[FRIAM] quotes and questions

[Marcus Daniels]

I was thinking of a deep Boltzmann machine having an energy function made of up of millions of quadratic and linear terms, but pick your favorite deep-fake sampling tech.   There's a growing literature on learning physical systems using generative adversarial neural networks.    I have seen these techniques mimic high resolution physical simulations to high fidelity, so learning how bubbles and heat in a hot tub works seems doable.   Yes, then there's the matter of learning more and more indirect latent features of the physical system and learning distributions of the latent features themselves.    

-----Original Message-----
From: Friam <friam-bounces at redfish.com> On Behalf Of glen
Sent: Friday, May 13, 2022 4:08 PM
To: friam at redfish.com
Subject: Re: [FRIAM] quotes and questions

I don't think that's adequate to match Dave's "feeling" requirement. It seems to me "feeling" is affective or emotional. I disagree with Dave, though. Emotions *might* be expressible in equations, again if we read "equations" liberally ... e.g. not identities but equivalences, maybe not pure equivalences, allowing a little flex and slop, and generatively, where materials/inferences are "used up" by the process, where the transformation is lossy.

An emotional behavior may be generated simply by allowing higher order processes. So the robot would have to do some kind of sensor fusion (composition of collections of sensor IO) and dissolution. (Operations over distributions might suffice.) I don't see any reason to think that "equations" are limited in their ability to express that. But I can see that some *kinds* of equations are inadequate. E.g. I doubt standard ODEs would suffice. We'd at least need PDEs. Bottom turtle might be ≥ 18 large: 1st and 2nd xyz derivatives and roll, pitch, yaw. What's that 12 now? Plus maybe another 6 for the velocity and acceleration of the rpys? And they'd have to be stochastic, I'm guessing, at least for irreversibility. I'm not sure how to manage the "used up" part ... maybe with the boundary conditions? A lot of ground might be saved if we could slice it into organs and do the [de]fusion over the organs rather than arbitrary collections of base objects.

IDK. It seems premature to simply assume it can't be done. *I* can't do it. But I'm not the sharpest tool in the shed, either.


On 5/13/22 15:47, Marcus Daniels wrote:
> I am sure I have said it dozens of times before:   Create a robot covered in sensors of similar pressure and temperature sensitivity.  Have it sit in the tub and use some algorithm to learn the distribution of the sensors and how relates to the performance of its own motor system.
> 
>> On May 13, 2022, at 3:36 PM, Prof David West <profwest at fastmail.fm> wrote:
>>
>> On 5/12/22 13:56, Jon Zingale wrote:
>>> An interesting property of turbulence is that it need not be a statement about fluids, but rather a property entailed by a system of equations.
>>
>> McGilchrist would assert that the "reality" that is apprehended by the left-brain is precisely that set of abstract equations. However, the right-brain apprehension of "reality" is the totality of the experience of sitting in the spa and feeling the bubbles and jets caress your body.
>>
>> The latter is not expressible in equations.
>>
>> davew
>>
>>
>>
>>
>>> On Fri, May 13, 2022, at 1:47 PM, glen wrote:
>>>> On 5/12/22 10:32, Steve Smith wrote:
>>>> I personally don't think "Turbulent Flow" is an oxymoron.
>>>
>>> Exactly! That's the point. By denouncing negation, I'm ultimately 
>>> denouncing contradiction in all it's horrifying forms. It's judo, 
>>> not karate.
>>>
>>>> On 5/12/22 13:56, Jon Zingale wrote:
>>>> An interesting property of turbulence is that it need not be a statement about fluids, but rather a property entailed by a system of equations.
>>>
>>> I'm a bit worried about all the meaning packed into "property", 
>>> "entailed", and "system of equations". But as long as we read 
>>> "equations" *very* generously, then I'm down.
>>>
>>>> On 5/12/22 19:54, Marcus Daniels wrote:
>>>> Unitary operators are needed.  Apply a Trumping operator you get a Biden and apply another one to get a Trump back.    To make this work a bunch of ancillary bits are needed to record all the wisdom that Trump destroys.    I am afraid we are dealing with a dissipative system, though.
>>>
>>> IDK. The allowance of unitary operators seems to be a restatement of 
>>> orthogonality. In a world where no 2 variates/objects can be 
>>> perfectly separated, there can be no unitary operators. (Or, perhaps 
>>> every operator has an error term. f(x) → y ∪ ε) I haven't done the 
>>> work. But it seems further that we can define logics without 
>>> negation and logics without currying. Can we define logics with 
>>> neither? What's the expressive power of such a persnickety thing? Is 
>>> it that such a thing can't exist? Or merely that our language is 
>>> incapable of talking about that thing with complete faith? Biden is 
>>> clearly not not(Trump), at least if the object of interest is "too 
>>> damned {old, white, male}". If that's the object, clearly Biden ≡ 
>>> Trump and ∀x|x(Trump) = x(Biden) ∪ ε, where |ε| >> |x(Trump)-x(Biden)|.


--
Mɥǝu ǝlǝdɥɐuʇs ɟᴉƃɥʇ' ʇɥǝ ƃɹɐss snɟɟǝɹs˙

-. --- - / ...- .- .-.. .. -.. / -- --- .-. ... . / -.-. --- -.. .
FRIAM Applied Complexity Group listserv
Fridays 9a-12p Friday St. Johns Cafe   /   Thursdays 9a-12p Zoom  bit.ly/virtualfriam
un/subscribe http://redfish.com/mailman/listinfo/friam_redfish.com
FRIAM-COMIC http://friam-comic.blogspot.com/
archives:  5/2017 thru present https://redfish.com/pipermail/friam_redfish.com/
  1/2003 thru 6/2021  http://friam.383.s1.nabble.com/