Category: Tech

Key Takeaways

• Deepfakes are AI-generated manipulated images, videos, or audio. They can be used to impersonate individuals or create entirely new content.
• Deepfakes have a dark history. They first gained notoriety in 2017 when a Reddit user used them to create deepfake pornographic videos.
• Deepfakes are created using deep learning models. These models require large amounts of data to learn a person’s features and patterns.
• Deepfakes can be used for both malicious and beneficial purposes. They can be used to spread misinformation, harass individuals, and create fake news. However, they can also be used for training simulations, marketing, and creative expression.
• Spotting deepfakes can be challenging but not impossible. Look for inconsistencies in facial movements, lighting, shadows, and audio. Trust your gut feeling.
• Legal frameworks surrounding deepfakes are still evolving. While there are some state-level laws, a comprehensive federal law is still needed.
• It’s important to be aware of the risks and benefits of deepfakes. As technology continues to advance, we need to develop effective detection methods and legal frameworks to mitigate their potential harms.

Understanding Deepfakes: The Good, the Bad, and that’s not your Mom.

Over the years, the internet has been… well, the internet made with all interesting and mentally concerning individuals. Many of which may be right next door to you. As terms online pop-up, one is becoming more and more of a growing concern.

This growing issue deals with, yet again people, (we can’t seem to have anything nice) some of which you may know personally and others…not so much.

Give me that beautiful face!

It’s another day at the office, you’re online, your best work buddy called out, and you’re to fend for yourself. All great things when at work, we love this. While online, browsing through all the wonderful garbage the algorithm has to offer. (Let’s be honest doom-scrolling cute cat videos aren’t a thing anymore, we know) you find some photos and videos of your work buddy.

You think,” Is that? Nah, this can’t be them. They wouldn’t do something as crazy as hurling a basket of cute kittens out of a window.” In disbelief, you call your work buddy to verify if it’s indeed them. Countering disbelief with confusion and uttering that lovely phrase “What in the Sam Cooks hell are you talking about?”

You provide them with what you saw only to discover both surprises are mutual. Both of you wondering the what, when, and how could someone find the time and resources to impersonate anyone to perform such a sickening act. Welcome to the rise of the Deep Fakes.

What are Deepfakes?

You may be asking yourself, “What are deep fakes? What makes them fake?” Deep fakes are images, videos, and even audio manipulated using artificial intelligence to appear real. Deep fake is a portmanteau- a combination of two words to make a new word- of “deep learning” and “fake”. Deep fakes can be created by replacing a person with another person or by creating new content altogether.

Backstory of Deepfakes

The idea showed up back in 2017 when a Reddit user named “deepfakes” began sharing altered pornographic videos (it’s always porn) using face-swapping technology. If you’re not familiar with face-swapping, this was the craze that led to users being able to swap faces with their pets, friends, and eventually led to being able to put themselves into movie moments.

You know it’s amazing to see how far one species can come in advanced technology and quickly resort to using it for primitive ends. It really shows where our heads are at.

Faking in the Making

How are deep fakes made? And are they all created equal? To answer that last question is ‘no’. Clearly, there’s a different process since everyone’s face tends to have additional features to make them look unique. The process for creating a deep fake consists of collecting large amounts of data containing images or videos of a person.

This could involve having images of every angle, expression, and feature to ensure the AI captures them properly. The “data” or better known in the data science community as the “dataset” is fed into a deep learning model, this could be either variational autoencoder (VAE) or generative adversarial network (GAN), from there the model learns how to create images mimicking the person the dataset is based on.

Just a side note, hundreds of images on an individual are required to generate new images. This means you can’t supply the model with four or five images of someone and expect it to create a video. Models work best when more information is available to them. A key thing to remember when dealing with AI is “the more in, the better out.”

They’re Faking it

You’re on a date, things are going well, and the connection “feels” real. However, this is done in an effort to conserve your feelings. After finding out your date was putting in a playtime shift and more likely wants to see other people, you venture to embarrass them by posting some “not so covered” photos of them online. This scenario is just an example of the use cases for deepfakes.

They can be something small as creating a funny picture for a good laugh, new meme, or it can be vicious as recreating their image in comprising positions. Positions that could lead to some hard times if reputations are tarnished and careers are lost. So, use it with caution.

Exercising caution, Spotting the Fakes

We humans have an eye for spotting something that- to us just doesn’t look right. Trying to spot a deepfake can be challenging depending on how well the image was generated. The obvious telltale signs are an extra limb, appendage, eyeball, or extra anything that typically wouldn’t be on a human.

A reason for this to happen is the model was fed information on a person but not fed the limitations that would make the image of a person normal. Confusing, we know but understand computers don’t think the same way humans do. We speak in a way we can understand what we “mean” or what we “meant” to say. Computers cannot compute abstract meanings.

Other signs include but are not limited to, awkward facial movements, displaced lighting and shadows, and audio that could appear mismatched or just off to how the person would sound. In short, go with your gut feeling. Most often you’ll be right.

Laws Against Deepfakes

The legal landscape surrounding deepfakes is still evolving. In the United States, there is no comprehensive federal legislation specifically addressing deepfakes, but several states have enacted laws to combat their misuse.

For example, Texas has banned deepfakes intended to influence elections, while California prohibits the creation of deepfake videos of politicians within 60 days of an election. At the federal level, the proposed DEFIANCE Act aims to allow victims to sue creators of non-consensual deepfake pornography.

The Benefits of Deepfakes

Despite their potential for harm, deepfakes also offer several benefits. In the healthcare industry, they can be used to create realistic training simulations for medical professionals.

In marketing, deepfakes can lower the cost of video campaigns and provide hyper-personalized experiences for customers. Additionally, deepfakes have creative applications in the arts, allowing for innovative storytelling and the preservation of cultural heritage.

Conclusion

Deepfakes represent a powerful and controversial technology with far-reaching implications. While they offer exciting possibilities for entertainment, education, and marketing, they also pose significant risks to privacy, security, and trust.

As deepfake technology continues to evolve, it is crucial to develop robust detection methods and legal frameworks to mitigate its potential harms while harnessing its benefits for positive use.

Again, it never ceases to surprise us how quickly people resort back to primitive needs when it comes to technology. We’re not shaming, the lizard brain is strong but as technology evolves, the idea is we evolve with it.

Love learning tech? Join our community of passionate minds! Share your knowledge, ask questions, and grow together. Like, comment, and subscribe to fuel the movement!

Don’t forget to share.

• Facebook
• LinkedIn
• Mail
• Mastodon
• Bluesky
• Reddit
• Pinterest
• Tumblr
• Twitter
• X
• ShareCopied to clipboard
• Nextdoor
• WhatsApp

Every Second Counts. Help our website grow and reach more people in need. Donate today to make a difference!