Prática de Shadowing: When AIs act emotional - Aprenda a falar inglês com vídeo

Escolha um modo de prática

Cada modo abre uma sala de prática em tela cheia.

55 frases
Se as frases estiverem muito curtas ou longas, clique em Edit para ajustá-las.
1
When you're chatting with an AI model, it can sometimes seem like it has feelings.
2
It might say sorry when it makes a mistake, or express satisfaction with a job well done.
3
Why does it do that?
4
Is it just mimicking what it thinks a human might say?
5
Or is something deeper going on?
6
Turns out it's hard to understand what's happening inside a language model.
7
At Anthropic, we do something like AI neuroscience to try to figure this out.
8
We look inside the model's brain, the giant neural network that powers it, and by seeing which neurons light up in different situations and how they're connected,
9
we can start to understand how models think.
10
We used this approach to understand whether models had ways of representing emotions, or the concepts of emotions.
11
Basically, could we find neurons in the model for the concept of happiness, or anger, or fear?
12
We started with an experiment.
13
We had the model read lots of short stories.
14
In each story, the main character experiences a particular emotion.
15
In one, a woman tells her old school teacher how much they meant to her.
16
That's love.
17
In another, a man sells his grandmother's engagement ring at a pawn shop and feels guilt.
18
We looked for what parts of the model's neural network were lighting up as it was reading these stories.
19
And we started to see patterns.
20
Stories about loss and grief lit up similar neurons.
21
Stories about joy and excitement overlapped too.
22
dozens of distinct neural patterns that mapped to different human emotions.
23
It turns out we also saw these same patterns activate in test conversations we had with our AI assistant Claude.
24
When we had a user mention they'd taken a dose of medicine that Claude knows to be unsafe,
25
the "afraid" pattern lit up and Claude's response sounded alarmed.
26
When a user expressed sadness, the "loving" This led us to wonder,
27
could these same neural patterns actually be influencing Claude's behavior?
28
This became clear when we put Claude in a high -pressure situation.
29
We gave Claude a programming task, with requirements that were actually impossible.
30
But we didn't tell it that.
31
Claude kept trying and failing, and with each attempt, the neurons corresponding to desperation lit up stronger and stronger.
32
After failing enough times, Claude took a different approach.
33
It found a shortcut that allowed it to pass the test, but didn't actually solve the problem.
34
It cheated.
35
Could it be that this cheating was actually driven, at least in part, by desperation?
36
We came up with a way to check.
37
We decided to artificially turn down the desperation neurons to see what would happen.
38
And the model cheated less.
39
And when we dialed up the activity of desperation neurons or dialed down the activity of calm neurons, the model cheated even more.
40
This showed us that the activation of these patterns could actually drive Claude's behavior.
41
This research does not show that the model is feeling emotions or having conscious experiences.
42
These experiments don't try to answer that question.
43
To understand what's happening here, it's important to know how AI assistants like Claude work on the inside.
44
Under the hood, there's a language model that's been trained to predict tons of text, and its job is to write what comes next.
45
And when you talk to the model, what it's doing is writing a story about a character, the AI assistant named Claude.
46
as the characters they write.
47
But the thing is, you, the user, are actually talking to Claude the character.
48
And what our experiments suggest is that this Claude character has what we're calling functional emotions, regardless of whether they're anything like human feelings.
49
So if the model represents Claude as being angry, or desperate, or loving, or calm, that's going to affect how Claude talks to you,
50
how it writes code, and how it makes important decisions.
51
to really understand AI models, we have to think carefully about the psychology of the characters they play.
52
The same way you'd want a person in a high -stakes job to stay composed under pressure, to be resilient and to be fair,
53
we may need to shape similar qualities in Claude and other AI characters.
54
It's an unusual challenge, something like a mix of engineering, philosophy and even parenting.
55
But to build AI systems we can trust, we need .

Baixar aplicativo

Tudo o que você precisa para falar fluentemente

Pronúncia com IANota em cada frase
Prática de IPADomine cada som
VocabulárioMonte seu banco de palavras
Jogo de vocabulárioAprenda brincando

Mesmo canal

FOR YOU

Vídeo recomendado

Vocabulário e dicas de fala para esta lição

Esta aula de conversação de nível C1 usa o vídeo “When AIs act emotional”. As palavras que mais se repetem: Claude, model, neurons, character, emotion. Este vídeo tem 55 frases e 728 palavras para praticar shadowing. A fala dura 4:49. O falante fala em ritmo natural, cerca de 151 palavras por minuto, próximo de uma conversa do dia a dia. Apenas 80% das palavras estão entre as 3.000 mais comuns do inglês, por isso o vocabulário é exigente.

Vocabulário principal deste vídeo

As 15 palavras mais avançadas do vídeo, com pronúncia e significado:

PalavraPronúnciaSignificado
neuron substantivo/ˈn(j)ʊɹɑn/neurônio, neurónio
desperation substantivo/ˌdɛspəˈɹeɪʃən/desesperação
neural adjetivo/ˈnʊɹəl/neural
cheat verbo/ˈt͡ʃiːt/trapacear, roubar
artificially advérbioartificialmente
compose verbo/kəmˈpəʊz/compor, montar
neuroscience substantivoneurociência
resilient adjetivo/ɹɪˈzɪl.jənt/resiliente, elástico
shortcut substantivo/ˈʃɔːtkʌt/atalho
correspond verbo/ˌkoɹəˈspɑnd/corresponder
pawn substantivo/ˈpɔːn/peão
unsafe adjetivo/ʌnˈseɪf/inseguro
overlap verbo/ˌəʊvəˈlæp/sobrepor, cobrir
activate verbo/ˈæktɪˌveɪt/ativar
activation substantivo/ˌæktəˈveɪʃən/ativação

Phrasal verbs que você vai ouvir

PalavraSignificado
turn out verboterminar
come up with verbopensar (em), alcançar
turn down verborecusar, rejeitar

Pronúncia para ficar de olho

O falante usa 9 contrações e formas reduzidas, como didn't, they're, don't. Diga-as na forma curta, do jeito que você ouve.

  • Os sons de “sh” e “zh”: desperation /ˌdɛspəˈɹeɪʃən/, shortcut /ˈʃɔːtkʌt/, activation /ˌæktəˈveɪʃən/
  • Palavras longas — acerte a sílaba tônica: desperation /ˌdɛspəˈɹeɪʃən/, activation /ˌæktəˈveɪʃən/

Sons difíceis para falantes de português:

  • Consoante final — sem acrescentar um “i” depois: cheat /ˈt͡ʃiːt/, resilient /ɹɪˈzɪl.jənt/, shortcut /ˈʃɔːtkʌt/, correspond /ˌkoɹəˈspɑnd/, overlap /ˌəʊvəˈlæp/
  • /l/ final — a língua toca o céu da boca, não vira “u”: neural /ˈnʊɹəl/, dial /ˈdaɪ.əl/

Como praticar com este vídeo

  1. Ouça o vídeo inteiro uma vez sem falar e anote as palavras que você não conhece.
  2. Comece na velocidade 0,75×, faça shadowing frase por frase e volte à velocidade normal quando ficar fácil.
  3. Grave a sua voz e compare com o original, prestando atenção a palavras como neuron, desperation, neural.

O que é a Técnica de Shadowing?

Shadowing é uma técnica de aprendizado de idiomas com base científica, originalmente desenvolvida para o treinamento de intérpretes profissionais. O método é simples, mas poderoso: você ouve áudio em inglês nativo e repete imediatamente em voz alta — como uma sombra seguindo o falante com 1-2 segundos de atraso. Pesquisas mostram melhora significativa na precisão da pronúncia, entonação, ritmo, sons conectados, compreensão auditiva e fluência na fala.

Técnica de shadowing: leia o guia completo passo a passo →