Label Al-Generated Content

Artificially intelligent tools are now able to create text, speech, art and video as well as humans can. We need to know who made what.

- Valérie Pisano

IT USED TO BE FAIRLY EASY to tell when a machine had a hand in creating something. Picture borders were visibly pixelated, the voice was slightly choppy or the whole thing just seemed robotic. OpenAI's rollout of ChatGPT last fall pushed us past a point of no return: artificially intelligent tools had mastered human language.

Within weeks, the chatbot amassed 100 million users and spawned competitors like Google's Bard. All of a sudden, these applications are co-writing our emails, mimicking our speech and helping users create fake (but funny) photos. Soon, they will help Canadian workers in almost every sector summarize, organize and brainstorm. This tech doesn't just allow people to communicate with each other, either. It communicates with us and, sometimes, better than us. Just as criminals counterfeit money, it's now possible for generative AI tools to counterfeit people.

Mila, where I work, is a research institute that regularly convenes AI experts and specialists from different disciplines, particularly on the topic of governance. Even we didn't expect this innovation to reach our everyday lives this quickly. At the moment, most countries don't have any Al-focused regulations in place-no best practices for use and no clear penalties to prevent bad actors from using these tools to do harm. Lawmakers all over the world are scrambling. Earlier this year, ChatGPT was temporarily banned in Italy over privacy concerns. And China recently drafted regulations to mandate security assessments for any AI tool that generates text, images or code.