txttletale:

i would simply respond that almost none of that is true.

‘depleting the global power and water supply’

something i’ve seen making the roudns on tumblr is that chatgpt queries use 3 watt-hours per query. wow, that sounds like a lot, especially with all the articles emphasizing that this is ten times as much as google search. let’s check some other very common power uses:

• running a microwave for ten minutes is 133 watt-hours
• gaming on your ps5 for an hour is 200 watt-hours
• watching an hour of netflix is 800 watt-hours

and those are just domestic consumer electricty uses!

• a single streetlight’s typical operation 1.2 kilowatt-hours a day (or 1200 watt-hours)
• a digital billboard being on for an hour is 4.7 kilowatt-hours (or 4700 watt-hours)

i think i’ve proved my point, so let’s move on to the bigger picture: there are estimates that AI is going to cause datacenters to double or even triple in power consumption in the next year or two! damn that sounds scary. hey, how significant as a percentage of global power consumption are datecenters?

1-1.5%.

ah. well. nevertheless!

what about that water? yeah, datacenters use a lot of water for cooling. 1.7 billion gallons (microsoft’s usage figure for 2021) is a lot of water! of course, when you look at those huge and scary numbers, there’s some important context missing. it’s not like that water is shipped to venus: some of it is evaporated and the rest is generally recycled in cooling towers. also, not all of the water used is potable–some datacenters cool themselves with filtered wastewater.

most importantly, this number is for all data centers. there’s no good way to separate the 'AI’ out for that, except to make educated guesses based on power consumption and percentage changes. that water figure isn’t all attributable to AI, plenty of it is necessary to simply run regular web servers.

but sure, just taking that number in isolation, i think we can all broadly agree that it’s bad that, for example, people are being asked to reduce their household water usage while google waltzes in and takes billions of gallons from those same public reservoirs.

but again, let’s put this in perspective: in 2017, coca cola used 289 billion liters of water–that’s 7 billion gallons! bayer (formerly monsanto) in 2018 used 124 million cubic meters–that’s 32 billion gallons!

so, like. yeah, AI uses electricity, and water, to do a bunch of stuff that is basically silly and frivolous, and that is broadly speaking, as someone who likes living on a planet that is less than 30% on fire, bad. but if you look at the overall numbers involved it is a miniscule drop in the ocean! it is a functional irrelevance! it is not in any way 'depleting’ anything!

'stopping us from thinking or writing critically’

this is the same old reactionary canard we hear over and over again in different forms. when was this mythic golden age when everyone was thinking and writing critically? surely we have all heard these same complaints about tiktok, about phones, about the internet itself? if we had been around a few hundred years earlier, we could have heard that “The free access which many young people have to romances, novels, and plays has poisoned the mind and corrupted the morals of many a promising youth.”

it is a reactionary narrative of societal degeneration with no basis in anything. yes, it is very funny that laywers have lost the bar for trusting chatgpt to cite cases for them. but if you think that chatgpt somehow prevented them from thinking critically about its output, you’re accusing the tail of wagging the dog.

nobody who says shit like “oh wow chatgpt can write every novel and movie now. yiou can just ask chatgpt to give you opinions and ideas and then use them its so great” was, like, sitting in the symposium debating the nature of the sublime before chatgpt released. there is no 'decay’, there is no 'decline’. you should be suspicious of those narratives wherever you see them, especially if you are inclined to agree!

plagiarizing human artists

nah. i’ve been over this ad infinitum–nothing 'AI art’ does could be considered plagiarism without a definition so preposterously expansive that it would curtail huge swathes of human creative expression.

AI art models do not contain or reproduce any images. the result of them being trained on images is a very very complex statistical model that contains a lot of large-scale statistical data about all those images put together (and no data about any of those individual images).

to draw a very tortured comparison, imagine you had a great idea for how to make the next Great American Painting. you loaded up a big file of every norman rockwell painting, and you made a gigantic excel spreadsheet. in this spreadsheet you noticed how regularly elements recurred: in each cell you would have something like “naturalistic lighting” or “sexually unawakened farmers” and the % of times it appears in his paintings. from this, you then drew links between these cells–what % of paintings containing sexually unawakened farmers also contained naturalistic lighting? what % also contained a white guy?

then, if you told someone else with moderately competent skill at painting to use your excel spreadsheet to generate a Great American Painting, you would likely end up with something that is recognizably similar to a Norman Rockwell painting: but any charge of 'plagiarism’ would be absolutely fucking absurd!

this is a gross oversimplification, of course, but it is much closer to how AI art works than the 'collage machine’ description most people who are all het up about plagiarism talk about–and if it were a collage machine, it would still not be plagiarising because collages aren’t plagiarism.

(for a better and smarter explanation of the process from soneone who actually understands it check out this great twitter thread by @reachartwork)

today’s students are worried they won’t have jobs because of AI tools

i mean, this is true! AI tools are definitely going to destroy livelihoods. they will increase productivty for skilled writers and artists who learn to use them, which will immiserate those jobs–they will outright replace a lot of artists and writers for whom quality is not actually important to the work they do (this has already essentially happened to the SEO slop website industry and is in the process of happening to stock images).

jobs in, for example, product support are being cut for chatgpt. and that sucks for everyone involved. but this isn’t some unique evil of chatgpt or machine learning, this is just the effect that technological innovation has on industries under capitalism!

there are plenty of innovations that wiped out other job sectors overnight. the camera was disastrous for portrait artists. the spinning jenny was famously disastrous for the hand-textile workers from which the luddites drew their ranks. retail work was hit hard by self-checkout machines. this is the shape of every single innovation that can increase productivity, as marx explains in wage labour and capital:

“The greater division of labour enables one labourer to accomplish the work of five, 10, or 20 labourers; it therefore increases competition among the labourers fivefold, tenfold, or twentyfold. The labourers compete not only by selling themselves one cheaper than the other, but also by one doing the work of five, 10, or 20; and they are forced to compete in this manner by the division of labour, which is introduced and steadily improved by capital. Furthermore, to the same degree in which the division of labour increases, is the labour simplified.

The special skill of the labourer becomes worthless. He becomes transformed into a simple monotonous force of production, with neither physical nor mental elasticity. His work becomes accessible to all; therefore competitors press upon him from all sides. Moreover, it must be remembered that the more simple, the more easily learned the work is, so much the less is its cost to production, the expense of its acquisition, and so much the lower must the wages sink – for, like the price of any other commodity, they are determined by the cost of production. Therefore, in the same manner in which labour becomes more unsatisfactory, more repulsive, do competition increase and wages decrease”

this is the process by which every technological advancement is used to increase the domination of the owning class over the working class. not due to some inherent flaw or malice of the technology itself, but due to the material realtions of production.

so again the overarching point is that none of this is uniquely symptomatic of AI art or whatever ever most recent technological innovation. it is symptomatic of capitalism. we remember the luddites primarily for failing and not accomplishing anything of meaning.

if you think it’s bad that this new technology is being used with no consideration for the planet, for social good, for the flourishing of human beings, then i agree with you! but then your problem shouldn’t be with the technology–it should be with the economic system under which its use is controlled and dictated by the bourgeoisie.