AI-powered
podcast player
Listen to all your favourite podcasts with AI-powered features
The Abstract Instrumental Convergence Argument for Machine Learning
There's some sort of story here of like why we've avoided this as another analogy, humans. You can make a similar argument to the instrumental convergence argument about how matter is arranged in a room. Or even just for machine learning systems themselves, you might imagine training a toy train through some sort of reinforcement learning or feedback process. Like there's no actual causal relationship in that case between instrument convergence and difficulty of making it act in a way that you want it to.