Tangentially related, this post git me reading about the spectre language. In the spectre docs you write: "There is a notable lack of contract based programming languages that enforce correctness at a low level.". What is the difference between a function checking arguments and throwing exceptions and contract-based functions, apart from perhaps a more concise syntax to specify it?
"Just" is exactly made for this, and it is amazing. You write a justfile that is somewhat similar to a makefile but without the painpoints and it provides a CLI interface of commands you want to run
This has not been true for at least a decade for dutch buildings: there are strict regulations requiring decent insulation for new buildings. Renovating and insulating old buildings is also encouraged, but not required by law.
I made something similar once, specifically targetted for guitar tablature https://tabviewer.app/
To make links shorter for sharing with others, I use a shortlink service. Pasting URLs of thousands of characters long can be problematic
I agree. It would also work out like a long term supervised learning process though. Humans showing how it's really done, and AI companies taking that as a gold standard for training and development of AI.
I'm surprised Pandoc markdown is not mentioned. You can make that semi structured quite easily, and write your own transformations using lua. It's powerful enough to write math papers and export into both pdf and html.
Regarding typing latex vs unicode, I use WinCompose/XCompose with a list of bindings that include most latex symbols. So instead of \cup I'd type <compose>cup
I use Unicode to type math, which is the closest you can get in plain text to what you see in the rendered output. The latex package unicodemath is amazing. As a bonus you can paste the code in chat applications when communicating with peers.
In Haskell with `head (sort list)` the entire list does not have to be sorted, depending on the sort implementation. Everything is lazy, so sort can sort the list just enough to return the smallest element.
Isn't that very wasteful or difficult to do in practice? If you consider that shrinkers generally take lower numbers to be 'simpler' than higher numbers, complexity-ordering requires you to generate all the numbers from low to high
This post has many upvotes, but all the comments ask questions about the usefulness of this, without any justifying response so far. I have the same question, and I wonder what's going on with this post?
No, it's about the distribution being injective, not a single sampled response. So you need a lot of outputs of the same prompt, and know the LLM, and then you should in theory be able to reconstruct the original prompt.
This is a cool example of how specializing a generic algorithm to a specific subspace can yield much better results. This is quite often the case in my experience, but we often don't bother utilizing properties that are specific to our problem space, and just apply the generic algorithm out of convenience (and because it is often good enough)
No? Intelligence is not telepathy. You can't just assume that your initial interpretation is the correct one, even more so if you interpretation doesn't make sense. Like in this case, where the parents interpretation of "abstract" is literally not applicable to this configuration language.