Nathaniel Virgo

My guess is it's much more than the annual budget to run arXiv.org, so my proposal is that the UK should found its own archive server for disseminating publicly funded research and stop giving taxpayer money to private companies who charge thousands of £'s to put a pdf file on a web site.

Does anyone know a good estimate of the total amount spent on open access fees / article processing charges by the UK government (or really any government)? I'm only interested in public funds, i.e. total cost to the taxpayer.

A while back I made my own tool for conjugation drills, but it kept saying things like "死んでください" (a polite request to die)

Flowers quash all bad feelings, indeed.

In the Gremlins adventure game on the Commodore 64, if you type "run" in the first room you'll end up in a weird glitched out room whose description is "I'm in a." You can wander through these weird rooms a bit and get to a later part of the game, but you won't have the items you need to proceed.

Oh, that looks very promising, thank you! I also stay away from subscription models, but I might try it.

(This thread brought to you by the fact that I just closed all my browser tabs in an effort to keep the room cool)

Basically, "I want to pause what I'm doing and look at another web site for a while" shouldn't mean the same as "I want to start a new browser process and leave all the Javascript running on the page I'm currently looking at". We must be collectively wasting huge amounts of energy because it does.

The interface could still look like tabs, it's just that they wouldn't run in the background unless you specifically request it. And there would be a mechanism where old tabs disappear and move into a nicely organised history list - but you can still restore them in the state you left them.

Advantages: (i) laptop not as hot in the summer; (ii) goodbye to "oh no, I've left too many tabs open" - it's basically like having infinite tabs but you don't have to worry about it; (iii) no problem if your browser crashes / you have to reboot - history gets retained and tabs = history.

I wish there was a browser that just didn't have the concept of tabs at all, but instead just kept track of your history in an organised way, such that you can always restore the state of a page you visited previously but never actually have more than one process running at once.

I always understood it as meaning taking multiple exposures at a different exposure level, in order to capture a higher dynamic range than a single exposure, regardless of how you go on to post-process it - the weird compressed look isn't mandatory. idk if the meaning has changed recently though.

in lineages descending from it. I don't think he mentions extinction explicitly, but that would make complete extinction of organisms with the trait less likely, because its descendants are adapted to many different environments, so statistically some can survive.

(3) it puts me somewhat in mind of Dawkins' evolutionary watershed idea (from before he became so dogmatic). See www.richarddawkins.net/wp-content/u... from the first ALIFE proceedings. He tells a story where a complex novel trait (the arthropod body plan) leads to a wealth of different adaptations

(2) a threshold might not be the right way to think about it. There are traits that make you resistant to extinction, and if you have a lot of them working in synergy it might add up to 'complexity'. But it's surely also possible to be complex in ways that don't help you survive extinctions as well.

(1) It's surely true that *large* organisms are more likely to go extinct in mass extinctions - that's an observation from the fossil record AFAIK. But that might not be directly about complexity, e.g. it could be more about the energy budget to sustain them.

I don't know any work on it specifically (and it's a long time since I was into the literature) but here are some thoughts:

(which is more or less what you said of course)

It must be at least partly because most old kinds of lights (tungsten, arc lights, flashbulbs, ...) work by either burning something or heating something up til it glows, so they emit black body radiation to a good approximation.

If it doesn't seem to understand it first time I never had much luck getting it to grasp it in a longer conversation, and I diagnosed this as being because the idea isn't in the training data. But I could be wrong about that - I don't really know how to test it.

I don't mean autonomous research, more like "I've had this idea but I don't quite see how to make the definition work, can you help me out" sort of thing. Sometimes this works well, but I assume it's because something similar *is* in the training data even if it seems really obscure to me.

In the past, I've found they're pretty good at telling me things that I don't know but someone probably does, but when I've tried using them to explore probably-novel things they sent me down blind alleys by saying stuff that was wrong in really subtle ways, so I tend to avoid it.

Do you feel they're any good at proving novel things yet, or are they just better at reproducing and restyling proofs that exist somewhere in the training data?

3968

Queen h4, queen takes, then bishop f3 is checkmate

Plus there's no guarantee your code will run the same if you shut down the notebook and open it again, because it might be relying on data from previous things you ran that aren't in the cells any more.

So the more we accept AI as part of our culture (or part of our bureaucracy, as the UK government explicitly wants to do), the less we'll be collectively able to innovate in those fields. It won't just freeze words, it'll freeze the whole way we conceptualise problems and formulate answers to them.

The printing press had a similar effect, so it's not really new. But still, I'm worried that AI will have this effect times a hundred. LLMs can only statistically regurgitate stuff from the training data, and if a certain kind of writing is only done by AI there will never be any more training data.

Maybe I really want to make a deeper point here. "Counterintuitive" is absolutely a word, but my computer wants me to say "counter-intuitive" instead. Spellcheckers can be useful but they also stifle innovation in language, freezing it in whatever state it was in when they became widespread.

Looking back at my notes I don't think I was successful. But if you have a hole in your life where it seems like something like that would fill it, then maybe we should talk.