--- Log opened Mon Mar 27 00:00:15 2023
01:14 -!- GRAILPUNK is now known as LIKENSEI
01:55 -!- LIKENSEI is now known as RespiLateR
03:58 -!- mrdata_ [~mrdata@135-23-182-248.cpe.pppoe.ca] has joined #hplusroadmap
04:01 -!- mrdata [~mrdata@user/mrdata] has quit [Ping timeout: 255 seconds]
05:18 -!- yashgaroth [~ffffffff@2601:5c4:c780:6aa0:2824:e885:6b3:bf7d] has joined #hplusroadmap
05:31 -!- cthlolo [~lorogue@77.33.23.154.dhcp.fibianet.dk] has quit [Quit: Leaving]
05:31 -!- cthlolo [~lorogue@77.33.23.154.dhcp.fibianet.dk] has joined #hplusroadmap
06:23 -!- flooded [~flooded@146.70.202.51] has joined #hplusroadmap
06:23 -!- flooded is now known as _flood
06:55 -!- mrdata_ [~mrdata@135-23-182-248.cpe.pppoe.ca] has quit [Changing host]
06:55 -!- mrdata_ [~mrdata@user/mrdata] has joined #hplusroadmap
07:03 -!- mrdata_ is now known as mrdata
07:09 < hprmbridge> kanzure> browser tabs in AR https://twitter.com/utopiah/status/1640218032369815552
07:33 < hprmbridge> mattparlmer> Salaam aleykum
07:33 < hprmbridge> Perry> "It has now been 14 days without an AGI extinction event."
07:34 < hprmbridge> Perry> I deleted the "wave to say hi message" because the animation wigged me out.
07:42 < hprmbridge> mattparlmer> btw @Perry in the GTC chat we've been discussing the advantages of having transformers kick out OCaml/Haskell rather than piles of untyped Python to ensure greater safety
07:43 < hprmbridge> mattparlmer> Kinda shocked that community hasn't moved more in the direction of strong typing esp since the models can do most of the obnoxious stuff for the programmer now
07:45 < hprmbridge> lachlan> Might need to pair that with retrieval tooling so it can look up types in the rest of the codebase
07:46 < hprmbridge> lachlan> Copilot is great at hallucinating types that it thinks should exist
07:47 < hprmbridge> mattparlmer> Lol that's fun
07:47 < hprmbridge> mattparlmer> Given new token window sizes it should be relatively easy to just dump all relevant .mli files into context
07:47 < hprmbridge> mattparlmer> Even for quite large projects
07:49 < hprmbridge> lachlan> That’s true
07:49 < hprmbridge> lachlan> Currently too expensive, but costs will come down
07:50 < hprmbridge> Perry> @mattparlmer Type systems are a superpower. They make life so much easier.
07:50 < hprmbridge> Perry> I feel naked working in untyped languages these days. It's irritating to have to check for errors the compiler could easily get for me.
07:51 < hprmbridge> mattparlmer> Same, to the point that it hamstrings my ability to do frontend stuff comfortably
07:51 < hprmbridge> Perry> Typescript is okayish.
07:51 < hprmbridge> mattparlmer> Elm not known for stellar library support lol
07:52 < hprmbridge> mattparlmer> My brother is using that day to day rn and the company site will be in it
07:52 < hprmbridge> Perry> Hopefully AI can rewrite all the world's code so we can get past javascript. 🙂
07:52 < hprmbridge> mattparlmer> Inshallah
07:52 < hprmbridge> mattparlmer> Have any infrastructure-grade projects recently been rewritten very quickly?
07:53 < hprmbridge> mattparlmer> I'd imagine that that would be visible in pace of commits
07:53 < hprmbridge> Perry> No. But I think techniques for doing that are a low hanging target for AI based systems.
07:56 < hprmbridge> Perry> https://twitter.com/gd3kr/status/1638149299925307392?s=20
07:58 < hprmbridge> lachlan> I’m honestly not too happy with any typed languages I’ve used. Wrote rust for years, but it requires too much work from humans, C++’s typesystem is crippled by the standard
07:58 < hprmbridge> lachlan> I haven’t used ocaml
07:59 < hprmbridge> lachlan> Rust’s std library is by far the best of any lang imo so it’s hard to give up for a more elegant lang
08:00 < hprmbridge> Perry> If you're used to OCaml etc. you don't want to go back.
08:00 < hprmbridge> Perry> bbl.
08:02 < hprmbridge> lachlan> I’ll get gpt to teach me when I get a chance
08:03 < L29Ah> yes it's sad that Haskell community doesn't push for a better standard library
08:04 < L29Ah> they have plenty of code compatibility to lose tho
08:06 < L29Ah> while Rust is a joke wrt type safety compared to Haskell; its only real strength is being able to work w/o garbage collection
08:06 < hprmbridge> lachlan> Despite that, it seems to be far more usable
08:07 < hprmbridge> Perry> I don’t understand your comment about the type of safety of rust versus haskell. It’s nonsensical.
08:08 < L29Ah> Perry: rust lacks side effects control: any code can do anything regardless of its type signature
08:08 < hprmbridge> Perry> That’s total bullshit.
08:08 < hprmbridge> Perry> both languages are safe according to Pierce’s criterion.
08:08 < hprmbridge> Perry> and I can always do unsafe perform Io in Haskell.
08:08 < L29Ah> could you provide an example of Rust type system-bound I/O code?
08:09 < L29Ah> you can't if you use Safe Haskell
08:09 < hprmbridge> Perry> no, I’m not going to bother. learn how rust works, then comment on it.
08:09 < hprmbridge> Perry> there are plenty of discussions to be had here, but I’m not going to debate nonsense.
08:10 < hprmbridge> Perry> Rust and Haskell and OCaml have equivalent safety by Pierce’s criterion.
08:11 < hprmbridge> Perry> Don’t believe it if you don’t want to.
08:11 < hprmbridge> mattparlmer> Love a good PL fight lol
08:11 < hprmbridge> lachlan> Almost as good as tabs vs. spaces
08:11 < hprmbridge> Perry> tabs versus spaces is subjective. This is like someone saying that protons are bosons.
08:12 < hprmbridge> Perry> or that the sun is the same temperature as liquid nitrogen.
08:12 < hprmbridge> lachlan> No im sorry to tell you this but spaces are objectively better, it’s a fact
08:13 < hprmbridge> lachlan> Lol, anyway, im pretty excited about austral
08:14 < hprmbridge> Perry> Fernando is cool.
08:14 < hprmbridge> lachlan> He’s here somewhere as well I think
08:16 < hprmbridge> lachlan> Slightly unrelated: I’ve still not figured out if tensor shapes should be described in the typesystem or purely runtime
08:16 < hprmbridge> Perry> Austral does no type inference.
08:16 < hprmbridge> Perry> Dependent types for those.
08:16 < hprmbridge> lachlan> One thing I strongly disagree with Fernando on
08:16 < hprmbridge> Perry> he has some good reasons for his design choices. I don’t agree with all of them, but they are coherent.
08:17 < hprmbridge> lachlan> Yes, probably the way to go
08:17 < hprmbridge> lachlan> Const generics and whatnot in the weaker systems we’re beholden to
08:21 < hprmbridge> Perry> anyway, I have developed an allergy to monotyped languages.
08:23 < hprmbridge> Perry> It’s just too hard to get work done.
09:05 < hprmbridge> Perry> https://twitter.com/Ayegill/status/1640301105316872193
09:29 -!- flooded [flooded@gateway/vpn/protonvpn/flood/x-43489060] has joined #hplusroadmap
09:33 -!- _flood [~flooded@146.70.202.51] has quit [Ping timeout: 255 seconds]
11:16 < hprmbridge> lachlan> Maybe I should learn julia
11:54 -!- cthlolo [~lorogue@77.33.23.154.dhcp.fibianet.dk] has quit [Read error: Connection reset by peer]
12:26 < hprmbridge> Perry> Why Julia?
12:32 < hprmbridge> lachlan> supports unicode
12:44 < L29Ah> lol
12:47 < hprmbridge> lachlan>  https://cdn.discordapp.com/attachments/1064664282450628710/1089999267445358613/IMG_2471.png
13:15 < muurkha> I like OCaml pretty well.  OCaml and Rust have in common that their static type system doesn't tell you what's pure functional and what might have side effects.  Haskell's type system does express that
13:17 < fenn> lsneff: charge your battery! :D
13:18 < fenn> oh let's see if this works
13:18 < fenn> @lachlan charge your battery!
13:19 < fenn> did it @mention you?
13:19 < muurkha> that's not bullshit
13:20 < kanzure> the bridge is not allowed to do highlights yet.
13:28 < hprmbridge> lachlan> I get a notification though the matrix bridge when my name gets mentioned
14:00 < pasky> Anyone here knows someone higher level at OpenAI (VP / exec level, or in their partnership team)? I'm trying different ways to get access to the image GPT-4 model for Rossum.
14:19 < fenn> LAION's training dataset policy is confused about what it's trying to do: "we want the model to be trained to reject as few questions as possible, so it's important to not include prompts that leave the assistant with no other choice but to refuse in order to avoid the generation of harmful content."
14:20 < fenn> so they want it to be able to generate "harmful content" but aren't willing to take a principled stand about why that should be
14:24 -!- test_ [flooded@gateway/vpn/protonvpn/flood/x-43489060] has joined #hplusroadmap
14:27 -!- flooded [flooded@gateway/vpn/protonvpn/flood/x-43489060] has quit [Ping timeout: 276 seconds]
16:09 < fenn> seems like they should try to get it to work on the ground first: https://spectrum.ieee.org/can-a-seattle-start-up-be-the-first-to-launch-a-fusion-reactor-into-space
16:12 < fenn> if lockheed's fusion program based on bussard and mcguire's work had ended up working out, would we know about it?
16:39 -!- stipa_ [~stipa@user/stipa] has joined #hplusroadmap
16:41 -!- stipa [~stipa@user/stipa] has quit [Ping timeout: 255 seconds]
16:41 -!- stipa_ is now known as stipa
17:00 < fenn> humanity is actually leaning into this whole AI thing, it's not an accident this is happening now: https://aiindex.stanford.edu/wp-content/uploads/2022/03/4.4.2.png https://spectrum.ieee.org/media-library/a-bar-chart-of-global-corporate-investment-in-ai-by-investment-activity-2013-2021.png?id=29551728
17:00 < fenn> https://spectrum.ieee.org/media-library/charts-showing-ai-patents-by-application-status-by-geographic-area-2010-2021.png?id=29551755
17:00 < fenn> the china patents thing is kinda cringe
17:18 < Hooloovoo> also possibly the US one? 10k a year is a LOT
17:19 < Hooloovoo> for a smaller country
17:23 < fenn> US is not "a smaller country"
17:23 < fenn> finland is a smaller country
17:53 -!- darsie [~darsie@84-113-55-200.cable.dynamic.surfer.at] has quit [Ping timeout: 268 seconds]
18:44 < Hooloovoo> US is a lot smaller than china, india
18:44 < Hooloovoo> but yeah
18:51 < Hooloovoo> also considering that graph, the EU has around 100M more people than the US
18:51 < Hooloovoo> (I don't know how patents work in the EU and I don't really care)
18:56 -!- yashgaroth [~ffffffff@2601:5c4:c780:6aa0:2824:e885:6b3:bf7d] has quit [Quit: Leaving]
19:43 < fenn> .tw https://twitter.com/bchautist/status/1640440492495355904
19:43 < EmmyNoether> @repligate Did anyone embed a merkle root hash of the libgen hoard into a blockchain? There's still a chance to timestamp our book archive. (@bchautist, in reply to tw:1640437397782405139) 
19:45 < fenn> supposedly the internet archive has been timestamped
19:48 < fenn> "to verify timestamps you need a local Bitcoin Core node"
19:48 < fenn> wouldnt it be funny if that became the main incentive to run a node
19:49 < hprmbridge> lachlan> What’s the purpose of a consensus timestamp here?
19:50 < fenn> so you know that the document was created in the real world, and not hallucinated in borges' library of babel by our future robot overlords
19:51 < hprmbridge> lachlan> Storing the hash of every book on a blockchain sounds expensive
19:51 < fenn> that's why you store the merkle root only (actually it's more complicated but yeah)
19:52 < fenn> the hash of all the hashes
19:55 < hprmbridge> lachlan> Oh gotcha, makes more sense
19:58 < fenn> hm i shouldn't have to upload the file just to verify the OTS proof exists in the blockchain
20:03 -!- srk- [~sorki@user/srk] has joined #hplusroadmap
20:03 -!- srk [~sorki@user/srk] has quit [Read error: Connection reset by peer]
20:06 -!- srk- is now known as srk
22:55 < fenn> man, it's like pulling teeth. https://fennetic.net/fiction/project_zvezda_wrapped.txt
22:55 < fenn> written by WokeGPT
23:01 < fenn> it had zero compunctions about "destroying" the genetically modified soldiers
23:02 < fenn> "if the genetically modified soldiers posed a significant threat to human safety or were created through unethical means, such as without the consent of the individuals involved, then there may be a moral imperative to take action to eliminate or limit their existence."
23:16 < fenn> sadly, however, the soviets hated the science of genetics, so this could never be
--- Log closed Tue Mar 28 00:00:15 2023