Friam

Predictive coding basedon deep learning

Classic

List

Threaded

11 messages Options

Steve Smith

Predictive coding basedon deep learning

Anyone here who codes regularly (daily?) who thinks this is useful?

As an aging has-been, I can see how it could extend my semi-competence
quite a long way... but could undermine something fundamental the way
GPS seems to be undermining geospatial awareness and wayfinding skills?

https://tabnine.com/blog/deep

============================================================
FRIAM Applied Complexity Group listserv
Meets Fridays 9a-11:30 at cafe at St. John's College
to unsubscribe http://redfish.com/mailman/listinfo/friam_redfish.com
archives back to 2003: http://friam.471366.n2.nabble.com/
FRIAM-COMIC http://friam-comic.blogspot.com/ by Dr. Strangelove

Marcus G. Daniels

Re: Predictive coding basedon deep learning

I think GPS increases geographical curiosity, not undermines it.

This technology is objectionable for a different reason: A good software architecture compresses out every pattern into parameterized macro or higher order function. There should be no need to say anything more than once. Whether or not that is achieved in practice is beside the point: Shoddy design should not be be enabled by developer tools.

On 7/28/19, 9:23 PM, "Friam on behalf of Steven A Smith" <[hidden email] on behalf of [hidden email]> wrote:

Anyone here who codes regularly (daily?) who thinks this is useful?

As an aging has-been, I can see how it could extend my semi-competence
quite a long way... but could undermine something fundamental the way
GPS seems to be undermining geospatial awareness and wayfinding skills?

https://tabnine.com/blog/deep

============================================================
FRIAM Applied Complexity Group listserv
Meets Fridays 9a-11:30 at cafe at St. John's College
to unsubscribe http://redfish.com/mailman/listinfo/friam_redfish.com
archives back to 2003: http://friam.471366.n2.nabble.com/
FRIAM-COMIC http://friam-comic.blogspot.com/ by Dr. Strangelove

============================================================
FRIAM Applied Complexity Group listserv
Meets Fridays 9a-11:30 at cafe at St. John's College
to unsubscribe http://redfish.com/mailman/listinfo/friam_redfish.com
archives back to 2003: http://friam.471366.n2.nabble.com/
FRIAM-COMIC http://friam-comic.blogspot.com/ by Dr. Strangelove

Steve Smith

Re: Predictive coding basedon deep learning

On 7/28/19 10:31 PM, Marcus Daniels wrote:

I think GPS increases geographical curiosity, not undermines it.

I would say it apparently does *for you*, and in fact it supports my own *deep curiosity*. I know (many) others who have lost any vestigal sense of wayfinding they might have had, including taking verbal directions or orienting on N/S/E/W or even "toward the mountains or toward the river" without lots of awkward "why should I have to know that?"

I think the "confirmation bias" thread is exposing what I'm nattering on about a bit better.

Eric Smith wrote:

The awareness that there is such an edifice, and that it is something constructed, seems very close to Husserl’s arguments that (in my language) we think of experience as a transparent window through which we passively receive a reality, but it is more like a painted surface on which we are constructing things we believe to be co-registered with something outside the window

To expand Erics metaphor... I believe that the "intermediating painted glass window", in contemporary technology is the "closed cockpit, situational awareness" of modern warfare. All the intermediation can be incredibly helpful. It can expand the dynamic range of our senses ( IR, radar, sonar) and it can provide some prioritization of focus (various alerts, etc.), but it can *also* blunt and distort our intuition and therefore understanding and the predictability that goes with it.

Perhaps what I'm railing about is no more than the Map/Territory conflation. As our maps get better (or more useful for specific purposes), we become less practiced outside the specific purposes encoded IN/BY the map.

Returning to the "painted window" analogy, we need to choose carefully the stylistic preferences of the painter, and for that, we need to recognize there is a painter, and that "they is us" and there *are* choices that can be made and they do matter. Does choosing a pointillist or cubist or surrealist style for our "window-painting" help us or hurt us? Do we, as the painters of our own windows develop good skills and a strong aesthetic awareness, or do we buy the cheap Hobby Lobby colored-plexiglass stain-glass-by-by-number kit and copy patterns we find on Pinterest or YouTube?

Ramble,

- Steve


This technology is objectionable for a different reason:  A good software architecture compresses out every pattern into parameterized macro or higher order function.  There should be no need to say anything more than once.   Whether or not that is achieved in practice is beside the point:   Shoddy design should not be be enabled by developer tools.  

On 7/28/19, 9:23 PM, "Friam on behalf of Steven A Smith" [hidden email] wrote:

    Anyone here who codes regularly (daily?) who thinks this is useful?  
    
    As an aging has-been, I can see how it could extend my semi-competence
    quite a long way... but could undermine something fundamental the way
    GPS seems to be undermining geospatial awareness and wayfinding skills?
    
    https://tabnine.com/blog/deep
    
    
    ============================================================
    FRIAM Applied Complexity Group listserv
    Meets Fridays 9a-11:30 at cafe at St. John's College
    to unsubscribe http://redfish.com/mailman/listinfo/friam_redfish.com
    archives back to 2003: http://friam.471366.n2.nabble.com/
    FRIAM-COMIC http://friam-comic.blogspot.com/ by Dr. Strangelove
    

============================================================
FRIAM Applied Complexity Group listserv
Meets Fridays 9a-11:30 at cafe at St. John's College
to unsubscribe http://redfish.com/mailman/listinfo/friam_redfish.com
archives back to 2003: http://friam.471366.n2.nabble.com/
FRIAM-COMIC http://friam-comic.blogspot.com/ by Dr. Strangelove

============================================================
FRIAM Applied Complexity Group listserv
Meets Fridays 9a-11:30 at cafe at St. John's College
to unsubscribe http://redfish.com/mailman/listinfo/friam_redfish.com
archives back to 2003: http://friam.471366.n2.nabble.com/
FRIAM-COMIC http://friam-comic.blogspot.com/ by Dr. Strangelove

Marcus G. Daniels

Re: Predictive coding basedon deep learning

Steve writes:

“Do we, as the painters of our own windows develop good skills and a strong aesthetic awareness, or do we buy the cheap Hobby Lobby colored-plexiglass stain-glass-by-by-number kit and copy patterns we find on Pinterest or YouTube?”

It depends what the window is looking at. Does the model need to be good or does it just need to get the job done?

Marcus

Steve Smith

Re: Predictive coding basedon deep learning

Marcus -

Steve writes:

“Do we, as the painters of our own windows develop good skills and a strong aesthetic awareness, or do we buy the cheap Hobby Lobby colored-plexiglass stain-glass-by-by-number kit and copy patterns we find on Pinterest or YouTube?”

It depends what the window is looking at. Does the model need to be good or does it just need to get the job done?

For any given instance, I fully acknowledge the value of "cheap heuristics" (e.g "profiling", "rule of thumb", "shoot from the hip"). What I'm trying to expose is the meta-heuristic of being a facile model builder/adopter/fitter... and how our technological prosthetics (precut colored plexiglass and stain-by-number patterns or GPS/routing systems that present opaque-to-the-user preferences or predictive SDE programming environments).

There is also a syntax/semantics tension in your question which seems to parallel the "ends/means" justification.

- Steve

Marcus G. Daniels

Re: Predictive coding basedon deep learning

Steve writes:

< What I'm trying to expose is the meta-heuristic of being a facile model builder/adopter/fitter... and how our technological prosthetics (precut colored plexiglass and stain-by-number patterns or GPS/routing systems that present opaque-to-the-user preferences or predictive SDE programming environments). >

When technology doesn’t work, take it apart and figure out what is wrong with it or how it could be improved. Human experts, or skilled practitioners, can hurt more they help because they have no incentive to unpack their expertise into reusable automated systems. The trick is to look at skills as technology and to be facile evolving the technology.

Marcus

Gary Schiltz-4

Re: Predictive coding basedon deep learning

Boy howdy (I've always wanted to say that) does that ring true here in Ecuador. There are just so many things that I took for granted back in the USA that I can't get here, be it technology, tools, food ingredients. If something I brought from there breaks, I make sure to tear it apart to see if it is something I can fabricate from what I can get here. It really pays to be a Jack/Jill of all trades. I've come to appreciate the value of knowing a little of everything, and not much of anything in particular.

On Mon, Jul 29, 2019 at 1:29 PM Marcus Daniels <[hidden email]> wrote:

Steve writes:

< What I'm trying to expose is the meta-heuristic of being a facile model builder/adopter/fitter... and how our technological prosthetics (precut colored plexiglass and stain-by-number patterns or GPS/routing systems that present opaque-to-the-user preferences or predictive SDE programming environments). >

When technology doesn’t work, take it apart and figure out what is wrong with it or how it could be improved. Human experts, or skilled practitioners, can hurt more they help because they have no incentive to unpack their expertise into reusable automated systems. The trick is to look at skills as technology and to be facile evolving the technology.

Marcus

============================================================
FRIAM Applied Complexity Group listserv
Meets Fridays 9a-11:30 at cafe at St. John's College
to unsubscribe http://redfish.com/mailman/listinfo/friam_redfish.com
archives back to 2003: http://friam.471366.n2.nabble.com/
FRIAM-COMIC http://friam-comic.blogspot.com/ by Dr. Strangelove

Marcus G. Daniels

Re: Predictive coding basedon deep learning

There is a sense in which even professionals can become helpless to the state-of-the-art in any field. For example, if Matlab or Mathematica don’t make it easy, then don’t even try. Or if a compiler produces slow code, change the source code of the program until it does produce optimal code rather than fixing the compiler. When there is no technology, people accept that they have to figure things in out. A problem has to be very serious to motivate the investment in literacy to get over an energy barrier introduced by tried-and-true but never perfect technology, and increasingly, only specialists make that investment. That said, having no technology is a ridiculous solution to the problem.

From: Friam <[hidden email]> on behalf of Gary Schiltz <[hidden email]>
Reply-To: The Friday Morning Applied Complexity Coffee Group <[hidden email]>
Date: Monday, July 29, 2019 at 12:33 PM
To: The Friday Morning Applied Complexity Coffee Group <[hidden email]>
Subject: Re: [FRIAM] Predictive coding basedon deep learning

On Mon, Jul 29, 2019 at 1:29 PM Marcus Daniels <[hidden email]> wrote:

Steve writes:

< What I'm trying to expose is the meta-heuristic of being a facile model builder/adopter/fitter... and how our technological prosthetics (precut colored plexiglass and stain-by-number patterns or GPS/routing systems that present opaque-to-the-user preferences or predictive SDE programming environments). >

When technology doesn’t work, take it apart and figure out what is wrong with it or how it could be improved. Human experts, or skilled practitioners, can hurt more they help because they have no incentive to unpack their expertise into reusable automated systems. The trick is to look at skills as technology and to be facile evolving the technology.

Marcus

============================================================
FRIAM Applied Complexity Group listserv
Meets Fridays 9a-11:30 at cafe at St. John's College
to unsubscribe http://redfish.com/mailman/listinfo/friam_redfish.com
archives back to 2003: http://friam.471366.n2.nabble.com/
FRIAM-COMIC http://friam-comic.blogspot.com/ by Dr. Strangelove

gepr

Re: Predictive coding basedon deep learning

In reply to this post by Marcus G. Daniels

It's in these crevices of the discussion that it becomes obvious that the "painted surface" analogy [†] fails completely. This is why Hoffman's "interface" theory is so attractive. The same core point is made, just with more explanatory power. It's analogous to how we play video games. Good examples are fighting games where you push buttons in different sequences to make the (physics-based) avatar do things in its environment. Our mental map of the controller interface has literally no similarity to the "physical" realities inside the game. But it *works*. You can control the avatar even though the control surface is nothing like the physics it's controlling.

As Hoffman points out in one of his papers, it can even be *bad* to have an accurate understanding of the controlled system. Competent players don't get hung up on, e.g. whether a rapier or a broadsword being wielded by their avatar reflects the real thing. They simply (randomly?) try lots of game play techniques and "git gud". What's being selected for is not a good/true/real mapping, but an effective mapping.

cf. Here's a video I haven't had the chance to watch yet: https://iai.tv/video/the-reality-illusion
FWIW, I find Hoffman's style and attitude in both his writing and presentation very off-putting. But I really like the fundamental idea.

[†] Which I first heard of by D. Dennett, I think, where our mind is projecting a movie onto an opaque screen. The world is projecting an image onto the other side of the screen. And there's some functionality to the screen so that when the two projections *match*, there's some feedback to both. When the projections are too different, then perhaps there's negative feedback.

On 7/29/19 11:28 AM, Marcus Daniels wrote:
> Steve writes:
>
> < What I'm trying to expose is the meta-heuristic of being a facile model builder/adopter/fitter... and how our technological prosthetics (precut colored plexiglass and stain-by-number patterns or GPS/routing systems that present opaque-to-the-user preferences or predictive SDE programming environments). >
>
> When technology doesn’t work, take it apart and figure out what is wrong with it or how it could be improved. Human experts, or skilled practitioners, can hurt more they help because they have no incentive to unpack their expertise into reusable automated systems. The trick is to look at skills as technology and to be facile evolving the technology.

============================================================
FRIAM Applied Complexity Group listserv
Meets Fridays 9a-11:30 at cafe at St. John's College
to unsubscribe http://redfish.com/mailman/listinfo/friam_redfish.com
archives back to 2003: http://friam.471366.n2.nabble.com/
FRIAM-COMIC http://friam-comic.blogspot.com/ by Dr. Strangelove

uǝʃƃ ⊥ glen

Marcus G. Daniels

Re: Predictive coding basedon deep learning

Not clear why one excludes the other. Experience helps to estimate whether a particular system will better yield to careful study or to chaotic perturbation. Witness the success of stochastic gradient descent in machine learning.

On 7/29/19, 11:57 PM, "Friam on behalf of glen∈ℂ" <[hidden email] on behalf of [hidden email]> wrote:

It's in these crevices of the discussion that it becomes obvious that the "painted surface" analogy [†] fails completely. This is why Hoffman's "interface" theory is so attractive. The same core point is made, just with more explanatory power. It's analogous to how we play video games. Good examples are fighting games where you push buttons in different sequences to make the (physics-based) avatar do things in its environment. Our mental map of the controller interface has literally no similarity to the "physical" realities inside the game. But it *works*. You can control the avatar even though the control surface is nothing like the physics it's controlling.

As Hoffman points out in one of his papers, it can even be *bad* to have an accurate understanding of the controlled system. Competent players don't get hung up on, e.g. whether a rapier or a broadsword being wielded by their avatar reflects the real thing. They simply (randomly?) try lots of game play techniques and "git gud". What's being selected for is not a good/true/real mapping, but an effective mapping.

cf. Here's a video I haven't had the chance to watch yet: https://iai.tv/video/the-reality-illusion
FWIW, I find Hoffman's style and attitude in both his writing and presentation very off-putting. But I really like the fundamental idea.

[†] Which I first heard of by D. Dennett, I think, where our mind is projecting a movie onto an opaque screen. The world is projecting an image onto the other side of the screen. And there's some functionality to the screen so that when the two projections *match*, there's some feedback to both. When the projections are too different, then perhaps there's negative feedback.

On 7/29/19 11:28 AM, Marcus Daniels wrote:
> Steve writes:
>
> < What I'm trying to expose is the meta-heuristic of being a facile model builder/adopter/fitter... and how our technological prosthetics (precut colored plexiglass and stain-by-number patterns or GPS/routing systems that present opaque-to-the-user preferences or predictive SDE programming environments). >
>
> When technology doesn’t work, take it apart and figure out what is wrong with it or how it could be improved. Human experts, or skilled practitioners, can hurt more they help because they have no incentive to unpack their expertise into reusable automated systems. The trick is to look at skills as technology and to be facile evolving the technology.

============================================================
FRIAM Applied Complexity Group listserv
Meets Fridays 9a-11:30 at cafe at St. John's College
to unsubscribe http://redfish.com/mailman/listinfo/friam_redfish.com
archives back to 2003: http://friam.471366.n2.nabble.com/
FRIAM-COMIC http://friam-comic.blogspot.com/ by Dr. Strangelove

============================================================
FRIAM Applied Complexity Group listserv
Meets Fridays 9a-11:30 at cafe at St. John's College
to unsubscribe http://redfish.com/mailman/listinfo/friam_redfish.com
archives back to 2003: http://friam.471366.n2.nabble.com/
FRIAM-COMIC http://friam-comic.blogspot.com/ by Dr. Strangelove

gepr

Re: Predictive coding basedon deep learning

Sorry. I didn't intend to imply that an interface must be false. But our tendency is to believe that our interfaces are true. If the closed-minded realists would admit that lots of data on *how* true our effective models are (or aren't) is needed, then we could study it thoroughly. It reminds me of the Nancy Reagans of the world who stifled research on marijuana and psychedelics for decades based on their presumptions.

On July 30, 2019 8:39:52 AM GMT+01:00, Marcus Daniels <[hidden email]> wrote:
>Not clear why one excludes the other. Experience helps to estimate
>whether a particular system will better yield to careful study or to
>chaotic perturbation. Witness the success of stochastic gradient
>descent in machine learning.

--
glen

============================================================
FRIAM Applied Complexity Group listserv
Meets Fridays 9a-11:30 at cafe at St. John's College
to unsubscribe http://redfish.com/mailman/listinfo/friam_redfish.com
archives back to 2003: http://friam.471366.n2.nabble.com/
FRIAM-COMIC http://friam-comic.blogspot.com/ by Dr. Strangelove

uǝʃƃ ⊥ glen