[FRIAM] New ways of understanding the world

[thompnickson2 at gmail.com]

Jon, 

Does "encoded a theory" get me off my own hook?

N

Nicholas Thompson
Emeritus Professor of Ethology and Psychology
Clark University
ThompNickSon2 at gmail.com
https://wordpress.clarku.edu/nthompson/
 


-----Original Message-----
From: Friam <friam-bounces at redfish.com> On Behalf Of u?l? ???
Sent: Tuesday, December 1, 2020 10:59 AM
To: friam at redfish.com
Subject: Re: [FRIAM] New ways of understanding the world

That sounds like a fairly standard way of distinguishing between intuitiionist and classical conceptions. For me, the problem boils down to whether or not you allow for *actual* infinities or only possible infinities. If you take a doing like "choose a token, apply unary_1 to it, then apply binary_1 to the original token and the output of unary_1", the result will be a theorem. And there's no, in principle, distinction between the doing and the theorem. But if you then say, "do that forever". Then the result isn't really a theorem because there's really no result. You'd have to add something else ... like a convergence operator. But convergence is persnickety. A better operator would be a similarity/distance operator.

Marcus' "try random stuff, possibly reproduce" allows for options in "reproduce" of strong or weak similarity. (Obviously, for copying a software app, you need pretty strong similarity, but perhaps not when you have complex gen-phen maps.) So it seems reasonable to include similarity operators, but not convergence operators. Then instead of "do that forever", you have "do that until similarity_1 < ε". Even if similarity_1 is NOT monotonic, you can stop when you find a procedure that's close enough to a copy to halt. And that feels like a theorem to me.

Of course, there's no reason those derived theorems have to already be present at the very start of the process. The machine can have a "priming" period where it has to "prove" a bunch of theorems that will *eventually* detect the "order". So Nick's "already possessed" is technically wrong, which is what Marcus points out. But, in some sense, those provable theorems are *already* expressible in the prior language. So Nick is spiritually right.

But re: self discovery as corollary to world discovery -- It seems like Wolpert's paper argues against that: https://arxiv.org/abs/0708.1362 The idea that there can only be a single strong inference device in the world implies an asymmetry between world and self discovery. But I can't really pretend to grok the contents of that paper. Maybe someone else here can?

On 12/1/20 8:09 AM, jon zingale wrote:
> It is a little strange to read "possessed a theory" from Nick because 
> he staunchly avoids language like "has consciousness". That said, I 
> read him here as saying that discovery of the self is a corollary to 
> discovery of the world. From my own perspective, theories are derived 
> from (founded upon?) doings, but are not the same thing as doings. 
> Tryings are perhaps more subtle, they evoke for me something like 
> proto-theorems or lemmas. -- 2₵

> On 11/30/20 1:24 PM, thompnickson2 at gmail.com wrote:
>> I was arguing  that given, say, a string of numbers, and no information external to that string, that no AI could detect “order” unless it already possessed a theory of what order is.  I found the discussion distressing because I thought the point was trivial but all the smart people in the conversation were arguing against me.


--
↙↙↙ uǝlƃ

- .... . -..-. . -. -.. -..-. .. ... -..-. .... . .-. .
FRIAM Applied Complexity Group listserv
Zoom Fridays 9:30a-12p Mtn GMT-6  bit.ly/virtualfriam un/subscribe http://redfish.com/mailman/listinfo/friam_redfish.com
archives: http://friam.471366.n2.nabble.com/
FRIAM-COMIC http://friam-comic.blogspot.com/