omekau on Nostr: AI requires a training set. A learning model. Who provides those and how can you ...

AI requires a training set. A learning model. Who provides those and how can you verify they aren’t biased in a way to cause harm, accidentally or intentionally? It is an incredibly complex system that is unpredictable which means it cannot be trusted. Maybe for asking questions like how to play chess, how to change your car’s oil, what’s the nth number of pi, etc. as those don’t require critical thinking.

Not understanding how a tech like bitcoin works is different than not understanding how an implementation of AI works. With bitcoin, you could understand how it works you’re just abstracted away from it for user experience purposes. With AI, even if you know the training model and data set, there’s an infinite possible inputs and each one of those inputs could theoretically have an infinite possible outputs (using conversational AI for example). Who knows what the implications of those are when people start asking for truths, facts, philosophical questions, morality, etc. IMO it will likely be catastrophic for human flourishing.