Mystery AI Hype Theater 3000

Episode 43: AI Companies Gamble with Everyone's Planet (feat. Paris Marx)

Emily M. Bender and Alex Hanna Episode 43

Technology journalist Paris Marx joins Alex and Emily for a conversation about the environmental harms of the giant data centers and other water- and energy-hungry infrastructure at the heart of LLMs and other generative tools like ChatGPT -- and why the hand-wavy assurances of CEOs that 'AI will fix global warming' are just magical thinking, ignoring a genuine climate cost and imperiling the clean energy transition in the US.

Paris Marx is a tech journalist and host of the podcast Tech Won’t Save Us. He also recently launched a 4-part series, Data Vampires, (which features Alex) about the promises and pitfalls of data centers like the ones AI boosters rely on.

References:

Eric Schmidt says AI more important than climate goals

Microsoft's sustainability report

Sam Altman's “The Intelligence Age” promises AI will fix the climate crisis

Previously on MAIHT3K: Episode 19: The Murky Climate and Environmental Impact of Large Language Models, November 6 2023

Fresh AI Hell:

Rosetta to linguists: "Embrace AI or risk extinction" of endangered languages

A talking collar that you can use to pretend to talk with your pets

Google offers synthetic podcasts through NotebookLM

An AI 'artist' claims he's losing millions of dolalrs from people stealing his work

University hiring English professor to teach...prompt engineering



You can check out future livestreams at https://twitch.tv/DAIR_Institute.

Subscribe to our newsletter via Buttondown.

Follow us!

Emily

Alex

Music by Toby Menon.
Artwork by Naomi Pleasure-Park.
Production by Christie Taylor.

Alex Hanna:

Welcome everyone to Mystery AI Hype Theater 3000, where we seek catharsis in this age of AI hype. We find the worst of it, and pop it with the sharpest needles we can find.

Emily M. Bender:

Along the way we learn to always read the footnotes, and each time we think we've reached peak AI hype, the summit of Bullshit Mountain, we discover there's worse to come. I'm Emily M. Bender, Professor of Linguistics at the University of Washington.

Alex Hanna:

And I'm Alex Hanna, Director of Research for the Distributed AI Research Institute. This is episode 43, which we're recording on October 21st of 2024. Have you prompted ChatGPT for a made up answer lately? As a listener of the show, we'd hope not. But if you did, that request was sent to a data center, where OpenAI's oodles of servers and NVIDIA's GPUs processed the request and sent it back, uh, sent back your neat little pile of synthetic text within seconds.

Emily M. Bender:

In the process, that machinery used a small amount of electricity. It also heated up just a bit, requiring just a bit of help from a massive water powered cooling system to prevent the equipment from getting too hot. Not a big deal, right? But OpenAI brags that ChatGPT has 200 million users per week, and you know that most of them are doing more than one query in that week. And that's just one of the many large language models running in massive data centers around the world. As companies try to find new ways to scale their synthetic text extruders, there's a commensurately massive cost in energy, water, and CO2 emissions. So much, that in the US, AI is straining the electrical grid and jeopardizing hard fought climate targets to transition to renewable sources.

Alex Hanna:

Don't worry, the tech companies say AI is going to solve the climate crisis. If we just pour enough natural resources into it and accept enough carbon dioxide emissions as a result. But we have the perfect guest to help us break down the cost of all those prompts and the hand wavy justifications of Silicon Valley. Paris Marx is a tech journalist and host of the podcast "Tech Won't Save Us." He also recently launched a four-part series called Data Vampires, which I had the pleasure of being featured on, about the promises and pitfalls of data centers like the ones AI boosters rely on. Welcome, Paris.

Paris Marx:

Thanks so much. It's really great to join you.

Emily M. Bender:

Yeah, we've both been on your show and we're super excited to have you on ours.

Paris Marx:

Absolutely. I I've been a listener as well. So it's really exciting to finally be on to chat with you both.

Emily M. Bender:

Yeah. All right, well, we have some artifacts to, to talk through today. And, um, as Alex and I can say, they are rich texts, rich in the sense of manure. Um, um, and so.(laughter) I'm going to start us with this one. This is some reporting in Mashable by Chase DiBenedetto from October 7th of 2024. The headline is, "Google's former CEO: AI advances more important than climate conservation." And then subhead is a quote, "'We're not going to hit the climate goals anyway, because we're not organized to do it.'" Great. Okay, so I'm going to just read the first couple paragraphs here and then we can react."AI is demanding more and more energy for its immense processing needs, and while many of its leaders are addressing the climate concerns, others are letting artificial intelligence lead the way. Ex-Google CEO Eric Schmidt is among the latter, for one. Appearing at a recent Washington AI summit, Schmidt argued that current climate goals should be abandoned in favor of a no bars held approach to AI investment.'All of that will be swamped by the enormous needs of this new technology,' said Schmitt, referring to recent efforts to make AI more environmentally friendly. We may make mistakes with respect to how it's used, but I can assure you that we're not going to get there through conservation." Thoughts? Any reactions?

Paris Marx:

This is, this is so bad. Like, you know, like, obviously, we've been hearing the stories for a while about the amount of energy use and, you know, the climate concerns that these, um, you know, generative AI and the AI boom has created, right? But more and more, and like Schmidt is very much part of this, we're seeing these CEOs come out and basically be like, Hey, you know, AI will solve climate change so it's okay that we're funneling all of our resources, all this energy, creating all this emissions to power it all. Um, and like, it's almost refreshing in a way for him to just come out and say it, so that we we can clearly say, like, look, these people don't care anymore. They're willing to bet the entire future on AI and this is a terrible bet.

Emily M. Bender:

Absolutely terrible bet. And we're going to see later in the episode, some just drivel from Sam Altman, um, which I think shows very clearly why nobody should believe these people, even if you like wanted to maybe believe that this could be the solution.

Alex Hanna:

Yeah. I mean, we've said it before about the kind of view of how scientific progress is made, sort of just pouring AI on it and it seems like it's going to solve it through some mechanism. And Schmidt really outlines it out, Altman really outlines it like that. Like, okay, these are not, first off, these are not purely technical problems. And then second, you can't just kind of pour a language model on it, and it seems like that's going to resolve the issue, right?

Emily M. Bender:

Yeah. I want to do one more paragraph of this article. But first I want to take issue with the first one. There's a metonymy, which is annoying. So this first sentence here, "AI is demanding more and more energy for its immense processing needs." So that's metonymy in the sense of using the product to refer to the people and companies who are building the product. But because AI gets anthropomorphized all over the place, it sounds like the AIs are making this demand. Um, which, no. Okay. Uh, "Schmidt has his own AI investments, including the defense company White Stork, which is testing a new legion of AI powered military drones.'We're not going to hit the climate goals anyway because we're not organized to do it,' Schmidt continued.'I'd rather bet on AI solving the problem than constraining it and having the problem.'"

Paris Marx:

Uh.

Emily M. Bender:

We, we, we have the problem. Alex Hanna: We have, wait,"I'd rather bet on AI solving than constraining it and having the problem." Oh, okay. So constraining it is referencing to AI and then that's, yeah. This is magical thinking.

Paris Marx:

Oh yeah.

Alex Hanna:

Yeah.

Paris Marx:

It's also so fascinating too because like what this article doesn't mention is that the conference he was speaking at was one that his own organization, like, put together. It was the first, I think it's the AI Plus Energy Summit by this, I can't remember the name of the group off the top of my head, but it's a group that he funds, basically, right? And it's part of doing exactly what this paragraph says, is that, you know, it's part of his push to get AI to be more adopted by the Pentagon, by militaries around the world, but the U. S. military in particular, because Schmidt has very much made himself like one of these key people who is trying to push this adoption of AI for military purposes, as, you know, Silicon Valley is in general, I think, has adopted this warmer stance toward the military. But like, when you listen to his actual comments, he's kind of saying like, we are making things more, you know, more efficient, I guess, in the creation of these data centers, but the amount of AI and the amount of computation that we need for this AI is such a scale beyond the efficiency improvements that we're making that, you know, like he's saying here, we just need to, you know, kind of accept that we're going to totally blow the climate targets and then, you know, rely on AI to fix it for us. Because in, in this, this like uh discussion, he calls AI, not just a universal technology, but an alien intelligence that, that we don't understand. And it's like, oh my God, man, stop it.

Emily M. Bender:

Smitt gets the alien intelligence down here.

Paris Marx:

I don't know if it's in this article in particular, but like in his actual comments, that he makes at the, at the thing. He talks about it as an alien intelligence. Yeah.

Emily M. Bender:

You know, so that's a case where the, the journalist I think was actually being too kind, not quoting that part. Right.

Paris Marx:

Yeah.

Emily M. Bender:

Um, yeah. So, and also, so he's not saying we have to accept that we're going to miss the targets or that AI needs this. He's actually saying you have to accept that, that I slash we, the people building this stuff, um, are just going to need to blow past those targets cause we need to make a lot of money with this.

Alex Hanna:

Yeah.

Emily M. Bender:

Yeah.

Paris Marx:

It's spot on.

Emily M. Bender:

Yeah. Yeah. All right. So should we leave, uh, White Stork and move on to Microsoft?

Alex Hanna:

White Stork is just a wild name for a defense company that I could write an essay on, but yeah.

Emily M. Bender:

I kept wanting to say White Shark when I was reading it too, but okay.

Alex Hanna:

I keep on thinking Stork Industries. Um, just a little MCU extended cut. Yeah.

Emily M. Bender:

And there's also this weird like, uh, stork and, um, you know, pregnancy myth stuff. Right. So this is, um, giving birth to new intelligences. Yeah, we, we could definitely go on.

Alex Hanna:

Yeah. Giving eugenics realness. Uh.

Emily M. Bender:

All right--

Paris Marx:

You hate to laugh at it, but like, yeah, that's all you can do.

Alex Hanna:

Right.

Emily M. Bender:

That's, that's our job here. Ridicule as praxis, right? We laugh, we try to get you to laugh with us.

Paris Marx:

We move on to the good news of Microsoft, right?

Emily M. Bender:

Yeah, I mean, I, I do have to say that, I'm glad they're putting this information out there. It seems like they have collected a lot of data. One of the big complaints we frequently have with Google is that they don't--let alone OpenAI, although, hmm, how much is OpenAI actually included in this? How much is, you know, if Microsoft is running uh cloud servers and other people are doing commute on--compute on those servers, is Microsoft counting that as part of their impact or no, do you know?

Paris Marx:

That's a good question. I'm not entirely sure. And especially where we're looking at, you know, the 2024 environmental sustainability report, like we've seen this stuff continually increase. I would imagine some of that is in there as long as it falls on the Azure servers. Um, but I don't believe they like explicitly talk about it here. And I think your point around like the information that they're gathering is an interesting one because I feel like they also try to call attention to that in the report itself by noting that we're collecting all this information and we're sharing it so people can hold us accountable and blah, blah, blah, which is obviously a real contrast from the stories that we saw around Google trying to keep its water use in particular secret and not wanting to share those things. Um, so yeah, you know, it's just an interesting framing to say like, you know, okay, we're not totally meeting our targets and stuff, but, uh, you know, we're, we're sharing the information so you know it.

Alex Hanna:

Yeah. Actually I'm, I'm curious about this too, because if you scroll to the end and I don't know--

Emily M. Bender:

Of this page?

Alex Hanna:

Of this page, because I actually, this is, um, cause this is, "Pursuant to, um, California assembly bill number uh, 1305, which was signed into law October 7th, 2023. Microsoft may be required to make certain disclosures regarding its involvement in the voluntary carbon offsets market, use of voluntary carbon offsets, and or certain claims or goals regarding greenhouse gas emission reductions." And so, I'm looking at now, I wasn't aware of this, that, that they were--this is voluntary. So, um, if you purchase voluntary carbon offsets, um, then there is, then one needs to actually disclose this under 1305. So a sustainability report should be lauded, but once we get into the report, a lot of it is still very opaque, um, once you kind of dig into the details.

Emily M. Bender:

Sorry, I have to, I have to, um, follow the podcast rule of saying "cat," that purring is coming from Euler who is nearby, but not on screen. Um, and blissfully unaware of the climate crisis, being a cat.

Paris Marx:

Must be a nice place to be, you know? Yeah.

Emily M. Bender:

Um, so yeah, so this is interesting. They're basically, as I understand it saying, um, we might have specific things we have to say as of January 1st, 2025. Um, so we're going to figure out what those specific things are.

Alex Hanna:

Yeah.

Paris Marx:

Yeah. I think it's really good to point that out too. Right. Because so often these companies, um, you know, will, will follow a particular law, but claim that they're doing this for like totally altruistic reasons. And it's like, oh no, actually you've just been forced to do it. And that's why we're finding out more now. Um, so yeah, I think that's a really important thing to note.

Emily M. Bender:

So this is a wwebpage. This is kind of typical for Microsoft. They tend to be crowded on the web pages, compared to Google and I think Meta, who go for very spare design. And in this case to me as a reader it feels a little bit like I'm glad there's all this information but also I feel like it's overwhelming and it seems like there could have been a more like at-a-glance way to present it. Um, and that's not what they've gone for. They've gone for we're going to really impress you with all the details so that you walk away thinking we've covered all the bases.

Alex Hanna:

Yeah, definitely like lots of four column things. Yeah. Um, so do we want to go into the, I mean, the PDF is where the rubber meets the road and the PDF is even more overwhelming, but I mean, the PDF, there's, it's an 88 page PDF, but the, you know, like, and it's got this, you know, just to give people, it's got this like, um, Financial Times, like salmon background, and there's pictures of mushrooms and it's very serene and it makes you feel like you're going you know, go into the earth and, and feel the soil under your toes. Um, there's lots of pretty pictures. Um, and then there's the forward to it and they talk basically, um, I think there's, there's some pretty important things to note here, which, you know, there was a lot of reporting around this, especially around how they miss their metrics. Um, so in the fourth column, in the third column, it starts and it says, you know, there's four-- "In four areas we are on track, and each of these we see progress that has a potential to global impact beyond our sustainability work.

These are:

Reducing our direct operational emissions, scope one and two, um, accelerating carbon removal, designing for circularity to minimize waste and reusing cloud hardware, improving biodiversity and protecting more land than we use." So very, and then that's the, that's the nice fluffy stuff. And then the stark stuff is, "At the same time, there are two areas where we're not yet on track, and each of these were intensively engaged in work to identify and pursue additional additional breakthroughs. Reducing our scope three or indirect emissions, reducing our water use and replenishing more water than we consume in our data center operations." So then that's, that's, you know, really where the rubber meets road. Um, it's, it's this data center problem fueled by the AI gold rush. And then--

Emily M. Bender:

I like how they say"pursue additional breakthroughs" as if they're gonna find some magic bullet that solves it all instead of just like, you know stopped wasting all this compute on synthetic media.

Alex Hanna:

Yeah. And I do want to point out just in the next page on on page on on page 5, they say "Carbon reduction continues to be an area of focus, especially as we work to address scope 3 emissions. In 2023 we saw scope 1 and 2 missions decrease by 6.3 percent from our 2020 baseline. And that remains on track to meet our goals, but our indirect emissions, scope 3, increased by 30.9 percent, and across all 3 scopes Microsoft's emissions are up 29. 1 percent for the 2020 baseline." So the scope 3 emissions really--which I think is, um, particularly their, their, um, embodied missions, particularly in data center build out. Um, yeah, so, thoughts on that, Paris?

Paris Marx:

Yeah, bad news, right? This is why we want to exclude the scope 3 and not talk about that so much, because it makes the picture look a lot better. Um, you know, I think just to add a few things to what you were saying, I would say, bigger picture, obviously we've had the reporting on how Microsoft's emissions are, are up about 30 percent over five years, I think. I think that's the number for them and not for Google. One is 48 and one is 30 percent. I think with Microsoft it's 30 percent.

Alex Hanna:

Yeah.

Paris Marx:

Um. And, and, you know, that's for data centers and stuff like that. But we also had this reporting out of the Guardian that, that dug into, um, the sustainability reports and the emissions numbers of, you know, many of the major tech companies, and found that when you look at the data centers of, I believe it's a Google, Microsoft, Meta, and Apple, that the emissions that they reported are quite a bit less than the actual emissions that they they are, they are actually emitting. Um, and they said that their actual emissions are about 662 percent higher than what is reported in, um, their, you know, their documents because, you know, they rely a lot on buying renewable certificates and offsets and things like that so that they can claim that their actual emissions are lower than they really are. Um, and so, you know, when you look at these sorts of numbers, you need to ask yourself, like, what is actually going on here? To what degree are the emissions actually down? Because I feel like, you know, when we were referring to those columns on the previous page in, on page four, um, you know, you have the talk of accelerating carbon removal, and that refers specifically, I'm quite certain to a project that Bill Gates has been pushing in Iceland that is about kind of sucking the carbon out of the air. That is still a very small scale thing at this point, but that Microsoft is investing in because Google or sorry, because Gates is involved with it. Um, and that is what is what is what that's all about, right? Um, and we also know that when it comes to these renewable certificates, um, they can buy these, uh, in any part of the world, they can buy it for renewables at any time of the day, it doesn't need to be at the same time that they're using the energy, or like even in the same area that this, this energy is, is being used, um, and so it's quite deceptive. And the Financial Times reported, I believe it was last month, that that, um, a number of these major tech companies, Amazon and Meta in particular, are lobbying to change the rules around carbon accounting to make it so that it looks like they're emitting, again, like a lot less than they actually are, um, because they're buying these certificates that are not actually, you know, that don't actually mean that their emissions are being reduced. And I think just the final point I would add on that is when you think about water, it's very similar. And, you know, you even see it in, um, part of what they have here. I'm not sure, um, actually it's on this page under, under water, water positivity, um, where they talk about how they've contracted water replenishment projects estimated to provide more than 25 million, um, uh, cubic meters, uh, benefit over the lifetime of the projects, "enough water to fill about 10,000 Olympic-sized swimming pools." And again, like, you know, they're still using all the water. They're just somewhere else. They're, they're paying for these projects to bring water back. And we don't even know if it's in the same communities or what, right? Um, so it's all like, it's all very, I would say, very deceptive in trying to make the numbers look bad or sorry, trying to make the numbers look better when the numbers are actually very bad, you know, to begin with. Right.

Alex Hanna:

Yeah. No, absolutely. Go ahead, Emily.

Emily M. Bender:

I was going to say, you know, listening to your series on data vampires and the bit about these, um, data centers come in and basically just suck a community dry of their water. Um, and, you know, it's nice if Microsoft is helping some other community somewhere else with water, but that doesn't help the people who are living around the data center.

Paris Marx:

Exactly.

Alex Hanna:

Yeah. And just--

Paris Marx:

And even in some of these communities, even in some of these communities, like, you know, they will invest in water infrastructure in the communities as well. That might be part of like, what the trade off is, or the deal of building a data center or building more data centers on top of what they already have there. But even then, like the communities are still quite worried in a lot of the places where the they're being built, because, you know, they might be regularly subject to drought conditions, or, you know, they might have had issues with water access in the past. And just because you know, water infrastructure is improving or whatever doesn't mean that they're finding all this new water to, you know, put into to data centers. So it's a huge concern.

Emily M. Bender:

But but but they've got AI enabled acoustic analysis to reduce water loss.

Alex Hanna:

It's truly ridiculous. Yeah,

Emily M. Bender:

So this is in their water positive thing. It says, um, "First-of-their-kind replenishment projects like FIDO, which leverages AI enabled acoustic analysis to reduce water loss from leakage." And it's like, okay, you could use machine learning over acoustic data around pipes to detect leaks. Like that sounds like a plausible thing. Um, but what are you doing when you've found it? Right. And how much, you know, what's the data center cost of collecting all that data and running the acoustic analysis? Like, is it actually a net win?

Alex Hanna:

Right. I imagine, I mean, just, I mean, how much are you actually losing through that too, compared to just how much you need in the first place? I mean, one thing I wanted to point out from the Data Vampire series -- which is excellent, again, everybody listen to it, I mean -- was the interview with that journalist from the, The Oregonian who, you know, where they went to suit, you know, they, they, they in a suit against Microsoft in the city to find out how much water they were actually using. And it turned out there was 29 percent of the water usage in The Dalles in Oregon was going to this data, to the data center. And it was, that was Google, not, not Microsoft, but it's just really, really stark, really wild numbers.

Paris Marx:

Totally. And like, you know, I feel like that is a story that really caught people's attention because it occurred at a time when you know, I think data centers were like slowly getting onto people's consciousness. You know, this is like around 2021 when the suit happened and then 2022 when the Google finally released the numbers after trying to sue to keep its water use secret for for a year. And it felt like, you know, this was not only--generative AI, like, you know, started at the end of 2022, right? It was when that boom really happened, but people had already been aware of the amount of, um, increasing energy use and stuff from, um, you know, centralized computation, if we want to put it that way, from, like, crypto miners and things like that, right? So this was starting to, to kind of get on people's radar. And I feel like that, in particular, was a story that really kind of broke through to a wider public in finding out what was happening there, you know, because Google was so aggressive in trying to hide its water use, which, which meant that there was some reason that they were trying to do that. Right? Um, and I feel like more and more, like, it feels like every month there's a story about a new community in the United States, let alone what's happening around the world, that is dealing with, the consequences of building these massive hyperscale data centers in particular, um, you know, and, and, you know, these companies basically just moving in expecting to use all this energy and all this water and, you know, communities being like, wait, what does this mean for us? And it's very like, you know, the broader impact seem to be very secondary to like the short term potential economic, you know, improvements or incentives that can come with building one of these things in the community and a lot of kind of left behind areas that are targeted by these major companies. Um, and of course, then we see, like on the bigger picture, what these companies are actually doing and how how willing they are to sacrifice climate targets and and so much else just to pursue this vision that they have. And just to add one more thing about Microsoft in particular, like in 2020, they talked about taking a carbon moonshot, right, they were going to be carbon negative by 2030, and you see that repeated in this report. But then Brad Smith, the president, you, who again is, you know, quoted or, you know, said to be one of the people who wrote this forward. I'm sure it was some PR people and stuff who, who actually put it together. Um, but you know, he gave this, this interview a few months ago where he was basically saying that, you know, they declared this carbon moonshot in 2020, but now the moon is five times further away, four years on.\ And it's like, okay, you're admitting But like, you know, what does that mean for, for all the rest of us, you know?

Emily M. Bender:

Yeah. And that's such a, I mean, it was a ridiculous thing for him to say, but also so passive, like, oops, the moon, the moon went further away. It's like, no, actually you made some decisions.

Alex Hanna:

Yeah. That's such a, that's such a hilarious thing to say. The moon got further away. What? That's not what a moon--you need to maybe change the metaphor, if you're going to keep on moving the goalposts like that.

Emily M. Bender:

We would also have some--

Paris Marx:

Maybe a comet that flies around or something, right?

Emily M. Bender:

Yeah. We would have some massive problems if the moon was suddenly five times further away. That would screw up a lot of stuff. So before we get back to this, uh, beast of a report, I want to share a comment from Abstract Tesseract who says, "There's something almost poetic about the juxtaposition of obfuscating legal wallow text with like cottagecore mushroom and moss images," which is pretty great.

Paris Marx:

Definitely makes the report a bit more appealing, you know, like--

Emily M. Bender:

Yeah. And here, here we have some friendly looking workers inspecting, uh, wind turbines, and here we've got somebody taking some measurement of, you know, some kind of plants and what looks like marshland.

Alex Hanna:

It looks like a rice paddy and it's, and I was struck by this because this person has like a tablet in one hand and is, you know, you know, and it's, it's like a very, you know, it's, it's, it's trying to, um, pose this kind of like, technology and nature can live in harmony or something like that.

Emily M. Bender:

But, but also the, the thing that he's looking at that you said is a rice paddy looks like it's maybe one foot by two feet and around him he's in some kind of marsh, so did they like artificially put some rice plants right there for him to be inspecting in the picture?

Paris Marx:

My, my guess is he's probably like planning to plant something there. Like, like it brings to mind those like reforesting initiatives, right? Where they talk about how, oh, they're going to plant all these trees and that's going to make this big difference. And of course we just had this recent report that says, hey, the trees aren't sucking down carbon like we expected them to, this is a big problem.

Emily M. Bender:

No. And what is it? Finland just, um, like they had a big carbon sink, sink in terms of like the, the peat and it's no longer absorbing carbon.

Alex Hanna:

Yeah. I saw that reporting in the Guardian, which, yeah, so basically, and so sort of the, the subhead of that was like, well, that basically means, you know, any kind of, um, any kind of carbon capture initiative that has banked on, you know, carbon sink, you know, those are way off now. And, you know, the problem is getting exponentially worse.

Paris Marx:

So great. Love it.

Alex Hanna:

Yeah. There's a few more things on this report before we go to the Altman like stream of consciousness blog post shape thing. Uh, first is on page seven, there's the 2023 highlights and with like kind of a crayon green--it's very, uh, it's just. Anyways, it just, there's something, there's a dissertation, I mean, there's a dissertation in many of this, um, but there's also, this is just the imagery that they have here. So they have four blocks, and it's the carbon, which is the renewable energy used in fiscal year 2023. Um, and it's, it says, 23.6 million, um, megawatt, megawatt hours, which"would be enough uh, to power Paris with renewable energy for about two years." And it's got an Eiffel tower.

Paris Marx:

Awesome, I'm doing well.

Alex Hanna:

We're doing great. We can do two Eiffel towers. Um, yeah, just, it's, it's all for you, Paris. And then, um, and then there's water. Total contract on water replenishment since 2026. And then there's, I don't even want to say the, the, the kind of a particular number, but it is "enough water to fill 24,000 Olympic sized swimming pools." Um, then there's this thing here that, that I'm curious if you know about, um, Paris, where it says, "Our new data centers are designed and optimized to support AI workloads and will consume zero water for cooling." That's is that like a new thing or is it, or are they sort of saying no net water? Cause I'm kind of curious. I'm like, well, what is cooling them then?

Paris Marx:

Yeah. So, so there's two potential meanings there, right. Um, and I would say, I'm not sure specifically what they're referring to, but there are some data centers where they're relying more on say, Um, you know, air conditioning and whatnot, but then you use a lot more power, um, instead of using water, right? And, and I know that in some data centers, they choose to rely more on higher energy use rather than higher water use. Um, you know, depending on the region where they're building it and whatnot, right? Um, but I would say that likely refers to net water. Um, and so again, like going back to the thing that they said higher up, where they said that they were, um, you know, investing in these water replenishment projects and all these sorts of things to try to provide more water back. There's also discussions about, um, data centers not evaporating as much water and like trying to filter it to put it back in, you know, the, the places where they took it from or whatnot. But the issue with a lot of these data centers is they particularly usually want like drinking water, right? Like clean water, um, because then it's not going to cause any problems as it goes through the data center itself. Um, and a lot of this water after it's gone through the data center cannot be put back into like the drinking water supply, is my understanding. Um, the other thing is there are some parts of the world where they're talking about, um, you know, recycling the water, using it for other purposes, um, contributing, say like the heat to district heating, um, you know, uh, systems and whatnot. But in all those cases, based on the people who I've spoken to, like, those are like a few data centers here and there that are kind of like demonstration types of things. But a lot of this stuff is not rolling out across like the, the vast network of data centers that they're actually using. So it feels kind of, you know, like a kind of deceptive PR thing for me.

Emily M. Bender:

It says "our new data centers," but that doesn't. Yeah.

Alex Hanna:

Yeah. Yeah. Yeah.

Emily M. Bender:

And yeah. It's also interesting to me it says "they are designed and optimized to support AI workloads," whatever that means, "and will consume zero water for cooling." So they're not designed to consume zero water. So that suggests that there's there's room in there for it to be offsets, effectively.

Paris Marx:

Yeah, that would be my thinking or, you know, some some like net accounting, right?

Alex Hanna:

Yeah, yeah, or some kind of a mix of like an offset and like a gray water situation, anyways, uh, just kind of speculation. Uh, the next two is like wastes, total amount of waste diverted from laying landfills, incinerators, 18,500 metric tones, which is the "equivalent to the weight of over 45 commercial passenger jets," great comparison.

Emily M. Bender:

And picking on jets as a positive point of comparison is a choice.

Alex Hanna:

They could have, they could have chosen whales.

Emily M. Bender:

Yeah, or polar bears.

Alex Hanna:

Yeah, completely. Then the, uh, actual land protected, uh, which is curious. So about 16,000 acres, equivalent to 9,000 soccer pitches. Um, and then they promote this, this, this this Microsoft data center community environmental sustainability program."We're responding to local needs in communities that host Microsoft data centers and where our employees live and work." Um, yeah, so it's, I mean, it's, which is again, very much washing, you know, washing the kind of harm they've done to communities.

Paris Marx:

It brings to mind two things in particular, right? Like, on the one hand, how, you know, like we were talking about with, like, the peat in Finland, for example, how a lot of these companies buy, like, you know, these offset projects that are essentially like forests or land that can, um, you know, theoretically absorb all this carbon. Um, and it feels like that's probably what they're referring to here, but it also brings to mind like the stories that I've heard from northern Virginia, which is the biggest kind of agglomeration of hyperscale data centers in the world. Um, and how like, you know, obviously they're just like gobbling up farmland and any kind of land they can get their hands on to convert into hyperscale data centers to the degree that like a lot of people who live in those counties, um, you know, are finding it not only impacts their, their quality of life, but in some cases they're even being like bought out by data center projects or, or sometimes like the data centers will move in, their quality of life will decline and the data centers will be like, well, we can buy out your land and build more data centers on them. So you don't need to live next to this anymore. And in a number of cases, like they, they do that. Right. Um, so yeah, like especially that that second part where they're like, we're responding to the local needs. It's like, oh yeah, interesting. Okay.

Emily M. Bender:

Yeah.

Alex Hanna:

Yeah.

Emily M. Bender:

Oof.

Alex Hanna:

Oh yeah. This is really, really gross stuff. Um.

Paris Marx:

Yeah.

Emily M. Bender:

So you wanted to take us down to this, uh, the scope three stuff, right?

Alex Hanna:

The scope three stuff, I just like this visualization and just to describe, yeah, so this is page 12 of the report, which is a very funny, uh. It's not funny because it's sad, but it's, it's the kind of ways, just in terms of the emissions. So the emissions of scope 3, which account for 96 percent of total emissions, and they have this graph where the scope 1 and scope 2 emissions are these minuscule amounts. And then they say that the scope 3 categories include purchased goods and services. So I'm, I'm suggesting, I think that's probably more of kind of energy, um, and also purchased hardware and also capital goods. So in that case, it's kind of fixed assets. Um, and those are the vast majority of what it, you know, what it accounts for in terms of, um, actual, actual emissions. So, I mean, it is very much a data center thing. And then there's this, also, there's this huge gap here in the next page, which is their carbon accounting table. So their base year of scope three was, um, so for fiscal years, 2021 and 22, they have this graph that basically says--well, now I'm kind of confused because they use the same color for scope 3 on the left and scope 1 on the right. So, so incredible, incredible graph making for one, um, just making it as confusing as possible. Now I'm going to hopefully not flub this, doing this live on stream, but effectively what they're, they're showing this kind of, um, and then they've got this--okay. So what they have here is fiscal year 21, is this growth in emissions. So this is metric tons of of GHG emissions and CO2. Um, and they're showing this this this growth from their baseline, which is 2020 to 2023. And then they've got this hypothetical, their target emissions, over on the far left, which is their target carbon emissions and their target removal. Um, and even though there's kind of like almost a completely secular or rather monotonic increase from 20 to 23, they're like, "and somehow" fiscal year--

Emily M. Bender: Step three:

profit.

Alex Hanna:

Yeah. Question, question mark, question mark. We're going to not only, uh, quarter our target emissions. We're also going to remove, we're also going to maybe like multiply 10 times our removal? Uh, and it's really, um, yeah, I mean, a 4X, you know, 4Xing the moonshot indeed, you know. Uh, so it's really, really quite the claim.

Paris Marx:

The alien intelligence will be online by then, so it will help us figure all this out, you know, how to, how to make the magical process of getting from, from here to there. Um, it's fascinating to see how like, you know, when you're talking about, about 96 percent of the emissions being scope 3 and then early on, like, they talk about how, look, we've reduced our scope 1 and 2 emissions. It's like, okay, yeah, but that's, that's pointless, right? And even how they, they're like, um, relying on these carbon removal projects for so much of this like goal and you see like apparently in 2020 they removed a lot more carbon than they did in the following year. So like it's actually decreased, the amount that they have removed.

Alex Hanna:

That's right.

Paris Marx:

Yeah.

Emily M. Bender:

If these numbers are to be believed at all anyway.

Paris Marx:

Exactly. And like one of the things that we know like, you know I don't know if you pay as much attention to this in the United States but obviously Canada, you know oil is a major part of our economy. Um, and and we've been talking about like, uh carbon capture and storage and all that kind of stuff for years as has Australia and and many other countries, and there's always these big promises about like you the massive, like, kind of carbon capture that is going to be delivered by these projects once we funnel millions and billions of dollars into them, and what we find time and again is that, like, they're not actually able to deliver nearly as much as what we thought, because they don't work as well, and the technology, yeah, I think is very questionable, actually, um, and, and in many cases, like, it's just used to justify further you know oil extraction or coal extraction or whatever it is that the various country is doing. And like, I feel very much the same about this carbon removal. It feels like a justification to continue, um, with like a lot of what these companies are doing rather than really taking the difficult action that would be necessary to rapidly decrease emissions in the way that we know is necessary. But that, of course, Eric Schmidt says we're not going to do, so let's just rely on the AI, right?

Emily M. Bender:

Yeah, which actually sounds like a good segue to Altman, but I just wanted to put it here that it feels like what we really need is some sort of a third party auditing system, like obviously some of this information only Microsoft has.

Paris Marx:

Yeah.

Emily M. Bender:

But it feels like the, the second layer of this, there's sort of like analysis of what it adds up to should be done by a third party and not by them.

Paris Marx:

Yeah. Yeah. And even then, you know, like I mentioned earlier, there, there are these groups that are supposed to be responsible for carbon accounting and that kind of set the rules that these companies are supposed to go after. Um, but but what the Financial Times has reported is that Amazon and Meta in particular are very active in trying to lobby these groups to change the carbon accounting rules so that they can get away with using more of these offsets that are like not connected by time, can be anywhere in the world, you know, these, these renewable credits to say that they're emitting a lot less than they are when actually they're just like buying credit for renewables somewhere else in the world to say that the actual emissions they're producing are being like offset somewhere else. So, so even then it's like they have the power to try to fudge the numbers further, right?

Alex Hanna:

Yeah.

Emily M. Bender:

Yeah.

Alex Hanna:

Abstract Tesseract says, "Such a beautiful example of how quote 'quantitative' doesn't equal quote 'rigorous.'" Yeah, in this case, quantitative is, is is intentionally, uh, obfuscatory.

Emily M. Bender:

This feels like a lot of mathiness.

Alex Hanna:

Yeah.

Paris Marx:

Yeah.

Emily M. Bender:

Yeah. All right. So shall we then jump out of the frying pan into the fire here?

Paris Marx:

Let's, let's transcend to the intelligence age.

Emily M. Bender:

Okay. So here is a, I hesitate calling it an essay. I guess we have to say it's a blog post, um, by Sam Altman where he oddly, the URL for this is IA.SamAltman.Com. So he created a separate little subdomain for this piece, um, titled "The Intelligence Age." And the date is September 23rd of this year. And there is a piece of artwork that undoubtedly was, um, plagiarized via Midjourney or whatever OpenAI's product is.

Paris Marx:

Yeah, Dall-E, I think.

Emily M. Bender:

Dall-E, yeah. Okay.

Paris Marx:

I find it interesting as well that, you know, we're all talking about AI, and of course now we're going to the intelligence age, which is IA in his subdomain. Yeah, which doesn't seem, doesn't seem like a coincidence, you know.

Emily M. Bender:

No, he must be very proud of that.

Paris Marx:

Yeah.

Emily M. Bender:

All right. I will, I will read the first little bit. um, and then at the point that I get too appalled, we will stop and talk and then we'll keep going."In the next couple of decades, we'll be able to do things that would have seemed like magic to our grandparents. This phenomenon is not new, but it will be newly accelerated." I'm hanging in there."People have become dramatically more capable over time. We can already accomplish things now that our predecessors would have believed to be impossible." Um, okay. We could have, okay, keep going. Um, "We are more capable, not because of genetic change, but because we benefit from the infrastructure of society of being way smarter and more capable than any one of us. In an important sense society itself is a form of advanced intelligence." Okay. Uncle. That's as far as I could get before we have to talk about this. Um, so the thing that like was already grating on me was talking about people as capable, partially because I don't like how they talk about large language models as capable. Um, but also this has a very like the eugenics is not far from the surface here, even in being sort of denied, we're getting this these benefits not because of genetics. It's like, well, why was that even on the table as a thing that might have been. You know, um. But, okay, I'm going to finish this paragraph and then let you two jump in."Our paragraphs and the generations that came before--" Our paragraphs."Our grandparents and the generations that came before them built and achieved great things. They contributed to the scaffolding of human progress that we all benefit from. AI will give people tools to solve hard problems and help us add new struts to that scaffolding that we couldn't have figured out on our own. The story of progress will continue and our children will be able to do things we can't. Thoughts?

Alex Hanna:

There's already. Two mentions of grandparents and one mention of children. And it is, I mean, the, the lineage thing definitely ekes out of the, of the eugenics sort of sense, you know, even though like it does have that sort of like pathos of like, the familial aspect. I mean, there's, yeah, it's also just like incredibly poorly written. Like, like he really needed an editor on this, but, um, you know, the man isn't known for, you know, he probably let, you know, some of his friends read it--

Emily M. Bender:

And tell him how good it was.

Alex Hanna:

I guess, I guess, I guess Mir, Mir, Mira was out. So she couldn't, and, um, she said some jobs haven't, you know, shouldn't have existed in the first place anyways. So--

Emily M. Bender:

He, maybe he had ChatGPT edit it.

Alex Hanna:

Yeah.

Paris Marx:

Yeah. Yeah, I find it interesting as well. Like, obviously he mentions magic right at the beginning and I feel like we were talking about that with the Microsoft report too, like, you know, finding the magic to make the numbers work there. Um, and this is supposed to be presented as like this, this magical thing, but, but that is also, you know, improving society basically. But it's also interesting when he talks about like the story of progress and, and, you know, history and how we built on things that have come before. Like it very much tries to set this like narrative or, or this notion that, of course, the tech industry has been trying to sell us for a little while, that you know, we can look back and see how, you know, we have, like, developed as a society over the course of thousands of years, or whatever, and, and how we've taken advantage of various technologies to have what we consider to be progress. And how they're very much trying to set up, like, the next stage of this like human story does involve AI and can only involve AI and this is like naturally our destiny to to continue in this direction. And and like I think many people would call it into question like the three of us here.

Emily M. Bender:

Yeah. Let's keep going. So actually this is such a rich text. I want to make sure to hit all the words of it. So, uh, he continues, "It won't happen all at once, but we'll soon be able to work with AI that helps us accomplish much more than we ever could without AI. Eventually we can each have a personal AI team full of virtual experts in different areas, working together to create almost anything we can imagine. Our children will have virtual tutors who can provide personalized instruction in any subject, in any language, and at whatever pace they need. We can imagine similar ideas for better health care, the ability to create any kind of software someone can imagine, and much more." And the thing that was so most infuriating to me about this is even though he starts with we are where we are today because society is, you know, something that allows us to become more, this is so isolating and atomizing. Right, everyone is surrounded by their AI experts and there's no community anymore.

Paris Marx:

Yeah. The community is between the AI agents now. Not the people.

Alex Hanna:

I mean, it's fully kind of, I mean, what we would call in science and technology studies, like technological determinists, you know, it's like we have the technology, the technology that's going to bring us into the intelligence age from the bronze age, you know, it's got that evo bio flavor.

Emily M. Bender:

Yeah. Yeah. All right. You want to read the next bit, Alex? I think you'll like the next bit.

Alex Hanna:

I'm trying to think about how we, I mean, we got to get to Hell.

Emily M. Bender:

Yeah. Maybe we should just do highlights. I know that you've got some favorites in here.

Alex Hanna:

There's a few. Let me, let me find. Okay, so, um, this is funny just because it has a riff on. I'm going to read two paragraphs. I just find it to be funny."Within these new abilities, we can have shared prosperity to a degree that seems unimaginable today. In the future, everyone's lives will be better than anyone's life is now." So yeah, great."Prosperity alone doesn't necessarily make people happy. There are plenty of miserable rich people, but it would meaningfully improve the lives of people around the world." Um.

Emily M. Bender:

We feel your pain, Sam.

Alex Hanna:

This, this next paragraph, I gotta, I gotta read it.

Emily M. Bender:

Yeah.

Alex Hanna:

"Here's one narrow way to look at human history. After thousands of years of compounding scientific discovery and technological progress, we have figured out how to melt sand, add some impurities, arrange it with astonishing, astonishing precision, at extraordinary tiny scale, into computer chips, run energy through it, and end up with systems capable of creating increasingly capable artificial intelligence." Um, just a completely materialist view of the world. We somehow arranged the right things in the right order. And we did the thing. To quote a great meme, yes. And if you added wheels to my grandmother, she would have been a bike.

Emily M. Bender:

Yeah. So, and then he said, he says,"How did, how did we get there? In three words, deep learning worked." And it's like all of this like patently ridiculous magical thinking. He basically says, "Humanity's discovered an algorithm that could really truly learn any distribution of data, um, or really the underlying rules that produce any distribution of data." No, false. That's not true, but he would like it to be. Um, and we have to get to the thing about, um, physics.

Alex Hanna:

Oh yeah. Oh, let me read this. It's terrible. Where is it?

Emily M. Bender:

It's here.

Alex Hanna:

Oh, here. Yeah."Although it will happen incrementally, astounding triumphs, fixing the climate--" As if that's something that can be fixed as a check mark."Establishing a space--"

Paris Marx:

As we know, Eric Schmidt thinks so. So.

Alex Hanna:

"Establishing a space colony--" What."--and discovery of all of physics--" As if physics is, is like a, like a progress bar."--will eventually become commonplace, uh, with nearly limitless intelligence--" As if that is a quantity, quantity."--and, abundant energy, the ability to generate ideas, generate great ideas and the ability to make them happen, we can do quite a lot." So it's really just like this, uh, first off, I blame the,"The Three-Body Problem." I feel like it was a very quantitative view of like--I don't blame solely that. That's like a good thing. I hate those books, by the way, if you like them, I'm sorry. They, they call like game theory, like cosmic sociology. And I took that personally. But like also, um, but like also the idea that like these things are to be solved or like check marks is like, no, that's not what science is.

Emily M. Bender:

Yeah, like there's, there's some like, I like your progress bar metaphor. We are, we are 75 percent of the way done with physics. It's mostly fully discovered. We just got that 25 percent, like that's not science. No. Um, and also abundant energy. Where does that come from? Oh, I guess the AI is figuring it out for us. Like the. Yeah.

Paris Marx:

Yeah. Well, it brings to mind like Altman's comments in, in January, right, where Bloomberg asked him about like the, uh, you know, the climate impacts of like his AI ambitions basically. And he was like, you know what, it's going to require a lot more energy than we thought we needed. And that's either going to require a technological breakthrough in nuclear energy in particular, or we're just going to have to geoengineer the planet for a while until, you know, we like figure out the technological solution. And it's like, yeah, again, like this idea of like total sacrifice for AI and, and for these tech people, like they, they seem to think it like won't affect them at all and don't care about the effects on anybody else.

Emily M. Bender:

Yeah. And I just, like, I want the world to know how deeply unserious these people are so that we can, you know, turn away from them and get down to the serious business of the political work of addressing the climate crisis. Like it's, you know, yeah. All right. So Alex, are we doing a musical transition this time?

Alex Hanna:

Sure.

Emily M. Bender:

Okay. What genre?

Alex Hanna:

Oh, gosh. Uh, surprise me. I don't have anything.

Emily M. Bender:

Um, okay. Musical theater. Cause I like making you do musical theater. Um, this is a musical about, um, the demons in AI hell when, uh, the tech bros up on earth discover that they have a source of abundant energy that they can mine out of the hellfire. And so this is the, this is the opening act, I guess, about how that's going to happen, how it's going to disrupt everything.

Alex Hanna:

Oh, okay. Interesting. I guess that would have to be like, it's an, if this is the opening act of musical theater, it's like, introducing, you know, the great, like, you know, the main character goes, what a wonderful day. I can't wait to go to the sulfur mines. Going to the sulfur mines. Gonna have myself a croissant. I'm going to eat it and then torture anybody that-- Nothing rhymes with croissant, I really worked myself into a corner there. That's all I have. Abstract Tesseract said we should have opened with a data vamp about a data vampire, which is so good.

Emily M. Bender:

Yeah. All right. Abstract Tesseract, I got to look to the chat and see if, you know, if next, next episode, if you've got an idea for the prompt, um, drop it in early so that I can hopefully use it., all right. Can you see my Fresh AI Hell screen now?

Alex Hanna:

Yeah.

Emily M. Bender:

All right. So you took the cosmic sociology thing personally. I'm taking this one personally, Alex. This is from Rosetta Voice and it says, "Hello, dear linguists, embrace AI or risk extinction. This is a message that endangered languages should heed. At Rosetta Voice we are excited to introduce a new resource, a standard reference text designed to support--" Basically they are trying to get work out of people, probably free work, um, to translate this into many, many languages so that they, Rosetta Voice have access to all that data and they think they are um, doing language preservation through that. Which is just, linguists are really angry. Um, and just that embrace AI or risk extinction. No, I refuse.

Paris Marx:

Yeah.

Emily M. Bender:

I refuse.

Paris Marx:

Rightfully so. Yeah.

Emily M. Bender:

All right. Um, so next one, Alex, you want this one?

Alex Hanna:

So this is from HigherEdJobs.Com and it's a posting from Kennesaw State University."Assistant professor of English to start--" Um, I'm assuming next academic year and of the responsibilities, um, so the, uh, so you, uh, so Emily has highlighted, there's a "specialization in professional writing specifically with a focus on AI writing applications." Ugh, um. Oh, is this tenure track? That's a nine month contracted position. So I don't know if this

Emily M. Bender:

is-- That's tenure track. That just means that they're not paid over the summer.

Alex Hanna:

Yeah. Which is a huge normalized scam in American higher education, but anyways.

Emily M. Bender:

Oh, here it is. This is, this was the thing.

Alex Hanna:

Yeah. So teaching one of the responsibilities, um, is "teaching undergrad or graduate courses and or graduate courses in professional writing, including prompt engineering, writing with AI, digital tools for writing, digital applications for creative writers and first year composition." Which is kind of funny that first year composition is kind of tossed in there. You're going to start at like a first year composition that everybody has to take. And prompt engineering, you know, which is really, um, terrible stuff.

Emily M. Bender:

Yeah. And the fact that this is an English department that has jumped on the bandwagon this way is like, no, resist, refuse. You don't have to do this.

Paris Marx:

It's so disappointing sometimes how like, and I'm sure you'll know all about this Emily, like how higher education can like jump on these bandwagons like universities and stuff so quickly to try to get funding or like whatever it is and it's always so disappointing. It's like, please like, you know, have a bit more self respect and like critical approach to all this.

Emily M. Bender:

Exactly. My university has this AI task force, which I was not invited to be on.

Paris Marx:

Surprise, surprise.

Emily M. Bender:

That just released a report, and they're doing this survey now of the community, and this is like, I don't know, 60 some odd questions about how you feel about this, that, and the other thing, and they never define AI.

Paris Marx:

Oh my god.

Alex Hanna:

That's incredible. Yeah. Yeah.

Emily M. Bender:

Okay, uh, speaking of higher ed, this one comes from the Chronicle of Higher Education, with the tag "Synthetic Synthesis." I approve of that tag, but I actually didn't like this reporting very much. So headline is, "Those voices on the podcast, listen closely," by Declan Bradley from October 4th of 2024. Um, and this I'd not logged in on this machine, so I'm just gonna have to tell you what I read. But basically, um, the, there's something from Google's NotebookLM. They've got this feature where you feed it in some data and it produces a fake podcast discussing what you fed it in. And so this is talking about using it on, um, the readings for a course. Um, and then down at the bottom they get to like the critics, which are always sort of shoved like 80 percent of the way through the article. And the concerns that people bring up is that this is going to give students an excuse to not engage deeply with the reading. And like nobody says, and it's going to be wrong a lot of the time. Like, this is, um, yeah. And, and the, I've had to listen to a couple of these and even though I generally try to avoid synthetic media. And it's very, like, jovial and it just, like, kind of gross and they are taking this, the structure of a podcast that suggests, like, they have thought about it and they know what to highlight and what's important, but then it's just sort of random things pulled out of the materials. It's bad. It's really bad.

Paris Marx:

Google's coming for my job. Like, this is not cool., Alex Hanna: AC Zhou in, in the chat a musical theater opening to AI Hell." Yes. Thank you,. Emily M. Bender: Yes. So true.

Emily M. Bender:

Also, you gotta wonder, like, what was the training data for these? Like they're probably using a bunch of YouTube data since it's Google. Um, but the other thing is that I think from the ones that I listened to, I think there is a template. That they are then, like, it's probably a series of prompts, um, that they're using, sort of, with the, you know, the input data. And in this article, people are like, well, yeah, this is good because we can control what it's talking about because we're only giving it good data. And once again, papier mâché out of good data is still papier mâché. Like, stop it. Speaking of art.

Alex Hanna:

This one's fun. So this is from Gizmodo, famous, "Famous AI artist says he's losing millions of dollars from people stealing his work." Uh, the subhead reading, "The guy who used MidJourney to create an award winning piece of AI demands copyright protections." This is by Lucas Roper, published on October 1. And this is the guy who created that piece of AI, that art that was reported in the New York Times. I think that one, some like a, like a, it--was it like the Iowa state fair or something?

Emily M. Bender:

Somewhere in the Midwest.

Alex Hanna:

Yeah, it was, um, I mean, not--

Emily M. Bender:

Oh, Colorado.

Alex Hanna:

Colorado. Sorry, not to, sorry. Those are very, two very, very different states. We will not like contribute to, you know, uh, erasing the Midwest on this podcast. But so, um, so let's see the, what's the first, what's the first paragraph, the lead, what does it say up there? So it says, uh, "Much consternation spread through the artistic community two years ago, when Jason M. Allen, an executive at a tabletop gaming startup," still already cursed,"submitted an AI generated painting to a Colorado digital art competition and won. Critics claimed that Allen had cheated, but the prize winner didn't have much sympathy for his detractors. Quote, 'I'm not going to apologize for it,' Allen said. Quote, 'I won and I didn't break any rules.' He also didn't seem to care much for the complaint that AI companies like MidJourney, the one he used to create his quote 'painting,' were poised to destroy the art market. Quote, 'This isn't going to stop,' Allen told the New York Times. Quote, 'Art is dead, dude. It's over. AI won. Humans lost.'" And then, uh, if you scroll down, it's, "Now in an ironic twist, Allen is upset that his work, which was created via a platform that's been accused of ripping off countless copyrighted works, cannot itself be copyrighted. And it's thus getting ripped off." Uh, schadenfreude increases. In March of last year, the US Copyright Office ruled that work derived from AI platforms contains no human authorship and therefore cannot be extended copyright protections. And then he's been trying to, you know, follow suit, et cetera, et cetera. It's really, uh, really great stuff.

Emily M. Bender:

Crocodile tears.

Paris Marx:

Yeah. Tiniest violin. Like, oh my guy, do not care. Yeah.

Emily M. Bender:

Yeah. All right. One last one from Wired from October 15th, 2024. Sticker is "gear." The, uh, author of the journalist is Boone Ashworth. And the headline is, "This talking pet collar is like a chatbot for your dog." Subhead, "A new smart collar aims to give pet owners the ability to talk to their fur babies, or at least fake it." Um, this is one of the, um, bits of quote AI hardware that's supposed to be released next year in 2025. And having read this, basically it's a prerecorded set of about 8,000 things. So it's not a large language model with a, with voice synthesis. It's actually, they got voice actors to record some stuff and you talk to it. And then you can also set like things like its politics ahead of time. And that like subsets--

Paris Marx:

Very important for my dog. Yeah.

Emily M. Bender:

Yeah. Then you have this conversation with your pet, but it's not with your pet. It's with the color that they're wearing. So that seems totally pointless.

Paris Marx:

Oh, I didn't understand. Okay. Now I get it. It's pretending to speak as your dog.

Emily M. Bender:

Yeah, but it's not your dog, but it also has a microphone on it, partially because I think it is responding when, when spoken to, but also it's got these luxuries surveillance features of like checking how often your dog is eating and drinking and stuff like that. And so you've got like always on microphone combined with something that sounds like it's coming from your dog, but isn't, like it's it's ridiculous.

Paris Marx:

Well, we all know how well AI hardware is doing lately. So I'm sure this will be a astounding success Yeah,

Emily M. Bender:

Yeah, and I don't know if you can hear Euler, but she's coming through with some commentary on that, genuinely from her.

Paris Marx:

Yeah.

Alex Hanna:

I'm reading in this and it's also seeing that the collars themselves start at $500 for cats and $600 for dogs. Well, sorry, 595, 495 and 595. And then there's also subscription fees, um, which, you know, yeah, luxury surveillance indeed, uh, when you come to it and yeah, that's so just really, um, and they, yeah, and it's called Shazam, as if that's not an overloaded product name to begin with. There's like 18 different products named Savan, uh, Shazam. So.

Emily M. Bender:

Yeah, and this, this, uh, subscription service, apparently, um, uh, let's see, you know, "the brain boost service is what brings all of the truly sentient quantities, such as empathy, reasoning, social awareness, and self awareness" to the collar on your, like, it's so absurd.

Paris Marx:

Who needs this? Like, I do not understand.

Alex Hanna:

This is really one of those conspicuous consumption sort of things that's going to be popular in like Palo Alto and nowhere else.

Emily M. Bender:

We'll have to watch for an update next year to see if it actually gets released. Um, yeah.

Paris Marx:

Or how quickly it implodes if it does. Yeah.

Emily M. Bender:

Yeah. Yeah. All right. I've got one last quote from the comments before we go to our outro. But, um, Abstract Tesseract says,"If irony was fuel, this would have enough power to push the earth six times further away from the moon."

Paris Marx:

We know it's just going to keep getting further away, right? Especially as we proceed more AI.

Emily M. Bender:

Yeah.

Alex Hanna:

I know, right?

Emily M. Bender:

All right.

Alex Hanna:

Anyhow.

Emily M. Bender:

Hey, I get to go first, Alex.

Alex Hanna:

You get to go first, yeah.

Emily M. Bender:

So that's it for this week. Paris Marks is a tech journalist and the host of the podcast Tech Won't Save Us. Paris, thank you so much for joining us today.

Paris Marx:

Absolutely. It was such a pleasure. Thanks so much.

Alex Hanna:

Pleasure's all ours. Our theme song was by Toby Menon, graphic design by Naomi Pleasure-Park, production by Christie Taylor. And thanks as always to the Distributed AI Research Institute. If you like this show, you can support us by rating and reviewing us on Apple Podcasts and Spotify, and by donating to DAIR at DAIR-Institute.Org. That's D A I R hyphen institute dot O R G.

Emily M. Bender:

Find us and all our past episodes on Peertube and wherever you get your podcasts. You can watch and comment on the show while it's happening live on our Twitch stream. That's Twitch.TV/DAIR_Institute. Again, that's D A I R underscore institute. I'm Emily M. Bender.

Alex Hanna:

And I'm Alex Hanna. Stay out of AI Hell, y'all.

Emily M. Bender:

Meow.

People on this episode