Thinking Machines Lab published a report arguing that current AI systems, which are typically trained in centralized locations and then frozen, exclude the people they serve. The lab proposes a shift toward distributed, customizable AI that extends human will and judgment.
The report outlines four technical directions to achieve this goal: training strong models with multimodal interaction and customizability; building tools that allow users to fine-tune and train model weights themselves; developing interfaces that widen the human-to-machine communication channel; and publishing research to help more engineers understand how models are made.
These directions aim to move both knowledge and alignment closer to users, ensuring AI systems remain shaped by their communities rather than remaining static artifacts.