'You learned the wrong way therefore the result is bad for science' What matters is whether the paper is good and contributes something real. How the author learned the subject is not an accurate proxy for that.

Writing good fiction with AI is hard. I dont know if any of what I've written so far is objectively good, but i know it's a lot better than slop and my latest attempts are a lot better than my first attempts.

What's sad about this and the inner quote is that boilerplate, flattening and hollowness are not inevitable consequences of using an LLM in ones writing process. But the tool makes it so easy to be lazy, most use it lazily and you loose the person behind it.

Using AI to support peer review seems unavoidable, but what quality checks should AI implement? We can take some lessons from metascience on the hard reward design problem that is AI review. I wrote a paper synthesizing a few points the emerging lit seems at risk of confusing. 1/

This is really easy to shift in harnesses. It was one of the first things I started to set specific instructions around

LLMs produce average answers given their context. What makes interesting answers is when their context is far from average.

Universities are arguably the key institution in society for marshaling the cognitive medium of writing to generate new and valuable knowledge and training people in that cognitive medium.

These are not the only 2 options. If you use AI and your talk is equivalent to an AI summary anyone can get, fully agree. But some people use AI as an interlocutor to stress test and refine their thinking and ideas and the result is nothing like an AI summary...

Once again for those at the back: offloading is bad, but not all AI use is offloading. This situation is a bad case of offloading BUT highlights a good rule of thumb: the measure is not if you used AI, its whether you can stand up and defend what you're saying and know it inside out.

This is the right response to the main thread.

This whole fable thing is Hasdrubal's inspections... marenonnostrum.com/idioms.html

Though the core thing both are grappling with is that AI enables low effort and low quality work to be made with ease so both get flooded with work that wastes people's time. But the whole discussion presupposes that genuinely high quality work is impossible with AI in the loop.

Also, the crutch: en.wikipedia.org/wiki/Shepher...

I wonder if people laughed at early programmable computers while pointing to all the tasks done at the time by analogue computers... General purpose systems which can do many many things have significant potential even if they are always beatable on some specific task.

"Every one of these issues deserves serious attention. None of them are well served by 'your ChatGPT query is killing the earth.'" But have you considered attention is hard and slogans are easy?

Car park canopy solar is 3-4 times more expensive than flat land utility scale solar. For the same price you can double the energy plus get a battery big enough to save energy to cover evenings. Consider that fossil fuel companies want us to spend more money on making less energy...

So an article about fact checking by a professional fact checker is factually misleading at best. And the media think this insanity is a viable way forward.

Using LLMs to simulate readers while writing a new book to refine how the plot progresses is partly undone because LLMs have general knowledge too broad compared to most real readers that they cant *not* see certain things...

If your interest in environmental issues closely tracks your interest in other topics and not the actual harms, you're not doing environmentalism, you're doing motivated reasoning.

If you actually care about the environment you care a lot more about these than data centers. Otherwise you are cosplaying environmentalism.

A fun version of this when writing fiction is them writing as if your editing exchanges are in universe public discourse which can and should be referenced...

I strongly support requirements for data center build out to be mostly renewable powered. That said, because economics is complicated [citation needed], 100% might not be optimal depending on your goal...

the arXiv stuff is a such a perfect litmus test sorry, antis, people should be able to use AI to help with research sorry, bros, people should be held accountable for the quality of their work I don't understand how any of this is controversial

Can you learn to see the world like Clay (Earth Observation FM)? Try playing ChipMatcb here: chipmatch.apps.lgnd.hosted.unionai.cloud

Ironically deliberate deceitful framing is actually quite a useful tactic in AI usage to mitigate sycophancy.

To put this challenge in perspective, chatgpt's rough back of the envelope would put converting the entire Midwest crop land over to just producing feedstock for bioplastics as not even 50% of global plastics demand. I doubt that's precise but we are not talking an insignificant amount of land.

Everything Is A Sequence Of Tokens

This is good news but absolutely doesnt mean "we don't need petroleum based plastics anymore" because the core problem isn't the material science, it's supply chains. We use a lot of plastics. We need a lot of the raw material used to produce them. 2 relevant feedstocks: petroleum or plant matter...

Also: humans aren't deterministic either and we dont use that to conclude humans are as a category useless.

Someone said they'd rather have lawns than data centers and they clearly Do Not Know About Lawns

"that's not a fair comparison, this would be like if a private corporation owns the lawn, doesn't let anyone except the rich use it, and we gave them tax breaks to do it" yeah golf really does sound crazy when you simply describe it plainly.

One particularly interesting one was where Claude chose a particularly strained but technically not invalid reading apparently because it was desperate to make the motivations and actions of an AI in the story legible and defendable....

Interesting LLM fiction interpretation weirdness: I'm writing a(nother) fiction book and run the draft manuscript though LLMs to see how it interprets certaim structures and themes. Often the question is "does the LLM fail to get X because it's an LLM or because I've not written well".

Let's do the test: 'Rather than memory palaces, dialogue with a *human* produces Ptomkin villages of intelligence, giving you the emotional satisfaction of believing you have insight while requiring none of the effort." True of much human dialogue. Doesn't mean real learning impossible with humans.

When I was at school I remember finding an advanced physics text book full of really interesting problems. I used a calculator because the arithmetic was not the point and was a distraction from engaging with harder content and higher level reasoning.

This sounds very familiar to me. I spend a lot more time reading, writing, speaking and listening to natural language than before and much less time coding. Dialogic free writing is a good phrase.

really good from @grimalkina.bsky.social. I’ve been thinking about writing something about a habit or maybe skill I have, which is that I’m good at treating things I hear as tentative and holding onto them for later confirmation (or not) www.fightforthehuman.com/cognitive-he...

guys, you’re missing the point. AI doesn’t have a purpose like parking lots, which help me fit fossil fuel burning more conveniently into my day

oil companies watching "environmentalists" focus all of their attention on data centers

To clarify this isn't to say it's bad if organisations pay for canopy solar. Just that the framing of it as a major solution as if space were the main constraining factor rather than being a consideration along side cost.

Small reminder that parking lot canopy solar is MUCH more expensive than utility scale field solar. For the same investment you'll get several times less capacity. It mostly solves for not having to think too hard about broader land use policy and economics.

This is not "whataboutism". If you are serious about these issues you need to know what the levers are, all of them, how big they are, and the costs/constraints of pulling them. If you're only thinking about one lever, you're thinking about the lever, not the system.

There are a lot of people who by their actions and blue whale sized omissions in their concerns show they don't actually functionally care about the environmental issues they say they care about.

I guess we're going to need to learn the hard way that you can't eat cotton... nor the varieties of corn grown for ethanol.

Everyone was harping on about model collapse, but as far as I can tell this is the clearest example of the proposed failure mode and while weird, hardly counts as collapse!

The question of LLMs is not "what can it do for me" but "what kind of cognitive ecology can I build with this?"

In parallel, I'd expect anyone in a position of authority to be able to articulate *why* on this matter. If your argument for changing a scientific classification is "because i have the feelies", you are too immature or feeble minded for public office.

The data center and renewable energy discourse reminded me of a thought experiment a explored a few months ago:

Rage as national newspaper uses manipulative framing for political ends.