Generative AI systems create text, images, audio and video that can be hard to tell apart from human work. That ability brings real benefits, but also new risks: confident falsehoods, realistic deepfakes that make misinformation easier to produce and harder to detect, and unresolved questions about copyright and who is owed credit, or payment, for the work these systems learn from.
The lessons begin with why large language models hallucinate, producing plausible-sounding but false answers. You then hear digital forensics expert Hany Farid on how deepfakes are made and detected, and what they mean for trust in what we see and hear. The last two lessons turn to copyright: a short overview of the three main legal issues, followed by a documentary on artists who say their work was used to train image generators without permission.
Copyright law around AI is still being tested in the courts and differs between countries, so treat these lessons as an introduction rather than legal advice. Watch them in order, then continue on to Work, Inequality & Society.
IBM's Martin Keen explains the different kinds of hallucination, why language models produce plausible-sounding falsehoods, and what users can do to reduce them.
Video 9minDigital forensics expert Hany Farid explains how deepfakes are made, whether they can be detected, and what they mean for truth online. The deepfake discussion starts around 12:00; the full conversation is long and optional.
Video 1h 25minA short overview of the three main copyright questions raised by generative AI: using works to train models, outputs that may infringe, and whether AI-created content can be protected.
Video 5minCNA Insider's Undercover Asia follows artists who found AI copies of their work, and explores fair use, web scraping, image datasets and the lawsuits and tools artists are turning to. A long, optional watch.
Video 46min