[Fis] Misinformation. 1. A Role for FIS 2. Terry's Strategy

[Mark Johnson]

Dear all

I have always been an AI sceptic. However, recently I found myself involved
in a medical diagnostic project using convolutional neural networks to
automatically diagnose diabetic retinopathy. (In fact we are getting the
machine to rank images in severity, rather than simply classify them)

Convolutional neural networks represent a particular strategy to process
image data in a way which is fractal oriented, using convolution processes
derived from DSP. If you want to see it in action, look here:
https://cs.stanford.edu/people/karpathy/convnetjs/demo/mnist.html

It's been remarkably successful at predicting the likely categories of data
it hasn't seen before (a key feature of machine learning), after being
programmed with redundancy of images (All this stuff works with
redundancy). The Visual Geometry Group in Oxford created a particular model
architecture called VGG16 which has been trained with thousands of images
from the internet. When you play with it, it's a bit spooky - not just for
what it gets rights, but how it gets things wrong in a way where we can say
"I can see how it thinks that"

The tricky point is that nobody knows why it works. There are various
theories.

A more profound question is that if we think machine learning might help us
in our pursuit of truth, with what theory do we ground that claim? What
insight into the operation of machine learning might ground an insight into
information and disinformation?

Best wishes

Mark





On Fri, 10 Jan 2020, 07:49 Joseph Brenner, <joe.brenner at bluewin.ch> wrote:

> Dear Friends and Colleagues,
>
>
>
> I believe I am allowed to make comments ‘on my own’, not as coordinator.
> “There will be recursion”.
>
>
>
> Barstow would seem to need all the benefit of the presumption of
> innocence of any intention of disinformation. Might FIS have a role to play
> in bringing our collective view in this matter to Barstow’s attention? In
> principle, if he is a scientist, he should be open to a critique. In this
> connection I note the comments of both Terry and Mark, the latter referring
> to my definition of an institution, the former to IS4SI.
>
>
>
> I hope that Terry is right that the AI tools of disinformation may be used
> to combat it, but this will not be sufficient. For me, the source and form
> of an ‘informational immune system’ may be such that it cannot be
> ‘automated’. *In encryptiam spes? * The problem is in ourselves, and
> secondarily in our current economic system, and this is where the
> intellectual effort is also needed. The corollary is that an ‘institutional
> response’, if we were to make one, should not be cast in terms of truth,
> but of science and meaning. “Battling over relativism and meaning” is an
> inevitable part of a responsible human condition, but I still would rather
> do so from inside FIS, IS4SI, ICPI (International Center for the
> Philosophy of Information) and IGSIS (Institute for a Global Sustainable
> Information Society) than from inside . . .
>
>
>
> Thank you and best wishes,
>
>
>
> Joseph
>
>
>
>
>
>
> ------------------------------
>
> *From:* Fis [mailto:fis-bounces at listas.unizar.es] *On Behalf Of *Terrence
> W. DEACON
> *Sent:* jeudi, 9 janvier 2020 19:14
> *To:* Mark Johnson
> *Cc:* fis
> *Subject:* Re: [Fis] misinformation, a lecture
>
>
>
> Dear colleagues,
>
>
>
> I am glad that we're having this conversation. It is not just timely, but
> urgent.
>
> The video that Stan posted of David Barstow's talk at the UC Berkeley
> Goldman School of Public Policy is chilling.
>
> Please take the time to watch it.
>
> Whatever else you want to quibble about with respect to the words "truth"
> "reinforcement" "coherence" or whatever
>
> the danger of not taking this problem seriously is monumental,
> existential, and deserving our serious attention.
>
> It is the challenge of understanding referential error-correction as
> opposed to the mere rectification of signal corruption.
>
> And although it is not merely an "academic" issue, it demands serious
> intellectual effort by those of us who study the very nature of
> information. But I fear that we are lagging behind in our theorizing and
> being overwhelmed in the same way that journalists are being swamped by
> spin factories and powerful demagogues. We are still arguing over the
> definitions of information, battling over relativism and meaning, and still
> lack a shared formal analysis of reference, interpretation, and
> informational causality. I see some faint glimmer of hope in progress made
> in encryption and decryption and in the way that blockchain systems help to
> provide a form of encrypted transparency. So even as AI is making deep
> fakes possible and social media enables disinformation to spread far more
> effectively than carefully vetted information, it may also be possible
> to explore how these same tools might be repurposed to provide a kind of
> informational immune system or automated therapies to combat information
> pathogens and information cancers. IS4SI has a role to play in this drama.
>
>
>
> — Terry
>
>
>
> On Thu, Jan 9, 2020 at 7:09 AM Mark Johnson <johnsonmwj1 at gmail.com> wrote:
>
> "the truth needs reinforcement" is slightly chilling don't you think?
> Isn't this an epistemological error?
>
>
>
> Does beauty need reinforcement? Or goodness?
>
>
>
> So what is this? An institution steps in to defends the grounds for its
> continued viability and claim to be the arbiter of truth. Its defense is
> amplified as the uncertainty of its environment increases and its
> judgements questioned. And its defense if itself (and "truth") increases
> environmental uncertainty, as (among other things) other institutions
> defend their competing versions of truth.
>
>
>
> Positive feedback isn't it?
>
>
>
> What's lost is not truth, but coherence.
>
>
>
> Mark
>
>
>
> On Thu, 9 Jan 2020, 14:44 Stanley N Salthe, <ssalthe at binghamton.edu>
> wrote:
>
> STAN
>
>
>
> https://ucsd.tv/search-details.aspx?showID=35394
>
>
>
> _______________________________________________
> Fis mailing list
> Fis at listas.unizar.es
> http://listas.unizar.es/cgi-bin/mailman/listinfo/fis
> ----------
> INFORMACIÓN SOBRE PROTECCIÓN DE DATOS DE CARÁCTER PERSONAL
>
> Ud. recibe este correo por pertenecer a una lista de correo gestionada por
> la Universidad de Zaragoza.
> Puede encontrar toda la información sobre como tratamos sus datos en el
> siguiente enlace:
> https://sicuz.unizar.es/informacion-sobre-proteccion-de-datos-de-caracter-personal-en-listas
> Recuerde que si está suscrito a una lista voluntaria Ud. puede darse de
> baja desde la propia aplicación en el momento en que lo desee.
> http://listas.unizar.es
> ----------
>
> _______________________________________________
> Fis mailing list
> Fis at listas.unizar.es
> http://listas.unizar.es/cgi-bin/mailman/listinfo/fis
> ----------
> INFORMACIÓN SOBRE PROTECCIÓN DE DATOS DE CARÁCTER PERSONAL
>
> Ud. recibe este correo por pertenecer a una lista de correo gestionada por
> la Universidad de Zaragoza.
> Puede encontrar toda la información sobre como tratamos sus datos en el
> siguiente enlace:
> https://sicuz.unizar.es/informacion-sobre-proteccion-de-datos-de-caracter-personal-en-listas
> Recuerde que si está suscrito a una lista voluntaria Ud. puede darse de
> baja desde la propia aplicación en el momento en que lo desee.
> http://listas.unizar.es
> ----------
>
>
>
>
> --
>
> Professor Terrence W. Deacon
> University of California, Berkeley
> _______________________________________________
> Fis mailing list
> Fis at listas.unizar.es
> http://listas.unizar.es/cgi-bin/mailman/listinfo/fis
> ----------
> INFORMACIÓN SOBRE PROTECCIÓN DE DATOS DE CARÁCTER PERSONAL
>
> Ud. recibe este correo por pertenecer a una lista de correo gestionada por
> la Universidad de Zaragoza.
> Puede encontrar toda la información sobre como tratamos sus datos en el
> siguiente enlace:
> https://sicuz.unizar.es/informacion-sobre-proteccion-de-datos-de-caracter-personal-en-listas
> Recuerde que si está suscrito a una lista voluntaria Ud. puede darse de
> baja desde la propia aplicación en el momento en que lo desee.
> http://listas.unizar.es
> ----------
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://listas.unizar.es/pipermail/fis/attachments/20200111/4470ac28/attachment-0001.html>