What are deepfakes and are they dangerous? | Start Here

Summary

Jennifer Lopez discusses the alarming rise of deep fakes, a technology using AI to convincingly impersonate people in videos. She explains the process, potential misuse in non-consensual pornography, and its impact on mental health and misinformation. Lopez emphasizes the importance of skepticism and awareness in a digital age where reality can be manipulated, and the ongoing efforts to counter this threat.

Takeaways

• 🎤 Jennifer Lopez discusses the dangers of deep fakes on her favorite show, emphasizing the importance of trusting genuine content.
• 🤖 Deep fakes are created using deep learning, a type of artificial intelligence, to impersonate people in videos and audio.
• 🕵️‍♂️ The process of making deep fakes involves turning videos into images, separating faces, and using machine learning to merge them.
• 🚀 Technology behind deep fakes is rapidly improving, making the creation process easier and more accessible.
• 🌐 There are creative and harmless applications of deep fakes, such as language campaigns and hiding identities for safety.
• 😱 Non-consensual deep fakes, particularly in pornography, are a significant issue, infringing on individuals' rights and causing emotional distress.
• 🔎 Deep fakes can be used maliciously in various contexts, including discrediting individuals and financial scams.
• 🌍 The impact of deep fakes extends to the global stage, influencing politics and causing misinformation.
• 👀 Spotting deep fakes can be challenging, but signs like differences in resolution or ghosting around the face may indicate manipulation.
• 🛡️ Researchers and organizations are working on solutions to detect and counter deep fakes, recognizing their potential threat.
• 📢 The script calls for increased awareness and skepticism, especially when encountering emotionally charged content online.

Q & A

Who is the speaker in the video script?

The speaker in the video script is Jennifer Lopez.

What is Jennifer Lopez excited about in the script?

Jennifer Lopez is excited about being on her favorite show and discussing her new album, as well as the topic of deep fakes.

What is the main concern Jennifer Lopez raises about deep fakes?

Jennifer Lopez raises concerns about the dangers of deep fakes, particularly how they can be used to manipulate and deceive people, and the difficulty in distinguishing what is real and what is fake.

What is the technology behind deep fakes?

The technology behind deep fakes is deep learning by a machine, which is a type of artificial intelligence used to impersonate people.

How does Jennifer Lopez describe the process of creating a deep fake?

Jennifer Lopez describes the process as involving turning a video into hundreds of images (frames), separating the face from the images, and then using machine learning to align and merge the faces, adjusting skin tones and other details in the final result.

What are some positive uses of deep fake technology mentioned in the script?

Some positive uses of deep fake technology mentioned include an anti-malaria campaign with David Beckham speaking nine languages, hiding the real faces of LGBTQ people in Chechnya, and reanimating photos of deceased relatives.

What is the primary ethical issue with deep fakes as discussed in the script?

The primary ethical issue discussed is the non-consensual use of deep fakes, particularly in creating non-consensual pornography, which infringes on individuals' rights and can have severe psychological impacts.

What are some of the negative impacts of deep fakes on individuals mentioned in the script?

Negative impacts of deep fakes include humiliation, embarrassment, and mental health issues such as depression and anxiety.

How have deep fakes been used in scams as described in the script?

Deep fakes have been used in scams by impersonating voices and faces to deceive people into transferring money, as was the case with a businessman who was tricked into transferring over $240,000 to a bogus bank account.

What are some signs that might indicate a video is a deep fake?

Signs that a video might be a deep fake include differences in resolution, ghosting around the face, and blurring around the ears or hairline.

What advice does Jennifer Lopez give on how to approach content that might be a deep fake?

Jennifer Lopez advises that if something makes you feel a strong emotion, it's a good time to take an extra second to check if it's real.

Outlines

00:00

🎭 Deep Fakes: The Art and Danger of AI-Driven Impersonation

Jennifer Lopez discusses the phenomenon of deep fakes, a technology that uses artificial intelligence to convincingly impersonate people in videos. She explains how deep fakes are created using machine learning to superimpose one person's likeness onto another's, as demonstrated by a video merging her face with another's. While there are creative and harmless uses for deep fakes, such as language campaigns and protecting individuals at risk, the technology also poses serious threats. Non-consensual deep fakes, especially in pornography, are rampant, causing emotional distress and damaging reputations. Deep fakes can also be used to spread misinformation and even influence political outcomes, as seen in the case of Gabon's President Ali Bongo. The video emphasizes the importance of being able to discern between genuine and fake content in an increasingly deceptive digital landscape.

05:01

🕵️‍♂️ Spotting Deep Fakes: Detection and the Fight Against Misinformation

In this segment, Jennifer Lopez continues the discussion on deep fakes, focusing on how they can be identified and the broader implications for misinformation. She outlines signs to look for in detecting deep fakes, such as inconsistencies in resolution or ghosting effects around facial features. Lopez also stresses the importance of skepticism, especially when encountering emotionally charged content. As deep fakes become more sophisticated, she points out the ongoing efforts by researchers and organizations like Microsoft, Facebook, the UN, Europol, and the FBI to develop tools and strategies to counter this technology. The segment concludes with a call to action for viewers to stay vigilant and not be deceived by the increasingly realistic fakes that can circulate online.

Keywords

💡Deep Fakes

Deep fakes refer to the use of artificial intelligence, specifically deep learning, to create realistic but fake video or audio content that impersonates real people, making them appear to say or do things they have not. In the video, Jennifer Lopez discusses the dangers of deep fakes, noting how they can be used to manipulate public opinion and deceive individuals, as well as the potential for misuse in non-consensual pornography.

💡Deep Learning

Deep learning is a subset of machine learning that involves the use of artificial neural networks with many layers, known as deep neural networks, to model and understand complex patterns. In the context of the video, deep learning is the technology behind creating deep fakes, allowing for the manipulation of video content to convincingly replace a person's likeness and voice.

💡Impersonation

Impersonation in the video refers to the act of assuming the identity of another person, which is achieved through deep fakes by superimposing one person's likeness and voice onto another's actions or statements. This is highlighted as a significant concern because it can be done without the consent of the person being impersonated and can lead to harmful consequences.

💡Non-consensual Pornography

Non-consensual pornography, as mentioned in the script, is a form of sexual exploitation where deep fakes are used to superimpose individuals' faces onto pornographic content without their permission. The video points out that a significant percentage of deep fakes online are used for this purpose, causing emotional distress and harm to the victims.

💡Misinformation

Misinformation in the video script denotes the spread of false information, which can be exacerbated by the use of deep fakes. The script discusses how deep fakes contribute to a broader issue of misinformation, making it difficult for people to discern truth from falsehood, and leading to the dismissal of actual facts.

💡Disinformation

Disinformation is the deliberate spread of false information, often for the purpose of influencing public opinion or causing confusion. The video mentions that deep fakes can be used in sophisticated disinformation campaigns by powerful states, further complicating the issue of trust in media and communications.

💡Consent

Consent in the context of the video is the voluntary agreement or permission given by a person to have their likeness or voice used in a particular way. The script emphasizes the importance of consent, particularly when discussing the unethical use of deep fakes to create non-consensual pornography or to impersonate individuals without their knowledge or permission.

💡Mental Health

Mental health in the video is discussed in relation to the psychological impact of deep fakes, especially for victims of non-consensual pornography. The script mentions the potential for deep fakes to cause depression, anxiety, and other mental health issues due to the violation of privacy and the difficulty in convincing others of one's innocence.

💡Technology

Technology in the video is portrayed as a double-edged sword, with the potential for both positive and negative applications. While it discusses the creative and harmless uses of deep fakes, such as language campaigns and hiding identities, it also highlights the dangers of misuse, including financial scams and political manipulation.

💡Authenticity

Authenticity in the script is the quality of being genuine or real. Jennifer Lopez emphasizes the importance of being able to trust what is authentic in a world where deep fakes are becoming increasingly sophisticated. The video suggests that as deep fakes improve, it becomes more challenging to determine what is genuine and what is fabricated.

💡Detection

Detection in the video refers to the methods and tools used to identify and verify the authenticity of digital content, particularly in the context of deep fakes. The script mentions that researchers and organizations are working on automated software to detect deep fakes, as well as the importance of being vigilant and skeptical when encountering emotionally charged content.

Highlights

Jennifer Lopez discusses the dangers of deep fakes, emphasizing the importance of trusting genuine content.

Deep fakes have evolved from slightly off look-alikes to highly realistic impersonations that are hard to detect.

The technology behind deep fakes is improving rapidly, making it difficult to assess what is real and what is fake.

Deep fakes use deep learning, a type of artificial intelligence, to impersonate people in videos.

Jennifer demonstrates the process of creating a deep fake using free software called Deep Face Lab.

The process of making deep fakes involves turning videos into frames, separating faces, and machine learning alignment.

Deep fakes are becoming easier to create as technology advances.

There are creative and harmless applications of deep fakes, such as language campaigns and hiding identities.

Deep fakes can be used unethically, often without consent, particularly in the creation of non-consensual porn.

96% of deep fakes online are non-consensual porn, highlighting a significant issue with women's rights and privacy.

Deep fakes can have severe impacts on mental health, causing depression and anxiety due to false representations.

Examples of deep fakes being used for discrediting individuals, such as a mother accused of defaming her daughter's rivals.

Financial scammers have used deep fake technology to impersonate voices and defraud businesses.

Deep fakes contribute to misinformation and can be used to create doubt about real events or statements.

The power of deep fake technology can threaten nations by creating realistic fake videos that can incite unrest.

Spotting deep fakes can be done by looking for inconsistencies in resolution or ghosting around facial features.

As deep fakes improve, they will become increasingly difficult to detect, necessitating advanced detection methods.

Researchers and organizations are working on automated software to detect and flag deep fakes.

The key takeaway is to be more aware and not be easily fooled by the content presented online.

Transcripts

hi everybody it's me jennifer lopez

i'm so excited to be on start here my

absolute

favorite show let's talk about my new

album

or we could talk about the dangers of

deep fakes

they've gone from look-alikes that seem

a bit off

this is a dangerous time trust in what

is genuine

and what is not to scary good videos

that have gone viral

we have now reached the state where the

impersonations

are extremely well done extremely hard

to detect

i genuinely love the process of

manipulating people online for money

the technology behind deep fakes is

getting smarter

while new apps are making it easier the

better deep fakes

become the harder it becomes to assess

what is real and what is fake

so what are deep fakes how can you spot

them

and why could something as fun as a face

swap

actually be the start of something way

more sinister

[Music]

let's break down the word itself the

fake and deep fake is pretty

self-explanatory

while deep refers to deep learning by a

machine which is a type of artificial

intelligence

and this is used to impersonate people

into making saying things you know

they've never said

or acting like that they've acted before

probably the best way to explain it

is to show you how we did it so i'm here

with haldoon start here's graphics

genius and he's taken a video

of me and a video of jennifer lopez so i

used a free software online called deep

face slab

the first step is to turn your video

into hundreds of images basically the

frames

and the second step is to separate the

face

from the images so the software is

pinning my eyes

and my mouth to the center of the frame

so that it's easier to align

with jlo's face and then we go to the

step three which is the machine learning

i need to keep pressing p to update

because like it takes so much time i

just like put the die

on the pee and then put some weight on

and go grab coffee

[Music]

and then we have the final step

basically is adjusting the

skin tune and merging the two faces on

the final result

the whole process actually took a few

days but making deep fakes is getting

easier as the technology improves

and it's improving fast so what can we

do with it

well there are plenty of useful creative

and harmless applications

to all this deep fakery this

anti-malaria campaign had david beckham

speaking nine languages

malaria isn't just any disease

this documentary used deep fakes to hide

the real faces

of lgbtq people in chechnya who were

afraid to be identified

there are slightly weirder uses as well

the website myheritage reanimates photos

of your dearly departed relatives

and some people find that comforting

what all of those examples have in

common

is that they're not about trying to

deceive people and the people involved

are all in on it the problem is when

deep fakes are made of people without

their consent

and so often that means women there are

people out there using pictures of

celebrities

and ordinary women and deep faking them

onto pornography actors

a study in the netherlands found that a

staggering 96

of the deep fakes online were

non-consensual porn

it's women's bodies uh identities um

and rights that are being transgressed

it's humiliating it's embarrassing and

particularly

with deep flex becoming so good it's

very difficult to convince people that

that isn't you because if it looks like

you it might as well be you

we we see big impacts on people's mental

health uh depression anxiety

and it goes beyond the world of porn a

mother in pennsylvania has been accused

of trying to discredit

three of her daughter's rivals on a

cheerleading squad where police said she

made deep fakes of them naked

drinking and smoking financial scammers

are using deep fake technology too

in 2019 criminals used ai software to

impersonate the voice of a businessman's

boss

on the phone they convinced him to

transfer more than 240

000 to a bogus hungarian bank account so

the dangers of deep fakes are already

real

and they're adding to a whole world of

misinformation

a world where it can already be hard to

know what's true and what's not

where actual facts are dismissed as

false conspiracy theories thrive

and powerful states run sophisticated

disinformation campaigns

most of the misinformation we see in

most of what people get affected by

is much lower tech things photos taken

out of context

things that are simple photoshop jobs

much simpler and cheaper ways of making

misinformation go

viral like this video of nancy pelosi a

senior u.s democrat who was made to look

drunk

just by slowing down the video it's

really sad and here's the thing

and i told this to the room it's really

sad

here's the thing and i told this to the

room but the power of deep fake

technology

takes it all to another level in the old

days

if you wanted to threaten the united

states you needed 10 aircraft carriers

and nuclear weapons and long-range

missiles increasingly all you need is

the ability to produce

a very realistic fake video what deep

fakes do

is create a climate of doubt to the

point where what's

actually real can be mistaken as

something fake

that's what happened in gabon

[Music]

back in 2019 president ali bongo hadn't

been seen in public for months

and when the government released this

video he didn't really look like himself

turns out the president had suffered a

stroke that changed how he looked

but some people were convinced he was

dead and that the video was a deep fake

a week later a group of soldiers

attempted a coup

whoa so how can we spot deep fakes

well there are some signs we can look

for there might be differences in

resolution

or if you see ghosting around the face

or blurring around the ears or hairline

chances are a computer made it what i

tell people is if something makes you

feel a strong emotion

either really good or really mad that's

the time to take an extra second

and check to see if it's real but the

reality is

that as deep fakes get better they'll

get harder and harder to spot

researchers at universities and

companies like microsoft and facebook

are working on automated software to

find and flag them

organizations like the un europol and

the fbi

are all actively looking into how to

counter deep fakes as a threat

we're always in this arms race of kind

of a new technology

exists people start using it for bad

things and then we kind of

adjust our understanding and move

forward there's nothing inherently bad

about the technology

but we know the harm it can do so what

you and i can do

is be more aware don't be fooled by the

rocks that i got

or the face that i got either

[Music]

if you want to see more start here

episodes subscribe to the al jazeera

youtube channel and our facebook

instagram and twitter pages where we

release new shows

every week see you there

you