A-Infos (en) Italy, FAI, Umanita Nova #21-26 - It applies itself (even too much), but it's not intelligent. The many implications of artificial intelligence (ca, de, it, pt, tr)[machine translation]

We wanted to discuss with LaTram, a cartoonist and illustrator, the 
direct impact that the use of artificial intelligence can have on some 
typically creative professions, as well as the general environmental 
implications of this technology's increasingly widespread use. Below is 
the interview. --- Q. Tell us about your work as a cartoonist and how 
you've seen it change in recent years with the development of 
information technology. A. I've been doing this job for about twenty 
years, as a cartoonist and illustrator, and I also work in graphic 
journalism. The profession has changed dramatically, in general, and has 
taken advantage of the many technological opportunities that have arisen 
over time. The introduction of artificial intelligence, however, has 
brought about a huge and massive change, constantly and rapidly 
evolving, making it difficult to metabolize and manage the impact. From 
one month to the next, increasingly updated and more powerful versions 
of artificial intelligence are created that work on ever-larger amounts 
of data sets and that, in an increasingly refined way, are able to 
understand prompts and better interpret requests.
Personally, I work analogically, that is, by drawing and coloring my 
works by hand, but I still have to put my productions online to make 
them known to publishers, as does the entire creative world to which I 
belong. Well, putting even a craft work on the Internet subjects it to 
the scrutiny of artificial intelligence.

Q. Could you explain, in broad terms, how the AI system works in the 
specific field of images?
A. In order to function well, artificial intelligence systems must be 
trained on an organized and structured collection of data, the data set 
, which software can analyze to perform the search and process the 
request. The larger the data set , the higher the quality of what the AI 
produces. The more you feed and "feed" the AI, the more it learns and 
then produces, in the visual field, highly detailed and professional 
images. This means that, in the creative field, working with 
illustrations for example, it can receive very specific instructions and 
reproduce exactly what is contained in the request. The internet is full 
of images, uploaded by illustrators themselves to give visibility to 
their work. With the advent of artificial intelligence, this quantity of 
data is available for use without any regulation to date. AI, at the 
moment, can use anything available on the internet for training: 
original images, edited images, photographs, etc.
A recent case in Spain provides a paradoxical and emblematic example of 
the indiscriminate use of AI. The poster for the Valencia Comics 
Festival was "commissioned" to artificial intelligence, which, based on 
the information received, created a poster virtually identical to the 
authentic one produced for a literature festival by Paco Roca, currently 
Spain's most famous cartoonist. A controversy ensued, the copied poster 
was withdrawn, and so on, but the case clearly demonstrates that if you 
commission a poster "in the Paco Roca style," you can end up with one 
identical, perhaps without even realizing it, with all the consequences 
that entails. Or you can commission a poster with certain 
characteristics, even without using a specific author's prompt, and the 
artificial intelligence autonomously decides that Paco Roca's style is 
best suited for that poster , without the client even knowing that it is 
Paco Roca's work... In short, it's crazy.
Furthermore, changing just a tiny detail in an image is enough to make 
it no longer your image but one that resembles it. This means that the 
originality of the creation is lost, as is all the research that 
preceded the creation of the work that is, the design, the composition, 
the choice of graphic language appropriate to the message to be 
communicated, the choice of palette, etc. AI can't yet reproduce this 
entire process, but it diligently creates an image that contains 
everything in the prompt, thus flattening and devaluing the work and, 
above all, using it improperly.

Q. From your description, it appears that not only is creative ability 
at risk, but also the very work of a significant number of workers in 
the graphic design sector.
A. Exactly. The issue of AI is obviously deeply felt by creative 
workers. To try to introduce protections, EGAIR (European Guild 
Artificial Intelligence Regulation) was founded, an organization that 
brings together workers from various fields, not just illustrators and 
cartoonists, but all image workers (graphic designers, advertising, etc.).
The aim is to demand institutional regulation of the use of images 
online, and steps have already been taken in the Senate and at the 
European level.
Among the various requests is the introduction of author consent. 
Currently, it is not possible to deny consent to the use of images as 
training sets, i.e., for training artificial intelligence; introducing 
some sort of consent could already be a first form of protection. 
Another request is that authors' names or pen names cannot be included 
in the prompt, thus prohibiting explicit commissioning of an image from 
AI, specifying that it reproduce the style of that particular author. 
Furthermore, it should be mandatory to include the words "entirely 
created with AI" or "partially created with AI" on all images processed 
by AI. Finally, there is also a lack of a clear distinction between 
copyrighted material, where the author is clearly acknowledged and 
identified, and "public domain" material: it is possible that 
copyrighted material may be used by AI as well. There is no clear 
classification of images found on the internet, and even copyright 
protection is insufficient, sometimes because the very definition of 
copyright is currently subject to potentially confusing interpretations.
EGAIR's proposal is to act at the institutional level, but there are 
other areas where the same problem could be addressed through 
sensitivity and solidarity. I'm referring to the fact that, 
unfortunately, the indiscriminate use of original images, through 
artificial intelligence processing, is widely practiced even in spaces 
for movement and in environments that should demonstrate attention to 
workers' rights, because that's what we are. A category that, moreover, 
isn't even protected by unions, national contracts, paid holidays, 
contributions, or sick leave. Nothing.
If we want to have a comprehensive view, if we truly want to bring about 
social change and overcome exploitation, we cannot ignore the needs 
ofThe entire workforce. It wouldn't take much to show concrete 
solidarity by avoiding the use of artificial intelligence when creating 
graphics, posters, and promotional materials, especially since many 
artists would be and are happy to lend their art, even for free. I've 
done it, and so have many of my colleagues.

Q. Beyond the specific impact of AI on the use of images, on a more 
general level, I'd like to ask you: is artificial intelligence really 
sustainable, as some claim, emphasizing the potential for optimizing and 
reducing energy consumption related to the ability to process such a 
large amount of environmental data?
A. I'm not an expert on the green transition or anything like that, 
which is all tied to the need to reshape capital, but, according to 
numerous official sources accessible to anyone, it seems to me that AI 
is anything but sustainable. Those who claim otherwise maintain that in 
a relatively short time we will develop a technology capable of 
minimizing the environmental impact of the data centers necessary for AI 
to function. In reality, the environmental impacts are there, they are 
there now, and they are irreparable and multiple; they concern 
electricity, soil, water, and critical minerals, despite all the 
reassurances and downplaying.
According to the recent United Nations report on the impact of 
artificial intelligence, for example, by 2030 it is estimated that 
electricity consumption could reach up to 945 terawatt hours per year, 
that is, triple the combined annual electricity consumption of Pakistan, 
Bangladesh, and Nigeria (source: Luca Aterini for Greenreport 
www.greenreport.it/news/scienza-e-tecnologie/62110-intelligenza-artificiale-consumi-naturali-nel-2030-lai-si-berra-lacqua-di-1-3-miliardi-di-persone 
) . And it's not "just" an electricity problem. To train and operate AI, 
enormous data centers are needed , colossal structures the size of 
several soccer fields, structures that house tens of thousands of 
ultra-high-powered processors. This means direct land consumption, which 
involves expropriation and desertification of large geographical areas 
where the infrastructure needed for AI production is located. It should 
also be remembered that artificial intelligence chips are powered by 
critical minerals: these minerals are typically found in geographical 
areas of so-called developing countries, where mining is less 
controlled, where land exploitation, drilling, and concessions to 
multinationals have overtly colonial characteristics and are often a 
contributing factor to wars. This too is land exploitation, but it's not 
"only" that: the phenomenon is intertwined with colonialism, 
militarization, and the exploitation of poverty. The enormous facilities 
intended for training and operating artificial intelligence are 
generally located near waterways or aquifers, because the expanses of 
processors present in gigantic data centers

They require enormous quantities of water for cooling.
Water to produce energy, water to cool the constantly working 
processors; quantities of water absorbed by AI's operation are then 
released into the environment, causing pollution, primarily of 
groundwater, with serious damage to the water supply of the surrounding 
population and with overall repercussions. The water footprint 
calculated for AI is enormous: it is predicted that with the 
increasingly widespread diffusion of artificial intelligence, global 
water consumption will be equivalent to the annual needs of 1.3 billion 
people.

Q. In your opinion, how much awareness is there of the fact that data 
centers generating artificial intelligence have such a strong 
environmental, as well as social, impact?
A. In my opinion, awareness is still lacking, and we see this, as 
already mentioned, also in the ease with which artificial intelligence 
is used even in movement circles, precisely those that should be paying 
attention to the problems this technology triggers, which are many and 
deeply interconnected: from environmental pollution, to forms of 
neocolonialism in the supply of energy resources, to the elimination of 
workers' rights.
But awareness is also lacking in the individual relationship with AI. A 
request to artificial intelligence consumes 10 times more energy than a 
simple search on a simple search engine . And while the spread of AI 
represents a huge profit opportunity for capital, we must not forget 
that it is our personal interaction, rather than the training of AI, 
that determines most of the work of these computers: a conversation with 
AI that involves continuous interaction, such as "tell me this thing, 
tell me more about it, etc." It consumes more than 200 times more 
electricity than a simple online search. If AI is used to generate an 
image, the consumption is 1,450 times higher, while it's 200,000 times 
higher to create a short video . So, it's us, with our often pointless 
interactions, who fuel this vicious cycle.

Q. Returning to the specific issue of using images with AI and therefore 
to your field of activity: a few decades ago, the experiences created by 
independent media and the possibilities offered by the internet for 
downloading videos and music were seen as a positive extension of access 
to information and the free enjoyment of cultural products. Couldn't 
artificial intelligence represent a horizontal process of this kind today?
A. Artificial intelligence could certainly be used more virtuously, but 
the scale the phenomenon is taking makes it very difficult, beyond an 
attempt at regulation, to envision a "self-management" of the AI 
technological tool.
With artificial intelligence, we have a manipulated generation, derived 
from the automatic processing of original images stolen online from real 
authors. This doesn't seem like a great achievement, either democratic 
or horizontal.
What need is there? With a sheet of paper and a pencil, or a computer 
and a graphics program, anyone can create something, expressing what 
they want, using a personal language consistent with what they want to 
express. Creativity is nothing more than an urgency; it's completely 
detached from the final aesthetic result, it has nothing to do with it. 
It's not that something must be beautiful and therefore made by an AI to 
be valid; this, in my opinion, is another of those mechanisms that 
capitalism has instilled in us, whereby if something isn't perfect, 
isn't beautiful, isn't wonderful, isn't shiny, doesn't have a flawless 
surface, then it's not valid. In reality, anything you feel the urge to 
say is valid; that's how it will surely arrive. And that's how 
communication truly happens.

edited by PN

https://umanitanova.org/si-applica-anche-troppo-ma-non-e-intelligente-le-molte-implicazioni-dellintelligenza-artificiale/
_________________________________________
A - I N F O S  N E W S  S E R V I C E
By, For, and About Anarchists
Send news reports to A-infos-en mailing list
A-infos-en@ainfos.ca
Subscribe/Unsubscribe https://ainfos.ca/mailman/listinfo/a-infos-en
Archive: http://ainfos.ca/en