The Tools We Still Need to Build with AI

As generative AI applications take the consumer internet by storm, we find ourselves left with questions about the data.

Where does the data come from to train these systems? How do we know where the systems get specific answers? How do we compensate people when those answers have a specific source of provenance? How do we incentivize people to keep publishing their work under threat of becoming fodder for automation?

Neither AI Paperclip doomsdayism nor Pollyannish genAI enthusiasm adequately answer these questions. Nor can regulation, since the technology’s already out there and legal precedent in tech trends both weak and slow. But existing software engineers, machine learning developers, and artists, too, can still build new tools and infrastructure to brighten a future infused with AI tools. This talk will imagine that future and how to get there.

Chelsea Troy
Machine Learning Operations Engineer, Mozilla