This story about backlash to an earth science specific chatbot at EGU seems to detail a lot of insider politicking that seems only obliquely related to the concerns over the use of Large Learning Models for scientific research and writing. I don’t understand the context of all this politicking enough to comment, beyond saying that if the questionable use of large amounts of copyrighted material to train an LLM is not the most ethically questionable thing being discussed, you’re probably not doing well.
Unfortunately, all the drama does seems to have actually muddied the serious discussions that need to be had about whether we really want ChatGPT and its cousins getting inserted into the research pipeline. My specific issues with using chatbot-style interfaces in research are three-fold:
Utility: I am not particularly excited by summarization, because regardless of whether the information provided is cited or uncited, I could never trust the summary to be accurate (this is a fundamental consequence of how LLMs work and don’t let anyone tell you otherwise). I would need to check everything in the summary, just like I would for a student paper or thesis. In that case, the annoying extra work is justified because you are training someone, in this case, it would just be annoying.
Misuse: of course, if you don’t care about accuracy (or are driven by the perverse incentives of the modern academy to deprioritize it), having ready available tools that can write large parts of your papers for you is an invitation to flood the world with a tide of pointless sludge. Writing scientific papers is a creative activity, because you are trying to say something new about the world. This is a difficult and frustrating (if ultimately very satisfying) process for scientists; it is impossible for an LLM. Generative AI can reproduce the form of a scientific paper, but cannot generate the new insights that are supposed to be at their heart.1
Environmental Impact: Earth Scientists in particular need to consider the disproportionate consumption of energy and water resources for training and deploying the LLM models that return this sludge of questionable accuracy.
To be clear, I don’t think all machine learning/AI is useless. It’s the generic “one chatbot to rule them all” approach that I think is flawed, and a corrupting influence besides. It seems to me that AI tools that help us to interpret data, or find relevant literature, should be more geared to giving us prods and signposts, rather than just doing our job for us.
The misuse of generative AI by professors is far more problematic than misuse by students, despite much more hand-wringing over the latter. Students have been rapidly throwing together random phrases without understanding what any of it means – and earning mediocre grades for it – since time immemorial.
After 15 years of discussion and exploration…Twelve members of the International Subcommission on Quaternary Stratigraphy (SQS) voted against the proposal to create an Anthropocene epoch, while only four voted for it.
To be honest, this outcome has always been a strong possibility. There has always been a tension between the unquestionable fact that humans are now a geological force that is measurably altering our planet (the Anthropocene concept), and the debate over how you draw the arbitrary line in the geological record that codifies our rise to geological significance (the Anthropocene epoch). The long-standing contradictions that arise from using the same word as an evocative thematic header for the climate crisis, and also as a proposed addition to the precise and fussy terminology of chronostratigraphy, have now come to a head, and in the worst possible way1.
There has always been an undercurrent of resistance to this concept amongst the people who sit on the International Committee of Stratigraphy (ICS), of which the SQS is a part. The establishment of the Anthropocene Working Group in 2009 always felt a little grudging, the clear enthusiasm of some members like group chair Jan Zalasiewicz notwithstanding. And scientifically, I can understand some of their unease: trying to codify the effect of humanity on a geologic record that is still in the process of forming, on a planet that is still in the throes of responding to the giant poke we are giving it, does fly in the face of procedures and methodologies used to define boundaries deeper in geologic time: in a sense, we are trying to define the ‘golden spike’ for a still-moving target.
That said, I have long been extremely uncomfortable with the way this has been covered in the media, who have spent the last decade breathlessly covering ‘the search for the golden spike’, an approach which has completely blurred the line between the fact of the Anthropocene concept and the debate over defining its boundary. As I have argued in the past:
The worry was always that the media would treat the refusal to draw a line in the sediment that officially says ‘Anthropocene starts here’ as a rejection of the idea that humanity’s hands are now on the planetary steering wheel. And here we are:
Image by Joana Rodrigues on LinkedIn, via Catherine Russell on Bluesky.
What happens now is important. The risk is that the SQS and ICS remain focussed on their internal debates, and don’t consider the impact of what they have decided on the current moment, where certain interests are willing to latch on any thing they can use to delay decarbonising our energy system. Geological history has been weaponised before (‘geology says Earth’s climate has always changed!’), and they will try to weaponise this, too.
So I dearly, dearly hope that they do not just look inwards, but reach outwards. They actually have a small window where they can grab the microphone and speak loudly and clearly that – golden spike or no golden spike – we are living in the Anthropocene2. It’s just too early to say exactly how the transition to the planet we are making will look in the rocks, tens of millions of years hence. But don’t doubt it: there’s going to be some really weird stuff for cockroach geologists of the far future to puzzle over.
As the Nature article discusses, there is some noise being made about how the vote happened, but 4-12 is not exactly close – and even if it did somehow get wrangled over the hurdle presented by the SQS, there are still two more: a full vote by the ICS, and ratification by the International Union of Geological Science. And if you can’t win over the people who study the Quaternary – who have had their own battles with the ICS – I wouldn’t be optimistic about convincing people who think the Cretaceous is young.
Even if they’d really, really prefer we not call it that. That ship has sailed.
The early months of the COVID-19 pandemic introduced us to the idea of ‘bending the curve’: that acting early to reduce infection rates made a huge difference in whether the peak in infections was manageable or not.
The exponential nature of disease transmission, and the 1-2 week incubation period in which COVID-positive people could infect others without showing any symptoms themselves, meant that the best time to start lockdowns, social distancing, and other actions to slow the rate of transmission was so early on the “very bad things are about to happen” curve that the bad things had yet to really manifest. We had to act when the absolute number of reported cases was very low, because that was the number of people who had been infected two weeks ago, and the next two weeks of exponentially increasing infections had already been locked in. The higher we ascended the slope of bad things before acting, the higher we would have to climb before we could start descending again.
“Bending the curve” when, absent some corrective action, very bad things are predicted to happen (orange line). When there is a lag between action and effect, the key decision point where corrective action will largely prevent these bad consequences (green circle and line) occurs long before things actually get bad; waiting until then (orange circle and dotted line) locks in a lot of misery.
To everyone’s cost, the optimal course of action from the public health perspective proved extremely difficult to implement and sustain. We change our trajectories to avoid bad things all the time: every time we steer around an obstacle or duck a low hanging branch, we are bending the curve of our lives towards a better outcome. But in those cases, the danger is clear and present to us, and the hazard we are avoiding with our corrective manoeuvres passes right before our eyes. Our instincts are not well-tuned to collectively turning the wheel when the danger is only visible on maps that very few of us can read; and when there is too large a lag between our actions and the consequences that flow from them, we can find it hard to even connect the two.
So our response to the spread of COVID-19 was often reluctant, and very much contested. In many, many places we did not turn those steering wheels as much as we could have, as soon as we should have. Our curves were more like the dotted orange line in the sketch above than the green one: we could have saved the lives and long-term health of many, many people if we had acted sooner. And yet, there was still a world of difference between acting later than we should have and not acting at all. In the end, we did bend those curves – and thank goodness we did.
In just the last couple of weeks in Vermont, we have had air quality problems due to wildfires, a heatwave, and now the worst flooding in decades.
The story of the summer is that this is not an outlier – everywhere you look, you see extreme weather events…
…Welcome to the world of the unbent climate change curve.
For that is indeed the world we live in: if we map the climate crisis on to my sketch above, the off-ramp to take the green curve was in the 1990s and early 2000s. We did not take that exit, and our failure to check the growth of greenhouse gas emissions since has led to planetary heating very close to the modeledpredictions from that time.
Our late start on trying to bend the curve means that we are committed to some climate change: the state of the planet in 2100 will be different from the state of the planet in 1900, with the consequent disruptions to the civilization we built assuming that previous state would persist – disruptions that will cause much loss and suffering.
In its 2020 renewables report, the IEA forecast an additional 1,092 gigawatts (GW) of global capacity would be built between 2022 and 2026. It raised this to 1,496GW last year… in its latest report, the agency estimates that an extra 424GW of renewables capacity will now be built over this five-year period… This is a 28% increase on the previous estimate and up 76% from two years ago.
Importantly, this has already made a difference. In 2015, we were headed for a 3.5ºC increase in average global temperatures by 2100, but the latest estimates have us on course for 2.5ºC:
“climate policies and clean technologies deployed over just the past eight years [since the 2015 Paris agreement] have already erased a full degree Celsius of global warming from the future world in 2100.”
Is 2.5ºC still too much warming for comfort? Of course! But our efforts to decarbonise, faltering and ill-matched to the moment as they often seem, have already measurably shifted our path away from the worst futures. What we do now still matters – a lot – in determining how different the Earth will become by the end of this century and beyond. As the figure below shows, our current emissions trajectory (the red line) still needs to get rapidly bent downwards to reach a path that will limit overall warming to 1.5-2ºC (green and blue lines). Even waiting just a little bit to accelerate decarbonization (grey line) will make at least temporarily warming the planet beyond these limits – with all of the disruption and loss that will come with it – much more likely.
If we keep trying – rather than surrendering to climate doomerism – we can keeping bending the curve of bad things further downward. We can make it crest earlier and much less destructively. We can make it so that when people a few decades from now look back on 2023, it is closer to “as bad as it gets,” than “you ain’t seen nothing yet.”
Hopefully this summer pressed home the lesson that calling Vermont a “climate haven” – as some do – is only potentially true in the sense of ‘less bad than elsewhere’.
It’s now been almost a year since NASA’s InSight lander – home of the first seismograph ever deployed on Mars – was declared dead. But the picture of the Red Planet’s interior being deduced from the four years of seismic data it collected continues to evolve.
A previous study identified seismic waves bouncing off a transition between solid and liquid, assumed to be the core-mantle boundary, around 1500 km down, giving the core a radius of around 1800 km out of a total planetary radius of 3390 km. This implied that in addition to dense, molten iron and nickel, there were also a lot of lighter elements like sulphur, carbon and oxygen hanging around in the Martian core – otherwise Mars’s gravity would be significantly higher than it is.
Between then and the end of InSight’s life, a marsquake on the far side of the planet from the lander produced waves that diffracted around the core mantle boundary. The travel time of these waves compared to ones that only travelled through the upper mantle did not fit the old model. To explain the new data, it is now proposed that the previously detected boundary between solid and liquid is in the lower mantle, and that the (also liquid) core is enclosed by a layer of molten silicates around 150 km thick. This reduces the core radius to about 1,650 km, which might not seem much, but reduces the volume of the core by about a quarter, allowing it to be significantly denser and have a more realistic composition.
This isn’t actually the first body in the solar system inferred to have a structure like this: the moon is also thought to have a molten mantle boundary layer of very similar dimensions above its core. In both cases, it impedes heat transfer out of the core, allowing it to stay hot (and molten) and also inhibiting magnetic field generation. Which explains why both the Moon and Mars – small worlds whose interiors should cool down pretty quickly – still have at least partially liquid cores.
The origins of the molten boundary layer itself are a little mysterious to me, so I’m not clear on why the Earth does not have such a layer. Is it a size thing, or a ‘our planet is weird’ thing? And here’s a speculative thought: we still don’t know why Venus doesn’t have a magnetic field. What if it’s because it also has a molten mantle boundary layer?
Here’s a new study that, if true, this would have some big implications for the nature of the sedimentary record for a lot of Earth History. Research from 1.2 billion year-old sediments in Scotland adds the oldest evidence yet against the idea that we had no large, single-channeled meandering rivers prior to rise of sediment-stabilising terrestrial plants in the Silurian, only braided ones with multiple active channels. Apparently cohesive clay minerals can stabilise a channel enough on their own.
Plants invading the continents does still have an effect, however: it turns rapidly migrating meandering rivers into less rapidly migrating meandering rivers, which leads to a big change in how these systems are preserved in the geologic record. The greening of the subaerial world would also still have had large impacts on geochemical fluxes from the continents due to increased weathering, and possibly also on timescales of sediment routing from source to sink.
Given that weathering patterns and rates were different, I do wonder about the rates and geographical distribution of clay production prior to the Silurian – it could be that whilst meandering rivers were possible, they were still rarer than they became post-greening.
Nice plan for content warnings on Mastodon and the Fediverse. Now you need a Mastodon/Fediverse button on this blog.
For lot's more videos on soil moisture topics, see Drs Selker and Or's text-book support videos https://www.youtube.com/channel/UCoMb5YOZuaGtn8pZyQMSLuQ/playlists
[…] Announcing STORMS | Highly Allochthonous on Recent News […]
Nice plan for content warnings on Mastodon and the Fediverse. Now you need a Mastodon/Fediverse button on this blog.