I co-wrote a blog post based on a new paper I co-authored. The work proposes a way to provide the public more power over the trajectory of AI development by focusing on data, one of the bottlenecks of modern AI development. Specifically, we sketch out and argue for a data trust to hold data from the digital commons — the cumulative intellectual and cultural contributions of humanity on the internet. AI companies use this resource to develop increasingly capable systems while imposing negative externalities on the rest of us, such as the threat of unemployment from automation and catastrophic risk.
Check out the blog post here.