Lisp

I'm curious, when you started moving, when you moved over to industry, we're talking late 90s, probably early 2000s, were there technologies from that time that are completely like irrelevant now? And are there technologies that you use that you still use?

Because my boss had done his PhD in knowledge acquisition and was like, a hard AI nut, I use that word non-pejoratively, obviously the solution was to use Lisp.

So not a lot of Lisp being used these days.

Still a lot of

fans, I would think, of Lisp.

Go ahead.

Yeah, I like Lisp.

Not knocking it, but yeah, the use of it now is, and Python beat it, right? It was like, I think, a lot of things that Python was good at Sorry, a lot of things.

Algol's name is Algorithmic Language.

Lisp's is List Processing.

They were domain-specific languages that somehow got morphed into so-called general-purpose languages.

But you take your so-called general-purpose language and you try to do a thing that it's not particularly designed for, like, you know, massive concurrency or massive distribution or massive whatever, and it completely falls down.

And it's hard to kind of square those two things.

I think for beginners, obviously, like you mentioned, the superficiality in both senses is very important.

But it's also what kind of we love about Lisp, right? The fact that it's home iconic, for example, right? Although you have some thoughts.

Oh, you don't want to

And I think having some time into the Erlang community for quite a long time basically told us what to do.

So we started from the very old times where you were using Emacs and the very basic, very complicated piece of code in Lisp, implementing a lot of things, to starting to understand how can we go from there to something that improves the experience.

And this is like...

I think the type checker is a very good example.

There has to be a very strong technical reason to choose a different tool, I think, because anybody who's a good programmer should be able to write any language now And I would say broadly read the code.

Unless it's something completely, I mean, let's take Clojure or any kind of Lisp, but any of like, let's say the C derivatives, you know, like if you know how to read TypeScript, could you not, you really couldn't read C sharp at all? Like you would look at that and it would just be completely foreign to you.

I don't know.

That's my opinion on that.

But just if you don't read about go.

You don't have to have written Go, but you may not get an appreciation for that technique and why that was chosen there versus a language that does have pattern matching, a functional language, for example.

And then you could go to something like Clojure or Common Lisp and just have your brain totally broken by the way that it handles programming.

And when we talk about Dynamic types and static types, reading about practical object-oriented programming as a book, I think is great because it teaches you how to use dynamic types versus outright rejecting them.

Yeah, yeah, yeah.

And this was programming in Lisp and doing expert systems.

And that feels like it's a century ago, because nothing today in AI assembles anything like that.

There's a quantum jump, pun not intended, in AI that went from expert systems to back propagation and, you know, Alex.

That would have been really nice.

But they also did have some programming languages that are very tightly tied with AI like Lisp and Prolog.

So I don't know how many people actually have the urge to go fire up their web browser and write some Lisp in their browser.

But, you know, if you do...

There you go.

He treated it as a black box experiment, refusing to look at a single line of the generated C plus plus code.

We have been promised end user programming for forty years with things like Lisp and Smalltalk.

It never really materialized beyond Excel spreadsheets.

Now anyone can literally speak their custom software into existence.

And so you see these, these beautiful, uh, sort of, um, nexus things in, uh, in computer science and in all of science, where there's a small set of rules that are general and that cover everything.

You know, like, uh, uh, the, you know, uh- The f- the first Lisp interpreter was that way, or, uh, or the rules of electromagnetic propagation, um, where you've just got like four equations, and out of that comes all of radio and, uh, and you know, the whole spectrum.

So I th- um, that's the thing about s- so there was that small is beautiful and that, and that desire to, to distill things down to a small point, and that was something that I loved about the, uh, the challenge of working with the microcoded computers at Xerox, which is you had to do that.

Actually if you, if you had a good idea, if you could get it down small enough, it could go into this memory that suddenly ran five times faster.

I would just point out, if you look below the surface, there's less there than Smalltalk has.

Similarly, if you look at a language like JavaScript, which people malign, uh, but I don't, uh, JavaScript is a little bit of a hodgepodge.

But underneath, it has, uh, both a, a model that's, that I think of as being a bit more like Lisp, and a really solid, uh, set of foundations, including a really great garbage collector.

It has fabulous optimization mechanisms.

Um, which kinda gets into the site-

Summon, summon the code in our minds with Lisp, okay? Or-

Yeah, with, with Lisp

... have a scheme.

[chuckles] It's coming.

Hard.

But I am jealous of those who can just do pure keyboard stuff.

Next thing I know, you're gonna be like a Neovim user and wr- writing things in Lisp.