Prof. Emily M. Bender(she/her)This is a great summary by @SashaMTL of the environmental and human costs of so-called "AI" technology.
>>
"For instance, with ChatGPT, which was queried by tens of millions of users at its peak a month ago, thousands of copies of the model are running in parallel, responding to user queries in real time, all while using megawatt hours of electricity and generating metric tons of carbon emissions. It’s hard to estimate the exact quantity of emissions this results in, given the secrecy and lack of transparency around these big LLMs."
>>
"it’s difficult to carry out external evaluations and audits of these models since you can’t even be sure that the underlying model is the same every time you query it. It also means that you can’t do scientific research on them, given that studies must be reproducible."
>>
@SashaMTL Is the horse out of the barn? Do we just have to stand by and watch this go down?
Indeed not. We've collectively handled other sources of pollution (e.g. lead in gasoline, CFCs harming the ozone layer) before and we can do it again.
Let's heed @SashaMTL 's call to get engaged with the regulatory process!
p.s. re "the horse is out of the barn". That metaphor is used to express helplessness, a "there's-nothing-we-can-do-now" attitude. But do people with escaped horses really say "Oh well, was nice knowing ya horsey"?
I'd guess probably not.
Mrs RavenNaw, my grandpa would sic all the kids on the horse and hilarity would ensue.
Because he knew the horse would always return on its own at dinner time.
Sam Wade@emilymbender @SashaMTL Probably not. Once a cat’s out of a bag, on the other hand, who can really be bothered?
Argot_Robbie@emilymbender @SashaMTL It's like boiling frogs, except actual frogs jump out of water as it's heated.
Or the definition of insanity, which has nothing to do with repeating an action with the expectation of a different result, since most insane people don't do that, and also since some actions have different results when repeated.
Tracy HallAll it means is "stop wasting time discussing the barn door latch and go get the damn horse"
Lady Artemis@video_manager @emilymbender @SashaMTL
I love this meaning to the phrase. This is the best. Someone put this on a billboard.
Steve Dodge@emilymbender @SashaMTL
I have a similar issue with the pendulum metaphor for social dynamics—whenever someone says that the pendulum has swung too far in some direction, they are basically saying that their preferred social arrangement is the natural, stable equilibrium, and that any attempt to change that is futile
I have a similar issue with the pendulum metaphor for social dynamics—whenever someone says that the pendulum has swung too far in some direction, they are basically saying that their preferred social arrangement is the natural, stable equilibrium, and that any attempt to change that is futile
Thomas@emilymbender @SashaMTL Few things in the social are truly reproducible (i.e., stochastic processes). Just because version controlled software, known parameters and deterministic behaviour regarding start conditions *can* provide mechanistic reproducibility this is by no means a precondition for scientific research. In the negative "evidence is evidence" (i.e., credible evidence of one occurrence can falsify a theory).
@emilymbender @SashaMTL It hasn't been my intention to defend OpenAI. My statement was about a simple observation regarding scientific research and reproducibility. On the contrary, scientists should always try to falsify what's claimed to be true, and they should lay bare any claim that's not falsifiable. Unfortunately 500 chars text is quite limiting 🙂
Alex Weatherall@emilymbender @SashaMTL It's also hard to estimate the exact *quality* of the emissions from ChatGPT since it's essentially making it all up as it goes along.
Christian Lynbech@emilymbender @dekkzz76 @SashaMTL All of this illustrates as well as anything how amazing a computer, the human brain is. It does qualitively much more than ChatGPT and its ilk, in only a tiny fraction of the energy consumption.
Some seem to think that LLMs will radically change how everything is done. I wonder if that will actually scale, are there even enough computing power to give everybody their own LLM?
Bornach@mapcar @emilymbender @dekkzz76 @SashaMTL
That depends on whether you can give everyone their own 4GB laptop. That seems to be the minimum hardware requirements for the smallest LLaMa AI although someone did manage to run Alpaca on a Raspberry Pi but generating one token per minute you might be waiting quite some time for an answer
https://medium.com/@martin-thissen/llama-alpaca-chatgpt-on-your-local-computer-tutorial-17adda704c23
That depends on whether you can give everyone their own 4GB laptop. That seems to be the minimum hardware requirements for the smallest LLaMa AI although someone did manage to run Alpaca on a Raspberry Pi but generating one token per minute you might be waiting quite some time for an answer
https://medium.com/@martin-thissen/llama-alpaca-chatgpt-on-your-local-computer-tutorial-17adda704c23
@bornach @emilymbender @dekkzz76 @SashaMTL It could certainly be that coming improvements in hard- and soft-ware will improve the energy consumption of AI but it probably will not help, as I incidentally just learned from Molly Woods newsletter.
If you worry about climate, sustainability and the future of the planet, this is a highly recommended read (and very relevant to the original post): https://www.mollywood.co/p/how-ai-kills-us-energy-use
