Ethics and AI: A Dip into Deepfake Technology
![Ethics and AI: A Dip into Deepfake Technology](/content/images/size/w2000/2024/07/ethics.png)
What is deepfake AI?
Deepfakes use a form of artificial intelligence (AI) called ‘deep’ learning to create uncannily realistic, but ultimately artificial images and videos.
💡Deepfake is a portmanteau of ‘deep’ learning and fake.
… and how does it work?
Deepfakes use two algorithms - a generator and a discriminator. The generator creates the initial fake content, while the discriminator analyses the accuracy of the content generated. This process is repeated so that the generator can refine the realism of its creations, and the discriminator can become better at discerning flaws for the generator to correct.
In combination, they form a generative adversarial network (GAN) which uses deep learning to recognise patterns in real images. For example, a GAN system views photographs and videos from various angles to capture all the details and perspectives, as well as analysing behaviours, speech, and movement patterns for the latter. These patterns are then run through the discriminator multiple times to fine tune the authenticity of the final product.
Deepfakes can be created either by using an original video source of the target, where their speech and actions are then manipulated by this technology to achieve the desired result, or the target’s face can be placed onto a video of another individual (i.e. face swap).
Friend or foe?
From an increase in the number of fake Robert Pattinson dancing videos, to a shocking image of an explosion near the Pentagon, deepfakes have made their existence well-known to modern society. Should we welcome this existence, or bar it at the door?
Pros: The Productive Uses
- Art and entertainment: Deepfake videos can enhance the experience of a gallery or museum. For example, in Florida, the Dalí Museum has a deepfake video of the artist introducing their artwork and taking photos with visitors. Furthermore, in the entertainment industry, deepfakes can be used to improve the dubbing of foreign language films, manipulate actors’ actions and voices for particular scenes if the actor is absent, and create satirical content and parodies.
- Customer support/call services: These services use deepfakes to provide customers with personalised responses and to complete simple tasks that customers may request through phone calls (❗AI overtaking human jobs).
- Medical Aid: Voice-cloning deepfakes can help restore the voices of those who have lost them.
Cons: The Distortion of Truth
- Fraud: Deepfake technology can be used to impersonate an individual to obtain private information such as bank account details. This is a clear breach of privacy and raises concerns about the threat deepfake AI poses to cybersecurity. In support, experts at CSIRO have estimated that deepfakes could ‘shatter trust in biometric authentication for 30 per cent of companies by 2026’. There is also growing concern about the use of voice cloning to hack voice-protected applications such as Google assistants.
- Political manipulation/misinformation: Deepfake images, videos, and audio files of politicians, celebrities, or trusted sources can be used to manipulate public opinion. In March 2022, the deepfake video of Ukrainian President Volodomyr Zelenskyy calling on his soldiers to ‘lay down arms’, caused panic and confusion in the midst of warfare before it was debunked and removed. Are we really able to bear the consequences of our own technology and creations?
- Stock manipulation: Forged deepfake materials can be used to influence a company’s stock prices, whether that be to sabotage competition, or to expand one’s own market. For example, the previously mentioned artificial image of the Pentagon going up in flames, caused a 0.3% decrease in the Dow Jones Industrial Index in four minutes, before the video was exposed as being fake. Is the presence of deepfake AI in business an ethical and sustainable practice? Can we dismiss it as harmless marketing, or does it intrude on consumers’ autonomy to make their own judgements about the businesses they’re supporting?
How to Tell Apart Real VS Fake
Finally, here are some things to look out for to detect deep fakes:
- Face
- Unnatural flickering or blinking of the eyes, or a lack of blinking
- Unnatural movement of the mouth and jaw
- Awkward facial and body movements
- Motion and emotion
- Do the words match the mouth movement?
- Do the words match the facial expressions and emotions being depicted?
- Lighting and colouring
- Look out for odd lighting and shadows
- Unnatural colouring
- Difference in texture
Hone your judgement of the truth and stay safe on the internet!
Bella Li
Also check out our podcast: CSE News - Ethics in Tech