For people who it isn’t as black and white as Kate’s option of joining a committee, what do you suggest doing for the environment. I understand your take on how being vegan or not using ChatGPT would make less change, I mean it all makes sense and I agree. But it does leave me wondering thinking, what can I do? Of course I could still be vegan or try lessen my carbon footprint but that won’t be doing much. What are some suggestions that you would say would help the environment more efficient. Preferably if you could come up with a variety of solutions it would be great, but I do understand that I should maybe do my on research on the topic instead of trying to quickly find answers.
It was funny seeing Ketan Joshi's name pop up again here: he's a deeply disingenuous commentator who mostly just looks for excuses to trash on people or ideas not in his clique. His whole exchange w/ Jordan Shanks some years back really highlighted the kind of person he is.
Can you link to the Jordan Shanks exchange? Don't really know Joshi's deal, he's just been sharing my stuff with extremely sloppy negative commentary that a high school student could poke holes in recently.
Absolutely understand, which is why I kinda held back on linking anything until you asked. It's incredibly niche and scattered which just kind of makes it a mess to dig through unless you're super into it.
Another excellent post, but... you excuse for people who believe factual lies about generative AI and the environment, and use it to attack generative AI companies or users as having been "tricked". I am not sure I agree - one should do their due diligence before going out of their way to criticise others.
Strongly agree. I can't imagine why people would care so much about one small use of power and a low contributor to climate change versus causes that are magnitudes larger.
Seeing everything as Horrible is what gives them their identity. ("I live in The End Times!") They have their Sacred Texts (Doomer blogs, etc.). Don't argue with crazy people - contribute to a constructive conversation with people who actually care about facts and want to make a difference. IMHO.
First, about peer review: I have been a researcher for over 15 years, I have been involved in dozens of published papers and have reviewed dozens more. You claim:
"A claim only needs to be peer reviewed if it’s something new that we don’t already know that hasn’t been verified by other studies. In these posts, I’m taking the numbers that we’ve been given by peer-reviewed studies or that are our best guesses, and saying “if these are true, this is how the energy costs of ChatGPT compare to other things we do.”"
This is not true. Review papers - scoping, narrative or systematic - as well as meta-analyses do exactly this: They collate research across a certain research question and provide a picture about the level of knowledge, either quantitatively or qualitatively. All of these go through just as rigorous peer review as the original studies, empirical or otherwise (which I know because I have both been a co-author on as well as peer-reviewed these). This is necessary because even if you base your argument and/or calculations on peer-reviewed literature, you may still be wrong! Your argument may be based on a fallacy or your calculations may be wrong.
The same is true about your post. What you're doing here is basically a review of the "literature". And you may be wrong in any of your conclusions. However, you can't even base most of your calculations on peer-reviewed findings! As you state yourself, much of the data is what OpenAI, Google et al. have deigned to put out, assumptions other people have made etc. This makes the conclusions even more contingent, and the premises, calculations and arguments even more in need of external verification.
Now, academic peer review is not a panacea and there has been a lot of well-founded criticism of the process and I personally appreciate efforts to make it more transparent and accountable. I also think post-publication peer review is a good idea - which is *in principle* similar to comment sections such as these, with the difference that it, too, is transparent and accountable, i.e. named individuals with known qualifications.
In any case, I don't think your defense about why this doesn't need peer review holds up. I appreciate that you have corrected some parts of your original post *where you agreed with the criticism*. In an actual scientific journal, peer criticism, especially on contentious topics, is often published alongside the original paper. Here, you have to scroll down to the comment section, and those comments that got the original author's (your) approval get a special mark above it. All of these set it apart from a scientific publication. That's okay! It's not a scientific publication and it doesn't need to be! But again, the criticism is valid, in my opinion.
The second point is about your Kate, Bob and Freddy example. This example is misleading and unfair, and I think you know this. Anyone who engages with climate change activism knows that issues are always both individual and systemic, and here you are confusing these levels. Anyone can decide for themselves to be a vegan (as you do) or use or not use ChatGPT. We have more or less complete control over these decisions. No-one can decide by themselves whether a nuclear plant keeps running. Not even 500 well-intentioned people have any definitive power to decide this. They can campaign for it, they can call their representatives, communicate with the decision-makers at the energy providers etc., but there is no way of knowing if this will achieve the goal. And even if it does, it is completely disingenuous to then put it as "These 500 people have now individually saved x amount of kWh/CO2e or whatever." (This is in addition to a, in my view, simplistic perspective what it takes to "keep a nuclear plant running". I personally think that it was a mistake to phase out nuclear before coal in my home country of Germany. However almost all nuclear plants would *still* have gone out of service because many would have needed extensive repair and maintenance, which would have been very, very resource-intensive and would have negated any environmental gains for quite a while.)
This, in my opinion, points to a further problem of your framing as a whole: Framing the problem as "making individual ChatGPT out to be a problem distracts from the real issues" is a fallacy. This assumes that we can only care about a limited number of (or a single) issues at once. If Bob is serious about the environment, he is probably not only a vegan but *also* limits his use of individual transport, checks the energy rating of the appliances he uses and, yes, limits how much he uses ChatGPT. In campaigning, activists have never focused on any one issue (food production, clothing, transport, etc.) exclusively.
This framing is in my opinion analogous to the argument that many climate distractors use in my country when they say "Why should we be the ones leading the way on climate action when others don't pull their weight? We only produce 2 % of all CO2e, that's nothing compared to China and the USA!" Sure, ChatGPT queries may not look like much. But most of them are *additional* to what we're doing anyway. Of course we can say that doing an extensive query would only need just as much energy as microwaving a bowl of popcorn. But then we haven't microwaved the popcorn.
You may disagree with my criticism of the general framing, but I hope the criticism of the Kate, Bob and Freddy example is clear: This is apples and oranges and, in my view, weakens your argument instead of strengthening it.
For people who it isn’t as black and white as Kate’s option of joining a committee, what do you suggest doing for the environment. I understand your take on how being vegan or not using ChatGPT would make less change, I mean it all makes sense and I agree. But it does leave me wondering thinking, what can I do? Of course I could still be vegan or try lessen my carbon footprint but that won’t be doing much. What are some suggestions that you would say would help the environment more efficient. Preferably if you could come up with a variety of solutions it would be great, but I do understand that I should maybe do my on research on the topic instead of trying to quickly find answers.
This is more contentious but I'm personally pretty into effective environmentalism, the goal is to find promising levers to pull for the climate to have an outsized impact. They have a good overview of advice here: https://www.effectiveenvironmentalism.org/articles/take-action/career-advice
Founders Pledge also has great overviews of philanthropy and projects happening in climate that could be really high impact: https://www.founderspledge.com/research/all-in
I think in general in doing your own research you should assume that ways you should make 2 assumptions:
1) Systematic changes to the energy grid are almost always going to dwarf individual lifestyle changes.
2) You can probably have more effects on this than you think if you look for ways to help out with greening the grid.
It was funny seeing Ketan Joshi's name pop up again here: he's a deeply disingenuous commentator who mostly just looks for excuses to trash on people or ideas not in his clique. His whole exchange w/ Jordan Shanks some years back really highlighted the kind of person he is.
Can you link to the Jordan Shanks exchange? Don't really know Joshi's deal, he's just been sharing my stuff with extremely sloppy negative commentary that a high school student could poke holes in recently.
I have to admit I am not nearly invested enough in Australian politics to make it through the videos but appreciate the share
Absolutely understand, which is why I kinda held back on linking anything until you asked. It's incredibly niche and scattered which just kind of makes it a mess to dig through unless you're super into it.
Another excellent post, but... you excuse for people who believe factual lies about generative AI and the environment, and use it to attack generative AI companies or users as having been "tricked". I am not sure I agree - one should do their due diligence before going out of their way to criticise others.
There I just meant the people who believe the lies, not the ones spreading them
Thanks for fighting the good fight of trying to improve everyone’s epistemics!
Strongly agree. I can't imagine why people would care so much about one small use of power and a low contributor to climate change versus causes that are magnitudes larger.
Some more facts:
https://www.sustainabilitybynumbers.com/p/carbon-footprint-chatgpt
Andy, the issue is that you've dared to question a piece of Dogma from the Cult of Doom.
https://www.mattball.org/search?q=doom
Seeing everything as Horrible is what gives them their identity. ("I live in The End Times!") They have their Sacred Texts (Doomer blogs, etc.). Don't argue with crazy people - contribute to a constructive conversation with people who actually care about facts and want to make a difference. IMHO.
Two comments about your responses to criticism:
First, about peer review: I have been a researcher for over 15 years, I have been involved in dozens of published papers and have reviewed dozens more. You claim:
"A claim only needs to be peer reviewed if it’s something new that we don’t already know that hasn’t been verified by other studies. In these posts, I’m taking the numbers that we’ve been given by peer-reviewed studies or that are our best guesses, and saying “if these are true, this is how the energy costs of ChatGPT compare to other things we do.”"
This is not true. Review papers - scoping, narrative or systematic - as well as meta-analyses do exactly this: They collate research across a certain research question and provide a picture about the level of knowledge, either quantitatively or qualitatively. All of these go through just as rigorous peer review as the original studies, empirical or otherwise (which I know because I have both been a co-author on as well as peer-reviewed these). This is necessary because even if you base your argument and/or calculations on peer-reviewed literature, you may still be wrong! Your argument may be based on a fallacy or your calculations may be wrong.
The same is true about your post. What you're doing here is basically a review of the "literature". And you may be wrong in any of your conclusions. However, you can't even base most of your calculations on peer-reviewed findings! As you state yourself, much of the data is what OpenAI, Google et al. have deigned to put out, assumptions other people have made etc. This makes the conclusions even more contingent, and the premises, calculations and arguments even more in need of external verification.
Now, academic peer review is not a panacea and there has been a lot of well-founded criticism of the process and I personally appreciate efforts to make it more transparent and accountable. I also think post-publication peer review is a good idea - which is *in principle* similar to comment sections such as these, with the difference that it, too, is transparent and accountable, i.e. named individuals with known qualifications.
In any case, I don't think your defense about why this doesn't need peer review holds up. I appreciate that you have corrected some parts of your original post *where you agreed with the criticism*. In an actual scientific journal, peer criticism, especially on contentious topics, is often published alongside the original paper. Here, you have to scroll down to the comment section, and those comments that got the original author's (your) approval get a special mark above it. All of these set it apart from a scientific publication. That's okay! It's not a scientific publication and it doesn't need to be! But again, the criticism is valid, in my opinion.
The second point is about your Kate, Bob and Freddy example. This example is misleading and unfair, and I think you know this. Anyone who engages with climate change activism knows that issues are always both individual and systemic, and here you are confusing these levels. Anyone can decide for themselves to be a vegan (as you do) or use or not use ChatGPT. We have more or less complete control over these decisions. No-one can decide by themselves whether a nuclear plant keeps running. Not even 500 well-intentioned people have any definitive power to decide this. They can campaign for it, they can call their representatives, communicate with the decision-makers at the energy providers etc., but there is no way of knowing if this will achieve the goal. And even if it does, it is completely disingenuous to then put it as "These 500 people have now individually saved x amount of kWh/CO2e or whatever." (This is in addition to a, in my view, simplistic perspective what it takes to "keep a nuclear plant running". I personally think that it was a mistake to phase out nuclear before coal in my home country of Germany. However almost all nuclear plants would *still* have gone out of service because many would have needed extensive repair and maintenance, which would have been very, very resource-intensive and would have negated any environmental gains for quite a while.)
This, in my opinion, points to a further problem of your framing as a whole: Framing the problem as "making individual ChatGPT out to be a problem distracts from the real issues" is a fallacy. This assumes that we can only care about a limited number of (or a single) issues at once. If Bob is serious about the environment, he is probably not only a vegan but *also* limits his use of individual transport, checks the energy rating of the appliances he uses and, yes, limits how much he uses ChatGPT. In campaigning, activists have never focused on any one issue (food production, clothing, transport, etc.) exclusively.
This framing is in my opinion analogous to the argument that many climate distractors use in my country when they say "Why should we be the ones leading the way on climate action when others don't pull their weight? We only produce 2 % of all CO2e, that's nothing compared to China and the USA!" Sure, ChatGPT queries may not look like much. But most of them are *additional* to what we're doing anyway. Of course we can say that doing an extensive query would only need just as much energy as microwaving a bowl of popcorn. But then we haven't microwaved the popcorn.
You may disagree with my criticism of the general framing, but I hope the criticism of the Kate, Bob and Freddy example is clear: This is apples and oranges and, in my view, weakens your argument instead of strengthening it.
(Edited for formatting and grammar)
"1-2 seconds on an entire 10kW GPU server would use up 2.7-5.4 kWh."
Is "kWh" at the end a typo? Shouldn't it be "Wh"?
The replies you get here sometimes are goofy goober in really unique ways
Sorry I block dumb-dumbs adios