Deepfake Technology
Lately, there's been a lot of worry about deepfake technology. These are tech tricks that can make fake audio, photos and videos of things that never really happened. While deepfakes have been around for a bit, it's only now that folks are asking the question—what's deepfake technology, especially since celebs are speaking up about being targeted? They're getting caught up in fake videos doing stuff they never did, like ending up in fake adult videos or politicians saying stuff they never actually said.
Concerns about deepfakes have sparked a bunch of actions to fight against them. Laws have been made to try and prevent people from making and spreading them. Big social media sites like Facebook and Twitter have already banned deepfakes. Also, conferences about computer stuff are packed with talks about ways to protect against technology deepfake.
Big tech companies joined forces in February 2024 to stop deepfakes from messing with major elections happening that year. Twenty companies, including Adobe, Microsoft, Google, Meta (which owns Facebook), and OpenAI, teamed up for a "Tech Accord." They promised to work together to make tools like watermarks and detection tricks to find, mark, and prove deepfakes - manipulated videos, audio, and pictures of public figures.
The rise of AI deepfake technology has made politicians worried. They are scared that the use of deepfake technology can make fake stuff to trick voters in big 2024 elections in countries like the US, UK, EU, India, and more. With the elections coming up, it's even more crucial to curb deepfakes from spreading false information.
So what is a deepfake technology, and why is it causing such havoc in the entire world?
A deepfake AI technology allows the user to create a type of fake video, audio, or image that's made using artificial intelligence (AI) technology. A deepfake technology app or deepfake technology website basically swaps one person's face or voice with another's, making it look like someone is doing or saying something they didn't actually do or say.
Deepfake AI Technology
Deepfakes can be created either by changing existing content by swapping one person for another, or they can create entirely new content where someone seems to be doing or saying things they never did.
The biggest disadvantages of deepfake technology is that it can spread fake info that looks like it comes from reliable sources. This can mess with people's trust in what they see and hear online. There's also worry that they could be used to interfere in elections or spread propaganda.
But not all deepfakes are bad; there are some advantages of incredible AI deepfake technology too. They can also be used for fun stuff like in video games or for helpful things like customer service, where they make automated systems sound more human. So while deepfakes can be dangerous, they also have some good uses.
You can understand the seriousness of the issues that even UPSC exam prep centres are now teaching about deepfake technology UPSC to help students get ready for the exam.
Deepfake technology software works by using two algorithms: a generator and a discriminator. The generator starts by making fake content based on what's wanted, and the discriminator checks how real or fake it looks. Then, they keep going back and forth: the generator gets better at making realistic stuff, and the discriminator gets better at spotting any mistakes. This cycle helps improve the quality of the fake content over time.
The generator and discriminator algorithms team up to form what's called a generative adversarial network, or GAN. This network uses deep learning to study real images and then uses that knowledge to make fakes. For deepfake photos, the GAN looks at different angles of the target person to capture all the details. For videos, it looks at various angles and also studies behaviour, movements, and speech patterns. This data goes through the discriminator multiple times to make sure the final fake looks as real as possible.
To elaborate deepfake technology, the deepfake videos are made in two main ways. One way is by using an original video of the target person and manipulating it to make them say or do things they never actually did. The other way is by taking a video of someone else and swapping their face with the target person's face, which is called a face swap. Both methods can create convincing fake videos that can be misleading or harmful.
Deep Learning
The term "deepfake" comes from merging "deep learning" and "fake." Deepfake technology meaning is utilising deep learning to make fake videos, audio or pictures.
This technology leans heavily on deep learning methods, especially generative adversarial networks (GANs), which were pioneered by Ian Goodfellow and his team back in 2014. GANs involve two neural networks, a generator, and a discriminator, collaborating to produce fake data that looks very close to real data. This discovery was a big deal and paved the way for deepfake technology to emerge.
The word "deepfake" was first coined in 2017 by an anonymous Reddit user who went by the username "Deepfakes." This user utilised Google's open-source deep-learning technology to produce and share pornographic videos that featured manipulated faces of celebrities.
In the beginning, deepfakes were mainly used for fun stuff like swapping the faces of celebrities or putting faces into movies. But as the tech got better, so did its ability to trick people. New deepfake technology has become really advanced, making it harder to tell what's real and what's fake. This made folks worried about how they could be misused, like spreading lies, ruining someone's reputation, or messing with politics.
To create a deepfake video of someone, the creator starts by feeding lots of real video footage of the person into a neural network. This helps the network learn how the person looks from different angles and lighting. Once trained, the network is combined with computer graphics techniques to put a copy of the person onto another actor.
Adding AI has made this process much faster than it would have been otherwise. However, it still takes time to make a believable video that puts the person in a completely made-up situation. To avoid any glitches in the image, the creator has to adjust many settings in the trained program. It's a complex process that requires careful tweaking.
To combat malicious deepfakes, various measures are being taken globally. In the U.S., laws have been enacted to criminalise deepfake pornography and prevent its use in elections, although their enforceability is limited to specific jurisdictions.
Other countries like China and South Korea are also taking action. In the UK, laws are under review to address deepfake creation within existing frameworks. Tech companies like Facebook, Twitter, and YouTube are implementing policies to detect and remove deepfakes, while research labs are developing tools to identify manipulated videos.
Deepfakes are created using software such as DeepFaceLab, FaceSwap, or FakeApp.
Whether or not deepfake is legal depends on the context and potential harm it causes. In India, the use of deepfake images for fraudulent activities or identity theft can be prosecuted under several sections of the IPC and the IT Act, 2000.
Deepfakes can be illegal if they are used for malicious purposes, such as fraud or defamation.
An example of a deepfake could be a video where a public figure's face is replaced with someone else's.
Deepfakes may be allowed for certain purposes, such as entertainment or artistic expression, but their use must comply with relevant laws and regulations.
Some deepfake software may be freely available, but the legality of their use varies depending on the jurisdiction and purpose.
Get The CEO Magazine to your Door Steps; Subscribe Now
