Climate Cosmos Bs
---
---
SIMON CLARK
CLIMATE COSMOS BS
0:00
This is a website called climatecossmos.com. At first glance, it appears to be a relatively normal website writing
0:06
articles about climate change. But spend more than about 10 seconds on it and you notice that things are a bit odd. Their
0:13
articles are very generic. They don't interview any experts. Their topics of articles are sometimes completely
0:19
bizarre, sandwiching articles with information about climate change between those that push an anti-science agenda.
0:26
And then they published a piece called um what the climate movement isn't
0:31
telling you. And the piece read, "I'll help you write an article about what the climate movement isn't telling you with
0:36
current facts and data. Let me search for the latest information." This website is AI slop. And this is Joey
0:43
Gsturn, a journalist at DMOG, the investigative reporting site specializing in climate misinformation.
0:50
A few months ago, I learned of Climate Cosmos and passed it over to Joey to do some digging. And what he uncovered was
0:57
a bizarre new form of climate misinformation. One that shows how the
1:03
information landscape around climate change is in danger and raises questions
1:08
about what we can and should do to safeguard that landscape.
1:18
To begin with, let's look at Climate Cosmos in a bit more detail. So, you'll have one piece that's like a
1:23
listical saying something like the top 10 snowiest cities in Hawaii or something nonsensey that's very
1:28
obviously like there to draw clickbait and then like sort of every third or fourth article would be like the secret
1:33
hidden costs of the green transition and then quote Bjorn Lomborg who is a kind of very prominent like anti-net zero
1:39
voice. So, we did more of a deep dive into the site, trying to look at who owns the site, who's behind it, you
1:45
know, all these kinds of things. On the about page, they listed staff members, and we found this site maybe 4 months
1:50
ago. Back then, the staff members had AI generated pictures and untraceable
1:55
social media footprints. So, at that point, it was kind of clear, okay, maybe these aren't even real people. Things have changed. They seem to have aligned,
2:02
swapped out these AI generated profiles for real people or what what we think of real people online. We did some digging
2:07
into who owns the companies. It turns out Climate Cosmos is part of a group called NextVision Media. This is a German startup media business. Next
2:15
Vision Media don't just do climate content. They have a few AI generated sites. There were like home decoration sites. There was all kind of like really
2:21
random seeming topics um where you'd think, okay, if this is a a media business dedicated to climate content,
2:27
why do they have a site called Home Harmony that's about how to like decorate your living room? And all of these websites feature AI generated
...
this story gets more interesting. MSN is one of the world's largest news aggregators. That means that it's a web
4:30
violated its policies on misinformation. And so when Microsoft was made aware of
4:35
the content, it was all pulled. But via the massive reach of MSN, Climate Cosmos's content had escaped into the
4:43
wild, where it was seized upon by an entirely different kind of actor in the
4:49
climate information landscape. From the 1960s, it was very clear to
4:57
scientists working in the fossil fuel industry that the fuels they were selling were leading to carbon building
5:02
up in the atmosphere. In fact, in 1959, physicist Edward Teller warned them that
5:09
this would lead to global warming and sea level rise. In 1965, President Johnson's science advisory committee
5:15
published the Restoring the Quality of Our Environment Report. President of the American Petroleum Institute, Frank
5:21
Ikard, said of the report, "One of the most important predictions of the report is that carbon dioxide is being added to
5:27
the Earth's atmosphere by the burning of coal, oil, and natural gas at such a rate that by the year 2000, the heat
5:33
balance will be so modified as possibly to cause marked changes in the climate beyond local or even national efforts."
5:40
And he did not dispute the links between fossil fuels, CO2, and global warming. In retrospect, it's kind of surprising
5:47
that the fossil fuel industry didn't immediately start disputing those links and climate science as a whole. Because
5:54
after all, if it became public knowledge that their products were threatening the
5:59
climate stability that we all depend on, their industry could just cease to
6:04
exist. Perhaps it was because no government was seriously considering actual climate policy until the 1980s.
6:11
And it was at this time when the existential threat to the industry became clear that the fossil fuel
6:17
companies started fighting back with all the resources of the richest, most
6:24
powerful industry the world has ever seen. They financed think tanks and
6:29
individual scientists who disputed the scientific consensus, saying variously that climate change wasn't real. real
6:36
but not caused by the fossil fuel industry or real but actually beneficial to humanity. Their techniques included
6:42
use of experts outside their fields of study, lobbying policy makers, portraying environmental campaigners as
6:48
outofouch liberal elites, harassment and legal action against individual scientists, and accusing scientists of
6:54
accepting money from industry while paying other scientists to produce studies that supported the industry's
6:59
continued survival. The fossil fuel industry was using a playbook to dispute the science that had been honed by
7:05
previous industrial misinformation campaigns, notably by the tobacco industry when they disputed the science
7:12
linking tobacco smoke and lung cancer. In fact, the fossil fuel industry used some of the same experts from that
7:19
tobacco industry campaign. I would highly encourage you to read Merchants of Doubt by Areskis and Conway, which is
7:26
the authoritative account of how this playbook came to be. Though for more recent reporting, you can also read the
7:32
Union of Concerned Scientists climate deception dossier or most recently
7:37
Barbara Fries's Industrial Strength Denial. The result was that from the 1980s, the climate information landscape
7:44
became a battleground. On the one side, you had climate scientists and institutions taking observations, doing
7:51
analysis, and projecting what kind of damage would occur to the environment if we continued burning fossil fuels.
7:57
Victory for this side meant getting accurate information about climate change to the public and ultimately to
8:03
politicians. The end goal being government policy to...
10:27
within the climate movement. Now as part of the article that Joey wrote about this investigation which is linked below
10:33
by the way he reached out to the three people he could confirm were actually real and not AI generated associated
10:41
with the site and this is what Matias Binda director of Climate Cosmos had to
10:47
say. We have no interest in publishing or amplifying misinformation. The articles you referenced were taken
10:53
offline as part of our ongoing quality review and we are tightening our internal checks to ensure every author,
10:59
biography, citation, and data point is fully verifiable before publication. And
11:04
indeed, all the articles that we've referenced so far in this video are only accessible through the Wayback Machine.
11:10
After Joey reached out to Climate Cosmos, all of their climate content and the content from its sister sites
11:16
disappeared from MSN, which you might take to mean that they've learned their lesson and will now be more careful in
11:22
their use of AI. But since Joey reached out to them, they have posted more articles in the same vein, such as this
11:28
one, highlighting the views of repeatedly disproven climate skeptics Richard Linden and John Christie. And
11:35
they now have a YouTube channel. Hopefully, I don't need to tell you that this is also AI slop. I mean, just look
11:42
at how frequently they're releasing videos and everything. As it stands, we really
11:48
don't know what the motivations behind Climate Cosmos are. The best clue we have is looking at one of the other
JAN OTTE
11:28
one, highlighting the views of repeatedly disproven climate skeptics Richard Linden and John Christie. And
11:35
they now have a YouTube channel. Hopefully, I don't need to tell you that this is also AI slop. I mean, just look
11:42
at how frequently they're releasing videos and everything. As it stands, we really
11:48
don't know what the motivations behind Climate Cosmos are. The best clue we have is looking at one of the other
11:54
definitely real people involved. Yan Otte. Otter, who owns a third of the
11:59
business, has repeatedly posted on social media about the role of AI agents in generating wealth and has reposted
12:06
political content, including from Elon Musk. He also posts on his LinkedIn page about the watch time and views received
12:12
by his various publications. As far as we can tell, he doesn't have a strong stance on climate action one way or
12:19
another. Which leads me to believe that the people behind Climate Cosmos are not anti-science, climate misinformation
12:26
activists, but tech bros trying to make money with AI. And how harmful can that
12:32
be? What the arrival of LLM has done is it's...
13:06
checked and like legitimate and accurate. It does not bode well for the future of journalism. Climate Cosmos is
13:11
not alone. There are increasing examples of AI generated content within the climate information space being used
13:18
both by those explicitly opposed to climate action and those who are just
13:24
seeking to profit from advertising revenue on sensationalist articles that get clicks because climate change is a
13:31
hot topic. No pun intended. Kimcosmos may seem harmless enough, but what they
13:36
are doing is, to quote Steve Bannon, flooding the zone, just pumping out
13:41
content, making it harder for the public to find accurate information about climate change written by experts.
13:49
Intention doesn't matter. The zone is being flooded with this mediocre, confusing, not fact check content. Um,
13:54
and that does crowd out good accurate climate content like that from Carbon Brief or you know any of the thousands
14:02
of good pieces of climate like science and journalism that are being done all the time. And this problem is only going
14:08
to get much worse. Not just because AI content can be churned out much faster
14:14
than expert-driven factchecked content, but because of the influence of AI
14:19
generated content on the next generation of generative AI tools. You see, the way
14:25
that AI tools are trained is through passing vast amounts of data and recognizing patterns in it. In the case
14:32
of a large language model, that means being fed a lot of text. And historically that has come from books,
14:38
transcripts of videos, and of course websites, not all of which were stolen for this purpose. But as the training
14:44
data sets used to create new large language models get larger and larger, AI companies are facing the possibility
14:51
of running out of new data to steal. And this paper from Nature estimates that
14:56
this could happen as soon as 2028. To get around this, companies are increasingly training their new AI tools
15:03
on data that was previously generated by an AI tool. Combined with the fact that
15:08
high quality training data, such as from academic journals or rigorous reporting,
15:13
is often behind a payw wall or is on a website that has a robots.txt file that blocks AI crawlers. The training data
15:21
used to create next generation models is increasingly formed of lowquality SEO
15:27
optimized and often AI generated content. And this generates a feedback
15:33
loop where misinformation can be amplified and become the mainstream. In a climate context, that means that next
15:40
generation LLMs risk paring the talking points of climate skeptic groups or
15:45
fossil fuel misinformation campaigns. In fact, this is one possible motivation behind climate cosmos. Creating large
15:52
volumes of climate skeptic talking points by just asking AI tools to churn them out such that it will form a larger
15:59
basis of the training data sets used to create future large language models. It's what we call LLM grooming. I should
16:06
say though that we have no evidence that this is what Climate Cosmos is doing. It's just a possibility. We still don't
16:12
know their actual motivations. Regardless, the information landscape around climate change is being flooded
16:18
by AI slop. Slop that is creating a feedback loop, intensifying the
---
---