libera/#lisp - IRC Chatlog
Search
10:04:18
gnoo
both are valid, and i think immediate is used? i remember seeing it when i tried to read SICP
10:08:24
x8dcc
I was reading it earlier, that's why I asked. I think this is mentioned in the metacircular evaluator part, but I was not sure if I understood it correctly
14:07:18
keeliro
Hello. Is it true that LLM can program? On which tasks do you think that LLM can effectively replace humans?
14:27:22
mrcom_
keeliro: I think LLMs have two particular features: (a) they find correlations, (b) they string correlations together. They're terrible at determining causation.
14:29:17
mrcom_
You can think of them as surveys-in-a-box. If you had a group of 1000 people, and asked them to vote on "what's the next word", over and over again, you'd have a LLM.
14:30:26
wasamasa
I dunno, it just strikes me as a hilariously inefficient way of asking a computer questions
14:31:22
wasamasa
but given the skill level of your average CS student, they could be an improvement I guess
14:33:54
keeliro
How would I convince a business person that they need a programmer and not someone to manage a LLM?
14:34:07
mrcom_
keeliro: Why does our computer keep dying at 20:00 every evening? Hmm. That's strongly correlated with the cleaning crew being on-site.
14:35:41
mrcom_
What does the customer want? Not what tech, what pushes their buttons, what makes them feel good?
14:45:15
jackdaniel
(I find current genai interesting, but marketing and shoving it onto users is revulsing)
14:52:41
mrcom
Once I realized that LLMs work by finding correlations, then changing internal modeling to incorporate the correlations (compressing the model), I went cold on the idea.
14:53:57
mrcom
I don't understand what "conscious awareness" means, no one does. But I'm 1000% convinced we aren't qualified to create a thinking entity. It will get very, very messed up.
14:58:41
mrcom
To end this meander on a cheerful note, I don't think current situations will last, things will change for the vastly better, but that's getting *really* off-topic.
15:11:29
keeliro
Why does Lisp not solve the World' problems. Isn't that what computers are supposed to do (repeat something or test or be precise, etc)?
15:18:29
jackdaniel
computers meant to be vehicles of thought, so instead of 'solving things' they were designed to amplify capabilities
15:19:38
jackdaniel
but that's not necessarily true in the current economy and the societal mindset - indeed many apps "solve problems"
15:22:43
keeliro
The main which think I get from this discussion is that Lisp is a language which can be easily redifined to the core. And that characteristic makes it able to solve unique complex situations which do not necesarily are correlations solvable by LLMs.
15:26:23
keeliro
What do you think about my conclusions? I know they might not be perfect. But I am slowly making some sense.
15:29:57
bjorkintosh
a handful of people realized computers were tools and have been using them as such since the very beginning.
15:30:24
bjorkintosh
these people do _not_ fetishize computers or languages. they simply use the goddamned thing and move on.
15:36:51
keeliro
I would like to make my Emacs take a youtube link and transform it to an invidious link and reproduce it with mpv without yt-dlp...I do have something of the sort. But it is not exactly what I described. Do you want to see it.
15:40:30
keeliro
like when you know it is you when looking at the mirror and not someone on the other side of the glass
15:40:58
x8dcc
Same nonsense, no offense. Explain why this isn't self-awareness and LMMs are: printf("I am a program");
15:42:42
keeliro
I don't even care about self-awareness. Maybe people care because they believe the fiction about machines controlling humans.
15:43:14
keeliro
They already do. But it is human decisions to control other humans by way of web pages.
17:38:14
jackdaniel
I"m not interested, but honestly you"d have a better chance if you had posted the thing
18:07:31
keeliro
so, if I am on a web page on invidious, it reproduces the video url with mpv without the need of yt-dlp
18:23:48
tux0r
my own alternative to yt-dlp - yaydl (code.rosaelefanten.org/yaydl) - redirects all youtube urls to invidious
19:30:18
keeliro
I don't want to run any code which has not been verified by someone which I trust...and it serves as a Lisp exercise too
19:35:32
tux0r
i initially thought i'd write yaydl in lisp too, but i wanted 1) to learn rust and 2) an easy way to add "modules", and rust's "inventory" is just great for that
19:43:04
wasamasa
keeliro: I have my doubts you have the drive to independently figure out youtube's many mechanisms to make playback of the video a major pain
19:44:20
wasamasa
keeliro: yes, lisp is nice to use and all that, but there is little point in rewriting everything in it
19:51:27
keeliro
but still, I would like something more in the lines of what Hyperbola GNU does. I know most people would not opt in. But it is my favorite choice.
19:56:54
tux0r
well, webdrivers were the easiest solution for the problem of some video sites loading their stuff via JS
20:28:30
tux0r
"i hope you gain contributors" is an evil thing to say, it roughly translates to "well I will certainly NOT help you lmao" :p
20:30:01
aeth
having to fake the heavyweight Chrome to maybe be able to get the content of the website (but 50/50 chance Cloudflare blocks you, anyway) just shows how bad the web has gotten
20:30:19
aeth
99% of these sites are mostly text and could've been done in 2008 with minimal/no javascript
20:33:48
keeliro
I just would not like to have anything to do with Rust...I did not mean to be evil to you.
20:35:04
tux0r
as far as i can see, the chance to avoid rust is almost over. i too wish lisp would have a similar impact on IT, but it doesn't.
21:03:47
keeliro
well...communication is a 2 way operation...both parties are wrong when communication does not happen, not one
21:04:04
x8dcc
In languages like Scheme, why is "begin" a special form? Why can't the interpreter evaluate its arguments before applying the procedure, and then just return the last expression "begin" received? Let me know if it's not clear what I mean
21:04:47
tux0r
i'm positive that, when one side makes a suggestion, the other side explains why that suggestion is not helpful and the one side soubles down on the suggestion, it's not a "both parties are wrong"
21:05:18
tux0r
but, as this discussion is moot by this point, i killfile'd you. just to have a calm evening. have a nice evening too.
21:08:21
tux0r
the easiest answer for "why does scheme do x?" is "because the standard says so"... ;-)
21:09:48
x8dcc
I am sure my logic is not right. I am guessing there is a reason why (at least SICP) mentions it as a special form
21:10:29
x8dcc
keeliro: Jesus, someone was rude on the internet, what a big deal. Just stop spamming, please.
21:12:25
tux0r
i would even deny that i was "rude", as all i did was explain why the suggestion won't work. i even said please. anyway, just add keeliro to your ignore list too, x8dcc.
21:18:27
x8dcc
Alfr: Funny, you say that, because my `lambda' doesn't support multiple expressions in the body :p
21:19:13
x8dcc
And I wanted to add precisely that. So I saw that `begin' was for that (independently of lambdas), and I was wondering why it's a special form
21:19:50
x8dcc
lambda, define, quote, cond... They all have some special rules regarding evaluation, but I don't see that with "begin"
21:30:42
x8dcc
Alfr: Thanks for the link. I still don't really see why it can't have normal evaluation rules, though :/