Re: <nettime> The Case for Gouging Out The Eye of The Cyclopes !

[franz schaefer via nettime-l <nettime-l@lists.nettime.org>]

Dear XLterrestrials, dear Felix,


> "And then there is the apprehension from the world of non-tech humans –
> climate-change activists distraught about the carbon footprint of the AI
> industry; students and workers who see mass unemployment and blank
> futures; artists, musicians and writers, like myself, who can sense the
> centrifugal churning that is sucking the beating heart from our bodies and
> drawing the steel threads of societal control together into one clenched,
> mechanical fist."
>  
> https://www.equator.org/articles/arundhati-roy-artificial-intelligence

Thanks for that link. I am really a big fan of Roy and often quote her. Also
the essay is quite good – it already acknowledges a lot of problems that are
about to come where others are still in denial.

I also like the part where she criticises Bernie Sanders. While Sanders is
one of the few on the left who really warns about AI, his suggestion to
nationalize AI companies only in the US was something worth criticising. AI
is a problem for humanity and not for the US. AI is built on all of
humanity's knowledge and thus must belong to all of us. So we need to
demand that it is open sourced. And in general it is sad to see that
nationalism and patriotism is so prevalent even in the left in the US.

But there is a part where I have a problem: Where she basically repeats
that story we heard for years now: that when there is talk about the
dangers of AI that this is mostly marketing. She writes:

  »Why would Whittaker and Gebru and their very specific indictments worry
  the oligarchs more than the prophets of all-out extinction like the AI
  scientist-turned-critic Geoffrey Hinton, or Jacob Coxon?  Perhaps because,
  like a big fireworks display, Hinton and Coxon’s warnings add frisson to the
  atmospherics.  They end up augmenting the oligarchs’ power play, which, when
  their companies go public, as they are set to do soon, will serve them very
  well indeed.  Their valuations will depend on how sophisticated – and, in
  the case of AI weapons systems, how lethal – their models are. « A. Roy
  from the article linked above

Now she does not draw the conclusion that the warnings are only marketing. 
But this is one of the lines of reasoning that has supported the AI-Denial
now for years.  And thus it is one of the reasons why the left is so late
with this topic.  And moreover: I still meet people who cling to this straw.

Trade Unions? One would assume they are all up in arms. After all when AI
replaces us at the work place we lose our ability to go on strike. But no.
The best you can get is some lukewarm recommendations for some regulations of
AI in the workplace. Well, all nice and good as long as there is a workplace
where people work.

Felix wrote:

> But it might be simply a normal technology, one that needs to work under
> normal economic constraints and (eventually) within biophysical
> boundaries.

We could hope that. If we actually manage to tame the cyclops. But I am a
pessimistic here: With the ignorance and denial on one hand and the sheer
speed of this exponential development on the other hand the future looks
rather bleak. Now, as I tried on my previous posting on this topic, it is
important to not fall into gloom and despair.

And I do not see the economic constraints: Again: If AI works, then its
worth is all of the global economy: 120T$.  Also the biophysical boundaries
are not really stopping this on its own.  You say the training of models now
takes 400 times more energy than the first ChatGPT models.  On the other
hand the energy per token went down by a factor of 200 in the last 3 years. 
At some point AI will be just "good enough" for many applications and not
much further training needed while it can be optimized to take less power
for the majority of tasks.  There is this company »Taalas« that was recently
bought by AMD: They can hard-code AI models on chip (which means the model
is baked in and cannot be replaced), but this again offers an order of
magnitude in cost- and power-saving.  So even if we manage to stop them
building new data centers (as we absolutely should), it will not stop the
AI.



mond


-- 
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
   .                Franz Schaefer       GPG: 3774ECD160719558
  ..              +43 699 106 14 590    Fingerprint: 5025 A74A
  ...             schaefer@mond.at         01DF F2AE 75E9 57C8
    ...          http://www.mond.at/       3774 ECD1 6071 9558

-- 
# distributed via <nettime>: no commercial use without permission
# <nettime> is a moderated mailing list for net criticism,
# collaborative text filtering and cultural politics of the nets
# more info: https://www.nettime.org
# contact: nettime-l-owner@lists.nettime.org