[NetBehaviour] The Uncanny Vibe

Eryk Salvaggio eryk.salvaggio at gmail.com
Wed Jul 14 00:46:51 CEST 2021


Hi Alan,
I have to disagree. In a neural network, the text that we're seeing
produced is the product of terabytes of human activity, scraped off of the
web, collectively, and generalized to language pairs. How the machine
determines how to answer a question is still a black box question. And yes:
the illusions are coming from me, in response to the text. It's projection:
I want to make that clear. But there are hooks and lures in the medium that
merit special scrutiny.

When you talk about human rights: yes, it's important. And understanding
our relationship to the "AI imaginary" is the first step to understanding
and unraveling the harms that algorithms can perpetuate! GANs, for example,
can create human faces from scratch, but you will notice that it is worse
at generating images of black women than white men. That's the result of
the bias inherent in the training material; the same biases incorporated
into algorithmic surveillance tools. You can see it manifested there, just
as you could analyze a film for evidence of toxic ideologies.

But we have been thinking about how we watch films, and how we describe
dreams, for much longer than we have talked about how we read and extend
ourselves into generated images and text. Though maybe there are deeper
histories to this (Weizenbaum observed it with ELISA, and faulted the
people seduced by it, but this is something else, I would argue!)

Thanks for the thinking.
-eryk

On Tue, Jul 13, 2021 at 6:21 PM Alan Sondheim via NetBehaviour <
netbehaviour at lists.netbehaviour.org> wrote:

> Hi Eryk,
>
> The illusions are coming from you, I think, not from the text. And I'm not
> sure what you mean by "unraveled" - it would be easy enough, I think, to
> find out the connections otherwise. (Not entirely true: I belonged to a
> systems group at Brown University when I was a student there; they were
> studying the state of a computer X(t) at a given time, when it was running
> a complex program. By that they meant, given a particular component, what
> would it be doing exactly at t? And they felt the problem was intractable.
> So the connections in terms of *output* would be discernable, but,
> fundamentally, not internally. That doesn't create consciousness or
> conscience of course, just an indicator of complexity.)
> I wonder, here might be where I'm coming from, it's old hat, if the
> unconscious is not a luxury. Where I'm coming from, and I admit I'm a
> pessimist, stems from my own reading, the usual litany of countries and
> populations under terrible stress and the need to bear witness in relation
> to them, Nauru for example, Syria, Myanmar, the Uyghur, but similar in our
> own back yard. And how can, a serious question in return, how can the work
> being done with AI (in relation to institutions, silos and so on)
> contribute to healing or working with them? For the problem is increasing
> violently across the planet...
> You say "the 'generative' text must offer the same potential" - but must
> it? Cinema and dreams are anchored in bodies and ostensible labor;
> something is at stake. Cinema-image and dreamwork speak among bodies,
> through bodies. AI has its own seduction, part of which I think might be
> that of the hunter? Ferreting out meaning? At least when I've worked with
> my own small programming in that direction. But it's a second-level, the
> phenomenology and dreamwork, the unconscious of the hunter, the programmer,
> not the unconscious of the prey (in a sense, I know I'm exaggerating)...
> A question possibly - how might the work through AI help solve the
> problems above? (For me this is a real issue, keeping me awake at night:
> what are all of us doing, what are the solutions to the increasing hells
> animals, plants, everyone and everything are going through? What is the
> conceivable? Or that old absurd question, What is to be done?)
>
> Best, Alan, wondering what a sick AI would be, an AI with cancer or
> Covid...
>
> On Tue, Jul 13, 2021 at 1:34 PM Eryk Salvaggio via NetBehaviour <
> netbehaviour at lists.netbehaviour.org> wrote:
>
>> Sorry Alan, to be clear I don’t think machines have an unconscious
>> either. I should have been more precise: I’m talking about the tools of
>> interpretation for the machine’s output, especially if one approaches
>> generated text in a psychoanalytic framework (Freudian / Lacanian). There
>> are illusions of an unconscious there, and so my question is: what are we
>> interpreting when we do that? What connections in the neural net can be
>> unraveled through that process?
>>
>> Freud used to say that dream analysis worked to give meaning to dreams
>> from the process of describing them. Cinema can work in the same way. The
>> “generative” text must offer the same potential, because the act of
>> interpreting it is human, and unconscious. Do cinema and dreams bring
>> things out in different ways? If so, what might generative text bring that
>> is unique to itself as a medium?
>>
>> I want to dive deeper into your article here, it looks great. Thank you
>> for sharing!
>>
>> -e.
>>
>> On 13 Jul 2021, at 12:58 pm, Edward Picot via NetBehaviour <
>> netbehaviour at lists.netbehaviour.org> wrote:
>>
>> 
>> I'm listening to it now and loving it!
>>
>> Edward
>>
>> On 12/07/2021 00:53, Anthony Stephenson via NetBehaviour wrote:
>>
>>
>>
>>> Date: Sat, 10 Jul 2021 10:33:26 -0400
>>> From: Eryk Salvaggio <eryk.salvaggio at gmail.com>
>>> Subject: [NetBehaviour] The Uncanny Vibe
>>>
>>> Here's an experimental text. For a while now, I have been creating pop
>>> music in collaboration with machine learning systems in a project called
>>> The
>>> Organizing Committee <https://www.cyberneticforests.com/music>.
>>>  ...  What do you all think?
>>>
>>
>>
>> ░░░░░░░░░░░░▄▄░░░░░░░░░
>> ░░░░░░░░░░░█░░█░░░░░░░░
>> ░░░░░░░░░░░█░░█░░░░░░░░
>> ░░░░░░░░░░█░░░█░░░░░░░░
>> ░░░░░░░░░█░░░░█░░░░░░░░
>> ███████▄▄█░░░░░██████▄░░
>> ▓▓▓▓▓▓█░░░░░░░░░░░░░░█░
>> ▓▓▓▓▓▓█░░░░░░░░░░░░░░█░
>> ▓▓▓▓▓▓█░░░░░░░░░░░░░░█░
>> ▓▓▓▓▓▓█░░░░░░░░░░░░░░█░
>> ▓▓▓▓▓▓█░░░░░░░░░░░░░░█░
>> ▓▓▓▓▓▓█████░░░░░░░░░█░░
>> ██████▀░░░░▀▀██████▀░░░░
>>
>> --
>>
>> - *Anthony Stephenson*
>>
>>
>>
>> *http://anthonystephenson.org/ <http://anthonystephenson.org/> *
>>
>>
>>
>>
>> _______________________________________________
>> NetBehaviour mailing listNetBehaviour at lists.netbehaviour.orghttps://lists.netbehaviour.org/mailman/listinfo/netbehaviour
>>
>>
>> _______________________________________________
>> NetBehaviour mailing list
>> NetBehaviour at lists.netbehaviour.org
>> https://lists.netbehaviour.org/mailman/listinfo/netbehaviour
>>
>> _______________________________________________
>> NetBehaviour mailing list
>> NetBehaviour at lists.netbehaviour.org
>> https://lists.netbehaviour.org/mailman/listinfo/netbehaviour
>>
>
>
> --
> *=====================================================*
>
> *directory http://www.alansondheim.org <http://www.alansondheim.org> tel
> 718-813-3285**email sondheim ut panix.com <http://panix.com>, sondheim ut
> gmail.com <http://gmail.com>*
> *=====================================================*
> _______________________________________________
> NetBehaviour mailing list
> NetBehaviour at lists.netbehaviour.org
> https://lists.netbehaviour.org/mailman/listinfo/netbehaviour
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://lists.netbehaviour.org/pipermail/netbehaviour/attachments/20210713/54c8e90f/attachment.htm>


More information about the NetBehaviour mailing list