Of God and Machines
theatlantic.com
theatlantic.com
Its a good article, pretty accessible IMO, and quite poetic.
> "In an attempt to regulate AI, the European Union has proposed transparency requirements for all machine-learning algorithms. Eric Schmidt, the ex-CEO of Google, noted that such requirements would effectively end the development of the technology. The EU’s plan “requires that the system would be able to explain itself. But machine-learning systems cannot fully explain how they make their decisions,” he said at a 2021 summit."
It's very clear that recommendation algorithms on Youtube, for example, are heavily biased in specific directions. The operators of the recommendation system must have access to many dials and filters that control input to and output from the black-box-algorithm, and it's those dials and filters that need to be transparently exposed. Ideally, the user should themselves have access to the filters and dials.
Indeed, the final result of the youtube recommender might well not be clear to the programmers even if they're satisfying their metric (it's been for a variety of ills, notably). But this is just a stronger argument for demanding explainability in this context.