[FRIAM] Derivatives of Turing Machines

[uǝlƃ ↙↙↙]

I haven't had time to think about any of this or read the paper, but here are the notes I took during the video simply to express that I am (lazily) interested.

• The idea of an infinitesimal binary delta *toward* another binary number reminded me of Conway's surreals.
  So rather than propagating uncertainty, could we propagate "lefts" and "rights" (or "ups" and "downs")?
• Keeping track of how many times you "use" a hypothesis is the functional equivalent of consuming data.
• Introduction and elimination allow for degenerate loops. If all hypotheses are "promoted", is Dereliction moot? I.e. how do we talk about infinite streams as input data?

How that relates to the idea of collections of locally coned machines churning out reality will have to wait for more relaxed times.

On 10/12/20 8:03 PM, jon zingale wrote:
> Recently, there was some discussion of linear logic that sent me down a hole
> where I found this lecture:
> https://www.youtube.com/watch?v=IW4LjjAWrO4&ab_channel=DanielMurfet
> 
> I am really enjoying it. What is the derivative of a Turing machine?


-- 
↙↙↙ uǝlƃ