@andykaufmanisnotdead on Tumblr

andykaufmanisnotdead reblogged

ilterzouomo

gatabella

Lauren Bacall, 1940s

andykaufmanisnotdead reblogged

sowhatifiliveinfukuoka

pentangeli

North by Northwest (1959)

andykaufmanisnotdead reblogged

nicolacava

andykaufmanisnotdead reblogged

sandmandaddy69

#Porto #Been there

andykaufmanisnotdead reblogged

astromechapunk

s-pyder

andykaufmanisnotdead reblogged

lablablaologa

vintageeveryday

An exhausted young couple during an all-night jazz session at the Royal Albert Hall in London, 1957. Photographed by Erich Auerbach.

andykaufmanisnotdead reblogged

7-a-m

marinella-ela

andykaufmanisnotdead reblogged

bloodmoses

leatherandmossprints

‘Three Cicadas’ by Vincent Van Gogh, c. 1889.

andykaufmanisnotdead reblogged

bloodmoses

deepseametro

Anonymous asked:

I dont get the environmental argument against ai bc look at video games or FUCKING GOOGLE and how much energy and water they use it feels disingenuous

deepseametro answered:

it's all sensationalist headline BS. that's not to say ML models don't use a lot of resources in their training - they do - but as you say - it's not anymore (in fact, less, in a lot of cases) than other industries, even other tech industries. i understand being dismayed water is being used for something like this but it's also used for, like, every server building in existence - personally speaking i'm more concerned about the use of water for things like unneccessary agriculture in areas that don't need it - 3/4s of water use in Arizona goes to agriculture (the vast majority of which is exported). yknow, where we're at 15+ years of one of the worst droughts here in recorded history!!!!! but yeah yeah the water coolant is the Big Problem

catpillmeowcel

The thing is that the water coolant part isn't, the worst part. These data centers are NOT constantly using water and generally only need to be filled once as they used closed circuit cooling, which is 10x more environmentally friendly than the other options. (Using shittons of electricity to constantly power hundreds of fans to keep it from melting itself)

It absolutely uses a lot of water but only really for the one time it needs to be filled:

Close Coupled Cooling - Wikipedia

en.wikipedia.org

The MAJOR issue that most mainstream media conveniently likes to ignore is the literal slave labor being used to train datasets.

The Exploited Labor Behind Artificial Intelligence | NOEMA

Supporting transnational worker organizing should be at the center of the fight for “ethical AI.”

NOEMA

That AI You’re Using Was Trained By Slave Labor, Basically

AI has a very dirty secret: that the people hired to train it are making literal pennies on the dollar, which is tantamount to slave labor.

Futurism

Exclusive: The $2 Per Hour Workers Who Made ChatGPT Safer

A TIME investigation reveals the difficult conditions faced by the workers who made ChatGPT possible

TIME

deepseametro

(this a great addition, thank you) - but i wanna take a second to address a sentiment i'm seeing in the tags in response to learning this:

so like. hey. it's great to be informed of this, and it's good that this pisses you off. but this is sort of my exact problem - whether you're meaning to or not, you're directing hate at ML technology specifically and not, like, the fact this is how the majority of things we use in the west are made.

like, do you say you "hate factory-built products", or do you hate the way factories are run overseas? do we hate imported agricultural products, or do we hate the systems in place that currently allow those products to get to us at such a low price?

nearly everything you use, consume, and interact with was made with slavery/slavery-like conditions. this is not an excuse of the fact the ML industry is participating in that. i just want to challenge the idea that machine learning or the industry is in any way uniquely evil, uniquely exploitative, uniquely powered by severe and troubling worker exploitation. (even the tech industry at large is in no way unique in this regard)!

there are ways to ethically make these things. there's a way to ethically produce nearly everything! if we hyperfocus on the specific conditions for this specific industry, it's going to ultimately do very little to stop this kind of worker abuse. it's not AI. it's the world, it's global capitalism, it's imperialism, and it's been ravaging the majority of the world's working class for a long time.

we don't generally consider people "bad people" for buying avocados or knitted sweaters or EV batteries - we gotta stop applying that logic to machine learning, too. oppose the industries!