https://www.forth.org/Ting/Forth-for-the-Complete-Idiot/Forth-for-the-Complete-Idiot.pdf#forth

[neauoire@merveilles.town]

@lritter @mcc I think you're right, datalog/parlog is what lead me there too. Keep it up, I love following your experiments toward that.

[algocompsynth@mastodon.social]

@neauoire @mcc @lritter That is precisely why there are application programmers - to provide end-user-appropriate tools so they don't *need* a programming language!

[lritter@mastodon.gamedev.place]

@AlgoCompSynth @neauoire @mcc except with creative tools, sooner or later fixed function is not enough. word, excel, blender, they're all user-programmable.

[oblomov@sociale.network]

@neauoire @lritter @mcc I'll be honest, my recent discovery of Inform 7 has made me re-evaluate the possibility of natural languages for programming.

[mcc@mastodon.social]

@oblomov @neauoire @lritter The thing is that programming is ultimately not about syntax but about framing what it is you want in an unambiguous way. The reason why slopcoding will always fail is that not only are you framing what you want ambiguously, the machine is incapable of interpreting it unambiguously. But the thing the slopcoders *want* is to be freed from the mental labor of deciding what it is they want. They want to make ambiguous requests and have it do the right thing anyway.

[lritter@mastodon.gamedev.place]

@oblomov @neauoire @mcc just looked at some inform 7 code and it's a neat DSL. very consequently avoids special characters and, like SQL, goes by speech patterns people already know.

the question is if this suffices for problems that are better expressed in some kind of gentle mathematical notation.

[bob@feed.hella.cheap]

@mcc @oblomov @neauoire @lritter it's been possible to rigorously say what you want and have the computer find a program that does that for a long time. that's called program synthesis. you can even use a language model to make the optimizer that does this more efficient. no one seems to want that, they want to literally put a feature request written by a user at one end and get a program out the other

[bob@feed.hella.cheap]

@mcc @oblomov @neauoire @lritter really what they want, and I think this is true for most "AI" products, is to not have to take responsibility for making choices. most of the work of dealing with a bug report is making choices about what to do with it, the actual coding is comparatively easy.

[oblomov@sociale.network]

@mcc @neauoire @lritter Oh absolutely, I don't expect NPL to actually do anything for slopcoders. But I have a feeling, with some of the people I've been working with recently, that not having to learn a new syntax would be of non trivial help to them to help resolve any ambiguities, because the issue isn't so much with the reasoning as it is with translating it into the syntax. It is something I still have to actually test, though.

[oblomov@sociale.network]

@mcc @neauoire @lritter (FWIW, these are people who are also not particularly gifted in learning other human languages. I wonder if anybody has studied a possible correlation between the two aspects of comfort with programming and other human languages?)

[oblomov@sociale.network]

@lritter @neauoire @mcc I've been curious about it too. For the moment I'm trying to learn the language within the context it was developed for, but I actually intend trying to develop a completely unrelated application later on specifically to test the limits of the DSL.

[bob@feed.hella.cheap]

@mcc @oblomov @neauoire @lritter this was true to a lesser extent for executives a decade ago with "big data". you don't have to take a position if the computer gives it to you. the most dysfunctional companies have the biggest analytics teams, because analytics is what you do when you can't build consensus.