[ExI] Tim May and DNA

[Stuart LaForge]

Quoting John Clark:

> That's true, and yet we know for a fact that the entire recipe for  
> building a human being is less than 750 meg long even though its  
> written in convoluted spaghetti code with massive amounts of  
> redundancy.

That figure of 750 MB is more the maximally-compressed information  
content of the haploid genome and not so much the recipe. Much of the  
redundancy in DNA is functional and so the code would likely have to  
be decompressed to be executable. For example, there might be little  
difference in the Shannon information content of the sequence TTAGGG  
repeated thousands of times or just once, but the information content  
does not consider that the purpose of repeating (TTAGGG)n telomeric  
sequence is to give the ends of chromosomes the flexibility to fold  
over themselves repeatedly to hide the tip of the chromosome in the  
center of a complicated knot where DNA-degrading enzymes called  
exonucleases can't access and digest them.

> I want to consider a very small amount of that 750 meg, the part  
> involving the brain hardware and even more important the part that  
> encodes the general learning algorithm that enabled Einstein to go  
> from learning which way is up on the day he was born to learning how  
> General Relativity works 36 years later.
> [. . .]

I think that this recent development in AI is relevant to your point:

https://interestingengineering.com/new-self-aware-robotic-arm-can-recognize-and-repair-itself
https://mindmatters.ai/2019/02/that-robot-is-not-self-aware/
http://robotics.sciencemag.org/content/4/26/eaau9354

"Abstract: A robot modeled itself without prior knowledge of physics  
or its shape and used the self-model to perform tasks and detect  
self-damage."

> All this makes me think the era of true AI may be much closer than  
> many people think, I wouldn't be surprised if the master learning  
> algorithm in its most efficient form is less than a meg in size;  
> with such a program and time to learn from the external world maybe  
> a human level AI could exist on a iPhone.

AI seems capable of most cognitive tasks these days except for logical  
inference perhaps. Then again aside from humano-centric prejudice, it  
may already be here:
https://www.businessinsider.com/this-robot-passed-a-self-awareness-test-that-only-humans-could-handle-until-now-2015-7

Stuart LaForge