The great AI copyright debate

Did AI steal from artists to learn? Unpack the fierce debate over AI training data, copyright, and who should win.
About this Event

Talk Description

AI models such as DALL-E, Midjourney, ChatGPT, Claude, Grok, Gemini and Stable Diffusion do not magically create images and texts from thin air. Rather, they are ‘trained’ using vast amounts of images, videos and text created and copyrighted by others, and copied off the internet – for example, Stable Diffusion was trained on 160 million images. So what about copyright infringement? Join Professor Alex Sims as she unravels the AI ‘training’ and copyright debate. On one side are the original artists and authors who say, most definitely, copyright has been infringed. On the other side are the AI companies who argue there is no ‘copying’ because copyrighted works are broken down into patterns, structures and statistical relationships using small fragments of data called tokens. The data is then free to use, as long as the original is not reproduced, or almost reproduced. Who will be the winner?
Bio

Alex Sims is a Professor in the Business School’s Department of Commercial Law. She has a PhD from Macquarie University, a MComLaw from the University of Auckland, and an LLB from the University of Otago. Her research areas include intellectual property law, particularly copyright law, law and technology and consumer law. She is a technical commentator for Radio NZ’s Nine to Noon programme.