Societal Implications of Generative AI Tutorial

The video explores generative AI, first presenting innovative applications like Google's DeepDream. It quickly addresses darker subjects, focusing on the ethical challenges of deepfakes, particularly their implications in terms of misinformation and national security. It also highlights issues of bias in AI, showing how training data can influence outcomes. A key point is the erosion of the concept of truth, where generative AI can sow distrust in the media and exacerbate social polarization. The video also acknowledges the opportunities offered by AI, while highlighting the tension between innovation and the need for regulation. It concludes by emphasizing the importance of education in navigating this complex landscape.

  • 5:58
  • 552 views
résultat(s)
00:00:03
Generative AI is at the heart of
00:00:05
the new technological frontier,
00:00:07
transforming our ways of seeing and creating.
00:00:10
From the hallucinatory visual art
00:00:12
of Mid Journey, Google's Deepdream,
00:00:14
to the sophisticated melodies
00:00:15
of Open a Is Muse Net,
00:00:17
creative boundaries seem to be fading.
00:00:19
However, with magic comes malice.
00:00:22
Deepfakes blur the lines between real
00:00:25
and fake, posing huge ethical problems.
00:00:27
Deep fakes can be used to create fake
00:00:30
videos or audios that feature public
00:00:33
figures or politicians saying or
00:00:35
doing things they never said or did.
00:00:37
This can be used to spread false
00:00:40
information or propaganda,
00:00:41
manipulating public opinion.
00:00:42
Deep fakes could be used to impersonate
00:00:46
national leaders or military officers,
00:00:48
creating videos that could trigger
00:00:51
diplomatic crises or even conflicts.
00:00:53
As deep fakes become more and
00:00:55
more convincing,
00:00:56
it becomes difficult for the public
00:00:58
to discern the real from the fake.
00:01:00
This could lead to a general erosion
00:01:02
of trust in video and audio media
00:01:04
where everything can be questioned.
00:01:06
Artificial intelligence,
00:01:07
including generative technologies
00:01:09
like deepfakes,
00:01:10
relies on data to learn and generalize.
00:01:13
These data come from the real world,
00:01:15
and just as our world has
00:01:17
inequalities and stereotypes,
00:01:18
these inequalities can
00:01:19
be reflected in the data.
00:01:21
When an AI is trained on biased data,
00:01:24
it can incorporate and
00:01:26
perpetuate these biases.
00:01:27
Thus, if deep fakes are used to,
00:01:29
say,
00:01:29
generate voice or face samples
00:01:31
that do not fairly represent
00:01:32
the diversity of the real world,
00:01:35
then the AI learning from these
00:01:37
samples will itself be biased,
00:01:39
leading to discriminatory systems.
00:01:41
For example,
00:01:42
an AI system for voice recognition
00:01:44
that is trained primarily on male
00:01:46
voices might struggle to recognize
00:01:48
and correctly understand female
00:01:50
voices or other voices that do not
00:01:52
match the model it was trained on.
00:01:54
With the spread of deepfakes,
00:01:56
objective reality could be undermined.
00:01:58
If everything can be falsified,
00:02:00
the very concept of truth could be eroded,
00:02:03
with profound consequences for society.
00:02:05
As deepfakes become more sophisticated
00:02:08
and indistinguishable from reality,
00:02:10
the public could begin to
00:02:12
systematically doubt the veracity
00:02:13
of any video or audio content,
00:02:15
whether it's news, documentaries,
00:02:17
or interviews.
00:02:18
This could lead to a society where people
00:02:21
no longer believe what they see or hear,
00:02:23
making the dissemination of reliable
00:02:25
information extremely difficult.
00:02:27
Media organizations might need to
00:02:29
adopt new methods of validation
00:02:30
and certification to prove the
00:02:32
authenticity of their content.
00:02:34
In a world where objective truth
00:02:36
is questioned,
00:02:36
individuals might turn more towards
00:02:39
information sources that confirm
00:02:40
their pre-existing beliefs,
00:02:42
reinforcing echo chambers and polarization.
00:02:45
Deep fakes could be used to support
00:02:47
conspiracy theories and given that
00:02:49
these videos would be difficult to refute,
00:02:51
these theories could gain in
00:02:54
popularity and influence.
00:02:55
This could lead to increased
00:02:57
fragmentation of society,
00:02:58
with groups becoming more isolated
00:03:01
and disagreeing on fundamental facts.
00:03:04
Democracies rely on
00:03:05
an informed electorate
00:03:06
to function effectively.
00:03:08
If citizens cannot discern the
00:03:09
true from the false due to the
00:03:11
proliferation of deep fakes,
00:03:12
this could compromise the democratic process.
00:03:15
Electoral campaigns could be
00:03:17
marked by falsified videos
00:03:19
intended to discredit opponents,
00:03:21
and elections could be influenced
00:03:23
by misleading information.
00:03:25
Trust in democratic institutions
00:03:27
could decrease,
00:03:28
leading to political instability and
00:03:30
a weakening of democratic values.
00:03:34
How to legislate this
00:03:36
rapidly evolving technology?
00:03:38
The debate between innovation
00:03:39
and regulation is more intense
00:03:41
than ever. Historically,
00:03:43
technology has often evolved faster
00:03:45
than societies ability to regulate it.
00:03:47
For example, when the Internet
00:03:48
began to become popular, it took years
00:03:51
before laws on privacy, copyright,
00:03:53
or cybersecurity were established. With
00:03:55
generative AI, the challenge is even greater
00:03:58
because the technology has the potential
00:04:00
to transform many aspects of our lives,
00:04:02
from media to politics to the economy.
00:04:05
The initial problems posed by the
00:04:07
emergence of deepfakes were mainly
00:04:09
ethical or related to misinformation.
00:04:11
However, initially there was no clear
00:04:13
regulation to penalize or frame the creation
00:04:16
and dissemination of malicious deepfakes.
00:04:19
Strict regulation can hinder innovation.
00:04:21
If severe restrictions are imposed on
00:04:23
research and development in generative AI,
00:04:26
this could prevent potentially
00:04:28
beneficial discoveries for society.
00:04:30
On the other hand,
00:04:31
overly strict regulation can push
00:04:33
innovators to relocate their research to
00:04:36
countries with more lenient legislation.
00:04:38
Suppose regulation imposes a long
00:04:40
and costly approval process for any
00:04:42
new application of generative AI.
00:04:45
This could discourage start-ups
00:04:46
and individual innovators,
00:04:48
thus favoring large companies
00:04:50
with significant resources.
00:04:51
The ideal balance between innovation and
00:04:54
regulation requires close collaboration
00:04:56
between policy makers, researchers,
00:04:58
businesses, and civil society.
00:05:01
It is essential to have open and
00:05:03
inclusive discussions to understand
00:05:05
the implications of technology and
00:05:07
create regulation that protects
00:05:09
society while fostering innovation.
00:05:12
Some countries have begun to
00:05:14
establish AI ethics committees,
00:05:15
composed of experts from various fields
00:05:17
to guide the creation of legislation.
00:05:20
These committees examine the ethical,
00:05:22
social and economic implications
00:05:23
of AI and advise governments on
00:05:26
how to frame the technology.
00:05:28
Generative AI, despite its dangers,
00:05:30
offers unprecedented opportunities.
00:05:32
It can improve lives, create solutions,
00:05:35
and enrich our culture.
00:05:37
Guided by responsibility,
00:05:38
ethics and collaboration,
00:05:40
we can shape the future of
00:05:42
generative AI for the good of all.
00:05:44
Faced with these challenges,
00:05:46
education is paramount.
00:05:48
Solid training allows understanding
00:05:49
the issues,
00:05:50
adopting tools and asking informed questions.

No elements match your search in this video....
Do another search or back to content !

 

résultat(s)
00:00:03
A IA generativa está no centro da
00:00:05
a nova fronteira tecnológica,
00:00:07
transformando as nossas formas de ver e criar.
00:00:10
Da arte visual alucinatória
00:00:12
de Mid Journey, Deepdream do Google,
00:00:14
às melodias sofisticadas
00:00:15
de Open a Is Muse Net,
00:00:17
As fronteiras criativas parecem estar a desvanecer-se.
00:00:19
No entanto, com a magia vem a malícia.
00:00:22
Deepfakes borram as linhas entre o real
00:00:25
e falsa, colocando enormes problemas éticos.
00:00:27
Deep fakes podem ser usados para criar falsificações
00:00:30
Vídeos ou áudios que apresentam público
00:00:33
figuras ou políticos dizendo ou
00:00:35
fazendo coisas que nunca disseram ou fizeram.
00:00:37
Isso pode ser usado para espalhar falso
00:00:40
informação ou propaganda,
00:00:41
manipulação da opinião pública.
00:00:42
Deep fakes podem ser usados para se passar por
00:00:46
dirigentes nacionais ou oficiais militares,
00:00:48
criação de vídeos que podem ser acionados
00:00:51
crises diplomáticas ou mesmo conflitos.
00:00:53
À medida que as deep fakes se tornam mais e
00:00:55
mais convincente,
00:00:56
torna-se difícil para o público
00:00:58
discernir o real do falso.
00:01:00
Isto poderia levar a uma erosão geral
00:01:02
de confiança nos meios de vídeo e áudio
00:01:04
onde tudo pode ser questionado.
00:01:06
Inteligência artificial,
00:01:07
incluindo tecnologias generativas
00:01:09
como deepfakes,
00:01:10
depende de dados para aprender e generalizar.
00:01:13
Estes dados vêm do mundo real,
00:01:15
e tal como o nosso mundo tem
00:01:17
desigualdades e estereótipos,
00:01:18
estas desigualdades podem
00:01:19
refletir-se nos dados.
00:01:21
Quando uma IA é treinada em dados tendenciosos,
00:01:24
pode incorporar e
00:01:26
perpetuar esses preconceitos.
00:01:27
Assim, se as deep fakes estiverem acostumadas,
00:01:29
diga,
00:01:29
gerar amostras de voz ou rosto
00:01:31
que não representam de forma justa
00:01:32
a diversidade do mundo real,
00:01:35
em seguida, a IA aprendendo com estes
00:01:37
as amostras serão, elas próprias, tendenciosas,
00:01:39
conduzindo a sistemas discriminatórios.
00:01:41
Por exemplo
00:01:42
um sistema de IA para reconhecimento de voz
00:01:44
que é treinado principalmente em homens
00:01:46
as vozes podem ter dificuldade em reconhecer
00:01:48
e compreender corretamente o sexo feminino
00:01:50
vozes ou outras vozes que não o fazem
00:01:52
corresponder ao modelo em que foi treinado.
00:01:54
Com a disseminação de deepfakes,
00:01:56
a realidade objetiva poderia ser posta em causa.
00:01:58
Se tudo puder ser falsificado,
00:02:00
o próprio conceito de verdade poderia ser corroído,
00:02:03
com profundas consequências para a sociedade.
00:02:05
À medida que as deepfakes se tornam mais sofisticadas
00:02:08
e indistinguível da realidade,
00:02:10
o público poderia começar a
00:02:12
duvidar sistematicamente da veracidade
00:02:13
de qualquer conteúdo vídeo ou áudio,
00:02:15
sejam notícias, documentários,
00:02:17
ou entrevistas.
00:02:18
Isto poderia conduzir a uma sociedade em que as pessoas
00:02:21
já não acreditam no que vêem ou ouvem,
00:02:23
tornar fiável a divulgação de
00:02:25
informação extremamente difícil.
00:02:27
As organizações de mídia podem precisar
00:02:29
adotar novos métodos de validação
00:02:30
e certificação para comprovar o
00:02:32
autenticidade do seu conteúdo.
00:02:34
Num mundo onde a verdade objetiva
00:02:36
é questionado,
00:02:36
os indivíduos podem voltar-se mais para
00:02:39
fontes de informação que confirmam
00:02:40
as suas convicções pré-existentes,
00:02:42
reforço das câmaras de eco e polarização.
00:02:45
Deep fakes podem ser usados para apoiar
00:02:47
teorias da conspiração e dado que
00:02:49
estes vídeos seriam difíceis de refutar,
00:02:51
estas teorias poderiam ganhar em
00:02:54
popularidade e influência.
00:02:55
Isto pode levar a um aumento do
00:02:57
fragmentação da sociedade,
00:02:58
com grupos cada vez mais isolados
00:03:01
e discordar de factos fundamentais.
00:03:04
As democracias dependem de:
00:03:05
um eleitorado informado
00:03:06
para funcionar eficazmente.
00:03:08
Se os cidadãos não conseguem discernir o
00:03:09
verdadeiro do falso devido ao
00:03:11
proliferação de deep fakes,
00:03:12
Isto poderia comprometer o processo democrático.
00:03:15
Campanhas eleitorais podem ser
00:03:17
marcado por vídeos falsificados
00:03:19
destinados a desacreditar os opositores,
00:03:21
e as eleições podem ser influenciadas
00:03:23
por informações enganosas.
00:03:25
Confiança nas instituições democráticas
00:03:27
podem diminuir,
00:03:28
conduzindo à instabilidade política e
00:03:30
um enfraquecimento dos valores democráticos.
00:03:34
Como legislar
00:03:36
tecnologia em rápida evolução?
00:03:38
O debate entre inovação
00:03:39
e a regulação é mais intensa
00:03:41
do que nunca. Historicamente,
00:03:43
A tecnologia evoluiu muitas vezes mais rapidamente
00:03:45
do que a capacidade das sociedades de regulá-lo.
00:03:47
Por exemplo, quando a Internet
00:03:48
começou a tornar-se popular, demorou anos
00:03:51
antes das leis sobre privacidade, direitos autorais,
00:03:53
ou a cibersegurança. Com
00:03:55
IA generativa, o desafio é ainda maior
00:03:58
porque a tecnologia tem o potencial
00:04:00
transformar muitos aspetos das nossas vidas,
00:04:02
dos meios de comunicação social à política, passando pela economia.
00:04:05
Os problemas iniciais colocados pelo
00:04:07
surgimento de deepfakes foram principalmente
00:04:09
ética ou relacionada à desinformação.
00:04:11
No entanto, inicialmente não havia
00:04:13
regulamento para penalizar ou enquadrar a criação
00:04:16
e disseminação de deepfakes maliciosos.
00:04:19
Uma regulamentação rigorosa pode entravar a inovação.
00:04:21
Se forem impostas restrições severas a
00:04:23
investigação e desenvolvimento em IA generativa,
00:04:26
Isto poderia evitar potencialmente
00:04:28
descobertas benéficas para a sociedade.
00:04:30
Por outro lado
00:04:31
uma regulamentação demasiado rigorosa pode empurrar
00:04:33
inovadores a deslocalizarem a sua investigação para
00:04:36
países com legislação mais branda.
00:04:38
Suponhamos que a regulamentação impõe uma longa
00:04:40
e dispendioso processo de aprovação para qualquer
00:04:42
nova aplicação da IA generativa.
00:04:45
Tal poderia desencorajar as empresas em fase de arranque
00:04:46
e inovadores individuais,
00:04:48
favorecendo assim as grandes empresas
00:04:50
com recursos significativos.
00:04:51
O equilíbrio ideal entre inovação e
00:04:54
A regulamentação exige uma estreita colaboração
00:04:56
entre decisores políticos, investigadores,
00:04:58
as empresas e a sociedade civil.
00:05:01
É essencial ter
00:05:03
discussões inclusivas para entender
00:05:05
as implicações da tecnologia e
00:05:07
criar regulamentação que proteja
00:05:09
ao mesmo tempo que fomenta a inovação.
00:05:12
Alguns países começaram a
00:05:14
criar comités de ética no domínio da IA,
00:05:15
composto por especialistas de várias áreas
00:05:17
orientar a criação de legislação.
00:05:20
Estes comités examinam as questões éticas,
00:05:22
implicações sociais e económicas
00:05:23
de IA e aconselhar os governos sobre
00:05:26
como enquadrar a tecnologia.
00:05:28
IA generativa, apesar dos seus perigos,
00:05:30
oferece oportunidades sem precedentes.
00:05:32
Pode melhorar vidas, criar soluções,
00:05:35
e enriquecer a nossa cultura.
00:05:37
Guiados pela responsabilidade,
00:05:38
ética e colaboração,
00:05:40
podemos moldar o futuro da
00:05:42
IA generativa para o bem de todos.
00:05:44
Face a estes desafios,
00:05:46
A educação é fundamental.
00:05:48
Uma formação sólida permite a compreensão
00:05:49
as questões,
00:05:50
adotar ferramentas e fazer perguntas informadas.

No elements match your search in this video....
Do another search or back to content !

 

RELATED COURSES This tutorial is part of the following courses

Part 02 - Societal and Ethical Implications of Generative AI

Address the challenges and opportunities of generative AI, exploring its societal, ethical impacts, and implications on truth, risks, and trust in the digital era.

  • Your trainer for this course
  • Cours E-learning
  • 540 Views
ALL COURSES

Reminder

Show

/

Join Live Training Session