Return to site

每日跟讀#760: About Technology - What is Deepfake?


· 每日跟讀單元 Daily English

每日跟讀#760: About Technology - What is Deepfake?

Deepfakes rely on a branch of AI called Generative Adversarial Networks (GANs). It requires two machine learning networks that teach each other with an ongoing feedback loop. The first one takes real content and alters it. Then, the second machine learning network, known as the discriminator, tests the authenticity of the changes.


GANs are still in the early stages, but people expect numerous potential commercial applications. For example, some can convert a single image into different poses. Others can suggest outfits similar to what a celebrity wears in a photo or turn a low-quality picture into a high-resolution snapshot.


But, outside of those helpful uses, deepfakes could have sinister purposes. Consider the blowback if a criminal creates a deepfake video of something that would hurt someone’s reputation — for instance, a deepfake video of a politician "admitting" to illegal activities, like accepting a bribe.


Next Article:

Facebook is making its own AI deepfakes to head off a disinformation disaster 防制假訊息災難,臉書積極產製人工智慧「深度偽造」影片

Facebook fears that AI-generated “deepfake” videos could be the next big source of viral misinformation—spreading among its users with potentially catastrophic consequences for the next US presidential election.


Its solution? Making lots of deepfakes of its own, to help researchers build and refine detection tools.


The rise of deepfakes has been driven by recent advances in machine learning. Algorithms capable of capturing and re-creating a person’s likeness have already been used to make point-and-click tools for pasting a person’s face onto someone else.


Facebook will dedicate $10 million. Together with Microsoft and academics from institutions including MIT, UC Berkeley, and Oxford University, the company is launching the Deepfake Detection Challenge, which will offer unspecified cash rewards for the best detection methods.


Source article: ;