AI is definitely changing maths -- senior/established researchers can now use AI models to speed up their work, which is going to affect early career researchers. At a recent maths and open science conference I attended [1] virtually every speaker detailed how they'd used AI to generate new results/ideas etc.
The change is happening, I don't think we can stop it now.
Yes, I share your optimism overall, although I think it is raising a question of what the future role of the researcher is (much like the current debate on developer roles).
I attended a conference on AI for maths and open science a few weeks ago, and was struck by just how many examples of AI-supported solutions there already are. Virtually every speaker had an example of either their own use of (often the frontier) AI models in solving a problem that was previously too hard (for various definitions of hard).
I wrote up a few notes [1], and most of the speaker videos are available via the conference website [2].
This is such a well written story, and congratulations Ben, it sounds like it's been a lot of hard but ultimately successful work!
I know you'll deservedly get a lot of credit for all your work in remastering the game, but you should also get credit for how you've woven this narrative together, it's a lovely read. Thank you for taking the time to write it up, and good luck with the Steam release, and whatever project you take on next! :)
Thanks for the quick reply -- yes, I imagine there is some prior art for this, although I'm often amazed at how hard it still is to print (even just to print to save a pdf) from many websites!
Thank you for working on and sharing this project, it's a wonderful example of how maps help tell / connect the stories through the years. It was also nice to see the piece you drew inspiration from that you linked to in another comment.
I'm curious -- each section (after the transitional animation) seems like it translates quite well to being a fixed page, and I'm wondering if you've looked at there being an easy way to generate a version that could be printed?
The reason I ask is that I'm working on a family history project which involves various locations, maps, newspaper stories and old photographs, and whilst something interactive feels like it would be an engaging way to do it, for the long term it's always good to have a printed copy of anything :)
I could imagine if there was an easy way to generate a print version of your articles, some people would pay for a copy (that was printed professionally in a suitable photobook-style format).
Apologies if there is a simple way to generate this and I've missed it in my cursory glance around the code :)
> For example, they seem to not care about instructions to only write a response and no explanation, thus making it impossible to use them in a pipeline
I was doing some local tidying up of recording transcripts, using a fairly long system prompt, and I saw the same behaviour you mention if the transcript I was passing in was too long -- batching it up to make sure to be under the max length prevented this.
Might not be what's happening in your case, but I mention it because it wasn't immediately obvious to me when I first saw the behaviour.
Ah yes, I missed some important context - one of the reasons for moving to a name without LaTeX in it was because we'd just released the first beta version of the visual editor (the rich text mode at the time), and the goal was to keep lowering the barriers to getting started with LaTeX, and to make collaboration easier for non-LaTeX users. And so Overleaf came from a search for a broader name than writeLaTeX.
That's an interesting point about the pronunciation - overleaf is a standard English word, and certainly seemed less confusing than writeLaTeX when said aloud, but I agree it's not perfect! Was the best we could find at the time (especially given the other needs mentioned above).
Took us about three months to pick Overleaf, we went for it largely because it was a single word, hard to confuse/mispell when said aloud (unlike writeLaTeX), had a connection with writing "over the page", and probably must importantly, we could get the .com domain.
Thanks Dang! You're right that's probably the better link to use, rather than the arXiv version -- I just have a liking for PDFs, which is why I linked to the preprint :)
Thanks w-m, glad you find Overleaf useful :) I can't quite believe it's been ten years since we started out -- time certainly does fly by.
I did appreciate the somewhat circular loop in posting this, so thanks also for noticing! And my H/T for spotting this originally goes to my friend & colleague Alf Eaton, who reshared Lukas's tweet: https://twitter.com/lukas_blecher/status/1696101110853910716
and in case anyone's interested in the meandering thoughts of two start-up founders in the scholarly comms space, Mark and I had a nice chat following a recent figshare milestone :) [1]
The change is happening, I don't think we can stop it now.
[1] https://scholarlyfutures.substack.com/p/ai-and-the-practical...