The EU’s struggle with general-purpose AI
If you’re on social media or other corners of the internet a lot, you’ve likely encountered AI systems with names like GPT-3 or DALL·E in recent months. These are some of the AI systems that have become surprisingly good at creating convincing snippets of text or computer-generated images. These models come with risks, too: for example, they can perpetuate harmful stereotypes against women or marginalized groups. How can these risks be addressed? And what part can regulation play in this?
In the EU, these AI systems are now referred to as “general-purpose AI” and policymakers are grappling with this exact question in negotiations of the EU’s so-called AI Act. In this discussion, we want to explore what effective regulation of “general-purpose AI” could look like in the EU, who needs to be held accountable, and how the equation changes once open source software enters the mix.