Andrew Trask Questions Concentrated Power in AI
Privacy researcher says AI models assume concentration and calls it a design flaw.
TLDR
In a reply posted on X, Andrew Trask, researcher at Google DeepMind and founder of OpenMined, wrote that unnecessary concentrations of power are net-bad. He stated that AI itself frames the issue around concentration in the model. Trask added that this amounts to a design flaw best addressed through superior alternatives such as federated approaches. The post does not detail any specific proposal or timeline.
Combined views
471
1 Source, first seen 29d ago