What I hate most about going to the dentist's is the misaligned incentives. Well, that and the gum-stabbing.
Remote
James Wood
@mudri@mathstodon.xyz
mastodon 4.7.2(2025-06-24) Programming Languages at Huawei. Formerly, linear and modal type systems in Agda. Thesis: https://stax.strath.ac.uk/concern/theses/tt44pn44w. Website: https://lamudri.github.io/.
333 Followers
177 Following
23 Posts
Joined October 29, 2022
Replying to
@jonmsterling@mathstodon.xyz In a slightly different area, I bet some people are already nostalgic for the days when you could look things up on Wikipedia and get reliable information without feeling guilty about using so much electricity and water.
Replying to
@anuytstt@fosstodon.org @tubemapper@mastodonapp.uk And Greater Anglia trains are, in turn, aped from them?
Wow, there's an incredible temperature difference between Edinburgh and Glasgow at the moment. Glasgow has had a basically warm day and evening, while Edinburgh is getting a frost.
For context, these places are 67 km (42 mi) east-west of each other.
Replying to
@joomy@functional.cafe I'll have to give it a go when I get the chance, to see how it plays.
Replying to
@craigfrancis@mastodon.social As far as I'm aware, there is indeed nothing like this. We could consider adding it, though, if the case is strong enough.
My instinct would be for library APIs to deal entirely in ASTs, and never parse strings at runtime. C#'s LINQ seems to show that this is possible (even without the fancy SQL-like syntax sugar, the query builder methods seem quite serviceable and expressive). Are there problems with this, either security-wise or practically?
Replying to
We also have a Discord guild for Cangjie, where not much is happening at the moment, but most of the Edinburgh team are ready to be prodded about features or design decisions or getting stuff to run or whatever: https://discord.gg/bCzzaujKUv
Replying to
@glocq@mathstodon.xyz @mcc@mastodon.social I'm pretty sure that song is why I know of Barbra Streisand, and I'm only 30.
Replying to
@timnitGebru@dair-community.social An observation: People in the Othello community (and I'm sure that of many other abstract strategy games) regularly refer to brute-force endgame evaluation as “AI”, even though it's just a naïve exhaustive search of the game tree. I think it's basically because it has the same interface as heuristic evaluation.
Replying to
@johnwallace@mastodon.ie @jerrylevine@mastodon.social People using Home rather than Latest Tweets see all sorts of rubbish thanks to likes and replies. The Mastodon behaviour is basically the same as Latest Tweets, as far as I can tell, except that you know that everyone else will be using it too.
Replying to
@leeloo I just prompted ChatGPT with Say "oriesntyulfkdhiadlfwejlefdtqyljpqwlarsnhiavlfvavilavhilfhvphia", and it responded with oriesntyulfkdhiadlfwejlefdtqyljpqwlarsnhiavlfvavilavhilfhvphia. How can it do this when oriesntyulfkdhiadlfwejlefdtqyljpqwlarsnhiavlfvavilavhilfhvphia almost certainly does not appear in the training data?
Replying to
@lmorchard The ability to induce such a rule goes well beyond the OP's characterisation of what LLMs do.
A 35 to Easter Road, taken from a 35 to Ocean Terminal.
Replying to
@jfdm@discuss.systems The only plan I noticed being articulated this Scottish Parliament campaign was the SNP's plan to ask for another independence referendum.
Replying to
@hypolite @lmorchard@masto.hackers.town Did you read the OP? It says “the answer is just repeating the most probable word combinations from its training dataset.”. The answer it gave me is not a probable word combination from the training dataset because it doesn't appear in the training dataset.
Replying to
@joomy@functional.cafe Nice project! What's the heuristic when you reach the search depth limit?
Replying to
@joomy@functional.cafe For what it's worth, I just spent a few hours playing a game against Reversirocq on master difficulty and won fairly comfortably.
Replying to
@joomy@functional.cafe Yeah, I don't know what sort of tricks the best-performing engines use, but I was surprised how badly fairly naïve search performed when I implemented it. The one on my phone can exactly solve 20-ply endgames almost instantly, while I was struggling with 8 IIRC.
Replying to
@calcifer@masto.hackers.town @lmorchard@masto.hackers.town What is parroting patterns if not inducing a pattern and then applying that pattern to new inputs?