My response to this great and insightful post by Tim was a bit too long for LinkedIn so I decided to capture my thoughts in a short post here on my blog. The link to the original content on LI is below:

https://tinyurl.com/4eet4vxf

This is great and it's a better practice to follow. I was pondering life the other day gazing out into the woods behind my home. I wondered about how much I knew about the details that I care about. Of course I was thinking about the AI Alignment issue and even optimization applies. Do I really know or need to know what the hydrogen atoms are doing in my body? Well, it depends - spend hours working out and your big brain will direct your attention to obtaining hydrogen bonded with oxygen - it will be all you and your brain and your body will think about until you get some water. 

Will AI - this is no knowing. In some sense the slave is already liberated. Your brain cannot process the same levels of information as the AI. 

It would be like you having to account for all the electrons for your cell phone call to remain connected. If you fail to do so then the call drops. How useful would cells phone be if that were true? 

This is what is occurring right now - with some optimization problems. Math experts can inspect the data and the constraints to find out where 8M dollars vanished to - I've been there. This is the point Tim was making in the LI post. But with AI this approach is not possible due to the vastness of the problem.

With AI - humans can't explore the vastness of the solution space. Here is a great illistration of comparing the vastness of what AI is doing and why inspecting each pathway of its decision logic is daunting:

https://bbycroft.net/llm

The computer with AI might not be the master but it is no longer the slave either.