Feel good Sunday: How to keep AI under control
With the rapid advancement of AI technology, there is a growing concern about the potential dangers of super intelligent AI. However, scientist Max Tegmark offers an optimistic outlook on how we can ensure AI works for us, not against us.
But what if the AI can’t create the necessary tools on its own? Tegmark suggests training one AI to learn your desired tasks and then using another AI to extract the learned algorithm and knowledge, similar to an AI neuroscientist.
This approach aligns with the progress being made in mechanistic interpretability, showing that creating safe AI systems is within reach.
For more insights on controlling AI, check out Max Tegmark’s TED talk:
https://www.ted.com/talks/max_tegmark_how_to_keep_ai_under_control
