrage décrivant une scène entre une personne et son assistant IA.
2
The script has what the person asks the AI.
3
but the AI's response has been torn off.
4
Suppose you also have this powerful magical machine that can take any text
5
and provide a sensible prediction of what word comes next.
6
You could then finish the script by feeding in what you have to the machine, seeing what it would predict to start the AI's answer,
7
and then repeating this over and over with a growing script completing the dialogue.
8
When you interact with a chatbot, this is exactly what's happening.
9
A large language model is a sophisticated mathematical function that predicts what word comes next for any piece of text.
10
Instead of predicting one word with certainty, though, what it does is assign a probability to all possible next words.
11
to build a chatbot.
12
What you do is lay out some text that describes an interaction between a user and a hypothetical AI assistant,
13
You add on whatever the user types in as the first part of that interaction,
14
And then you have the model repeatedly predict the next word that such a hypothetical AI assistant would say in response, and that's what's presented to the user.
15
In doing this, the output tends to look a lot more natural
16
if you allow it to select less likely words along the way at random.
17
So what this means is even though the model itself is deterministic, A given prompt typically gives a different answer each time it's run.
18
Models learn how to make these predictions by processing an enormous amount of text, typically pulled from the internet.
19
For a standard human to read the amount of text that was used to train GPT -3, for example, if they read "non -stop, 24 /7," it would take over 2600 years.
20
Larger models since then train on much, much more.
21
You can think of training a little bit like tuning the dials on a big machine.
22
The way that a language model behaves is entirely determined by these many different continuous values, usually called parameters or weights.
23
Changing those parameters will change the probabilities that the model gives for the next word on a given input.
24
What puts the "large" in large language model is how they can have hundreds of billions of these parameters.
25
you No human ever deliberately sets those parameters.
26
Instead, they begin at random, meaning the model just outputs gibberish, but they are repeatedly refined based on many example pieces of text.
27
One of these training examples could be just a handful of words, or it could be thousands, but in either case, the way this works is to pass in all
28
but the last word from that example into the model, and compare the prediction that it makes with the true last word from the example.
29
An algorithm called backpropagation is used to tweak all of the parameters in such a way
30
that it makes the model a little more likely to choose the true last word
31
and a little less likely to choose all the others.
32
When you do this for many, many trillions of examples, Not only does the model start to give more accurate predictions on the training data,
33
but it also starts to make more reasonable predictions on text that it's never seen before.
34
Given the huge number of parameters and the enormous amount of training data, The scale of computation involved in training a large language model is mind -boggling.
35
To illustrate, imagine that you could perform 1 billion additions and multiplications every single second.
36
How long do you think it would take for you to
37
do all of the operations involved in training the largest language models?
38
Do you think it would take a year?
39
Maybe something like 10 ,000 years?
40
The answer is actually much more than that.
41
It's well over 100 million years.
42
you This is only part of the story, though.
43
This whole process is called pre -training.
44
The goal of auto -completing a random passage of text from the internet
45
is very different from the goal of being a good AI assistant.
46
To address this, chatbots undergo another type of training, just as important, called reinforcement learning with human feedback.
47
Workers flag unhelpful or problematic predictions, and their corrections further change the model's parameters,
48
making them more likely to give predictions that users prefer.
49
Looking back at the pre -training though, this staggering amount of computation is only made possible by using special computer chips
50
that are optimized for running many many operations in parallel, known as GPUs.
51
However, not all language models can be easily parallelized.
52
Prior to 2017, most language models would process text one word at a time.
53
But then, a team of researchers at Google introduced a new model known as the Transformer.
54
Transformers don't read text from the start to the finish.
55
they soak it all in at once, in parallel.
56
The very first step inside a transformer, and most other language models for that matter, is to associate each word with a long list of numbers.
57
The reason for this is that the training process only works with continuous values, so you have to somehow encode language using numbers.
58
and each of these list of numbers may somehow encode the meaning of the corresponding word.
59
What makes Transformers unique is their reliance on a special operation known as attention.
60
This operation gives all of these lists of numbers a chance to talk to one another, and refine the meanings that they encode based on the context around, all done in parallel.
61
For example, the numbers encoding the word "bank" might be changed based on the context surrounding it, to somehow encode the more specific notion of a riverbank.
62
Transformers typically also include a second type of operation, known as a feed -forward neural network,
63
and this gives the model extra capacity to store more patterns about language learned during training.
64
All of this data repeatedly flows through many different iterations of these two fundamental operations.
65
And as it does so, the hope is
66
that each list of numbers is enriched to encode whatever information
67
might be needed to make an accurate prediction of what word follows in the passage.
68
At the end, one final function is performed on the last vector in this sequence, which now has had a chance to be influenced by all the other context from the input text,
69
as well as everything the model learned during training, to produce a prediction of the next word.
70
Again, the model's prediction looks like a probability for every possible next word.
71
Although researchers design the framework for how each of these steps work, it's important to understand
72
that the specific behavior is an emergent phenomenon based on how those hundreds of billions of parameters are tuned during training.
73
This makes it incredibly challenging to determine why the model makes the exact predictions that it does.
74
What you can see is that when you use large language model predictions to autocomplete a prompt,
75
the words that it generates are uncannily fluent, fascinating, and even useful.
76
If you're a new viewer and you're curious about more details on how transformers and attention work, boy do I have some material for you.
77
One option is to jump into a series I made about deep learning, where we visualize and motivate the details of attention and all the other steps in a transformer.
78
But also, on my second channel I just posted a talk
79
that I gave a couple months ago about this topic for the company TNG in Munich.
80
Sometimes I actually prefer the content that I make as a casual talk rather than a produced video, but I leave it up to you which one of these feels like the better follow on.
81
you you
Télécharger l'application
Tout pour parler couramment
Prononciation IAChaque phrase notée
Phonétique (IPA)Maîtrisez chaque son
VocabulaireEnrichissez votre lexique
Jeu de vocabulaireApprenez en jouant
Même chaîne
C1Differential equations, a tourist's guide | DE1
Daily ConversationUniversity Students
C12
Linear combinations, span, and basis vectors | Chapter 2, Essence of linear algebra
American AccentCambridge Exam PrepUniversity StudentsWorking Adults
C15
Fourier series
American AccentUniversity StudentsWorking Adults
C14
Euler's Formula Poem
American AccentIELTS PrepTOEFL PrepUniversity Students
C12
Vectors | Chapter 1, Essence of linear algebra
American AccentBusiness EnglishUniversity StudentsWorking Adults
C13
How might LLMs store facts | Deep Learning Chapter 7
American AccentIELTS PrepTOEFL PrepUniversity Students
C113
Essence of linear algebra preview
American AccentBusiness EnglishUniversity StudentsWorking Adults
C15
The inscribed square problem
American AccentUniversity StudentsWorking Adults
C14
The Newton art puzzle
American AccentUniversity StudentsWorking Adults
C15
Transformers, the tech behind LLMs | Deep Learning Chapter 5
American AccentBusiness EnglishUniversity StudentsWorking Adults
C12
Oh, wait, actually the best Wordle opener is not “crane”…
American AccentBusiness EnglishUniversity StudentsWorking Adults
C23
Hilbert's Curve: Is infinite math useful?
American AccentBusiness EnglishUniversity StudentsWorking Adults
C211
But what is a neural network? | Deep learning chapter 1
American AccentBusiness EnglishUniversity StudentsWorking Adults
C1But what is the Fourier Transform? A visual introduction.
Daily ConversationUniversity Students
C1Divergence and curl: The language of Maxwell's equations, fluid flow, and more
American AccentBusiness EnglishUniversity StudentsWorking Adults
C1But what is quantum computing? (Grover's Algorithm)
American AccentBusiness EnglishUniversity StudentsWorking Adults
C1Tie random ends: How many loops?
American AccentUniversity StudentsWorking Adults
FOR YOU
Vidéo recommandée
C12
105 phrases
Google’s AI Course for Beginners (in 10 minutes)!
American AccentBusiness EnglishUniversity StudentsWorking Adults
C111
102 phrases
World Models explained in 10min..
Business EnglishUniversity StudentsWorking Adults
B28
59 phrases
An ex-OpenAI researcher just deleted language from the LLM...
American AccentBusiness EnglishUniversity StudentsWorking Adults
B23
137 phrases
AI Agents, Clearly Explained
American AccentBusiness EnglishUniversity StudentsWorking Adults
C168
35 phrases
SEtalk: What a Second Language Does to Your Brain
American AccentIELTS PrepTOEFL PrepUniversity Students
C12
33 phrases
PyTorch in 100 Seconds
American AccentBusiness EnglishUniversity StudentsWorking Adults
C117
118 phrases
Computational Linguistics: Crash Course Linguistics #15
Business EnglishUniversity StudentsWorking Adults
C12
68 phrases
Prompt engineering - Introduction
Business EnglishUniversity StudentsWorking Adults
C11
125 phrases
Introduction to statistical prediction models
Daily ConversationUniversity Students
C11
143 phrases
MCP vs API: The protocol every developer needs to know
Business EnglishUniversity StudentsWorking Adults
Vocabulaire et conseils d’expression pour cette leçon
Cette leçon d’expression orale de niveau C2 s’appuie sur la vidéo « Large Language Models explained briefly ». Les mots qui reviennent le plus souvent : model, word, training, prediction, text. Cette vidéo contient 81 phrases et 1312 mots à répéter en shadowing. La partie parlée dure 7:51. Le locuteur parle à un rythme naturel d’environ 167 mots par minute, proche d’une conversation courante. 83 % des mots font partie des 3 000 mots les plus courants en anglais ; le reste mérite d’être vu avant de commencer.
Vocabulaire clé de cette vidéo
Les 15 mots les plus avancés de la vidéo, avec leur prononciation et leur sens :
| Mot | Prononciation | Sens |
|---|---|---|
| prediction nom | /pɹɪˈdɪkʃən/ | prédiction |
| parameter nom | /pəˈɹæm.ə.tɚ/ | paramètre |
| transformer nom | /tɹænsˈfoɹməɹ/ | transformateur |
| encode verbe | /ɪnˈkoʊd/ | encoder, coder |
| predict verbe | /pɹɪˈdɪkt/ | prédire |
| chatbot nom | /ˈtʃætbɑt/ | chatbot, dialogueur |
| probability nom | /ˌpɹɑ.bəˈbɪl.ə.ti/ | probabilité |
| computation nom | /ˌkɑmpjʊˈteɪʃn̩/ | calcul, comptage |
| refine verbe | /ɹɪˈfaɪn/ | raffiner |
| hypothetical adjectif | /ˌhaɪpəˈθɛtɪkəl/ | hypothétique |
| prompt verbe | /pɹɑmpt/ | inciter |
| researcher nom | /ˈɹiˌsɝ.t͡ʃɚ/ | chercheur, chercheuse |
| autocomplete nom | autocomplétion | |
| backpropagation nom | rétropropagation | |
| boggle verbe | /ˈbɒɡ.əl/ | être époustouflé, être confus |
Les verbes à particule que vous entendrez
| Mot | Prononciation | Sens |
|---|---|---|
| take over verbe | reprendre | |
| tear off verbe | /teɹ ɑf/ | arracher |
La grammaire de cette vidéo
Les structures que le locuteur utilise le plus, avec les mots exacts de la vidéo :
| Structure | Dans la vidéo |
|---|---|
| Voix passive be + participe passé — l’accent est mis sur ce qui arrive, pas sur qui le fait | has been torn · are repeatedly refined · is called |
| Present perfect have/has + participe passé — une action passée qui compte encore maintenant | has been torn · has had |
Prononciation à surveiller
Le locuteur utilise 3 contractions et formes réduites, comme you're, don't. Prononcez-les sous leur forme courte, telles que vous les entendez.
- Les sons « th »: hypothetical /ˌhaɪpəˈθɛtɪkəl/, algorithm /ˈælɡəɹɪðm̩/, mathematical /ˌmæθ(.ə)ˈmæt.ɪ.kəl/
- Les sons « sh » et « zh »: prediction /pɹɪˈdɪkʃən/, computation /ˌkɑmpjʊˈteɪʃn̩/, gibberish /ˈd͡ʒɪ.bə.ɹɪʃ/, iteration /ˌɪt.əˈɹeɪ.ʃən/, multiplication /ˌmʌltɪplɪˈkeɪʃən/
- Mots longs — placez bien l’accent: parameter /pəˈɹæm.ə.tɚ/, probability /ˌpɹɑ.bəˈbɪl.ə.ti/, hypothetical /ˌhaɪpəˈθɛtɪkəl/, deterministic /dɪˌtɜː.mɪˈnɪs.tɪk/, iteration /ˌɪt.əˈɹeɪ.ʃən/
Les sons difficiles pour les francophones :
- /h/ — il se souffle, il n’est pas muet: hypothetical /ˌhaɪpəˈθɛtɪkəl/, unhelpful /ʌnˈhɛlpfəl/, behave /bɪˈheɪv/
- /tʃ/ et /dʒ/ — à ne pas adoucir en « ch » et « j »: chatbot /ˈtʃætbɑt/, researcher /ˈɹiˌsɝ.t͡ʃɚ/, emergent /ɪˈmɝd͡ʒənt/, enrich /ɪnˈɹɪt͡ʃ/, gibberish /ˈd͡ʒɪ.bə.ɹɪʃ/
- /r/ anglais — langue recourbée, sans frotter la gorge: prediction /pɹɪˈdɪkʃən/, parameter /pəˈɹæm.ə.tɚ/, transformer /tɹænsˈfoɹməɹ/, predict /pɹɪˈdɪkt/, probability /ˌpɹɑ.bəˈbɪl.ə.ti/
Comment s’entraîner avec cette vidéo
- Écoutez la vidéo en entier une fois sans parler et notez les mots que vous ne connaissez pas.
- Commencez à la vitesse 0,75×, répétez phrase par phrase, puis revenez à la vitesse normale quand cela devient facile.
- Enregistrez-vous et comparez avec l’original, en faisant attention à des mots comme prediction, parameter, transformer.
Qu'est-ce que la technique du Shadowing ?
Le Shadowing est une technique d'apprentissage des langues fondée sur la science, développée à l'origine pour la formation des interprètes professionnels. Le principe est simple mais puissant : vous écoutez de l'anglais natif et le répétez immédiatement à voix haute — comme une ombre suivant le locuteur avec un décalage de 1 à 2 secondes. Les recherches montrent une amélioration significative de la précision de la prononciation, de l'intonation, du rythme, des liaisons, de la compréhension orale et de la fluidité.