Posts by ncweaver@thecooltable.wtf
 (DIR) Post #AYo69YbWkOg9GkOsme by ncweaver@thecooltable.wtf
       2023-08-17T01:58:39Z
       
       0 likes, 0 repeats
       
       @tomjennings @ai6yr Even though it is marginally bigger, I'm using USB-C on my board designs rather than Micro-B...  Micro-B really does need to DiaF.Oh, and a good reason to upgrade to Pi-4s: the power is USB-C.
       
 (DIR) Post #AZ0Z13suJ7SaFELlaq by ncweaver@thecooltable.wtf
       2023-08-23T02:06:38Z
       
       0 likes, 0 repeats
       
       @simon Or cortana
       
 (DIR) Post #AZGCc2p028AFFIXyCm by ncweaver@thecooltable.wtf
       2023-08-30T15:06:34Z
       
       1 likes, 0 repeats
       
       @bluerootsradio @e_urq This is so not surprising me.  These LLMs are stochastic parrot powered bullshit machines, and due to their nature have no notion of "correct" or "incorrect"As such they are only useful when generating bullshit is useful or the output is human edited by experts who can detect & remove bullshit.The former is spammers and scammers.  The latter is "are we sure we should not just have the expert write it in the first place?"
       
 (DIR) Post #AaHHzheCPf83qg4E8O by ncweaver@thecooltable.wtf
       2023-09-30T01:43:36Z
       
       1 likes, 0 repeats
       
       @GossiTheDog @briankrebs @malwaretech Well, lets face it, these instances of spicy autocomplete are all about creating text that sounds plausible.  So it really does sound like you should be suing John Doe for hacking...
       
 (DIR) Post #AaIX4ZCmv9XHaHz0ls by ncweaver@thecooltable.wtf
       2023-09-30T16:06:51Z
       
       0 likes, 0 repeats
       
       @simon They are, however, of remarkably limited utility.  For programming they are useful, because most programming is actually recycling boiler plate, so having a compressed view of all programs and auto-completing off of that is hugely useful.For translation they are useful, as it is a compressed representation of the two languages.Otherwise, they really are bullshit generators: spewing out an uncompressed representation of the lossy compressed data starting at a specific start.
       
 (DIR) Post #AaIX4bonCp7lgjmOHo by ncweaver@thecooltable.wtf
       2023-09-30T16:08:09Z
       
       0 likes, 0 repeats
       
       @simon And you can really tell.  Copilot has been in the evolution for years, as is Google Translate.The new hype was driven by the eliza effect on the bullshit generators.  I think a good dividing line is "everything after ChatGPT went public" is in the bullshit category unless proven otherwise.
       
 (DIR) Post #AaIX4dtTUO3e7pz9rU by ncweaver@thecooltable.wtf
       2023-09-30T16:10:57Z
       
       0 likes, 0 repeats
       
       @simon Additionally, on their own, LLMs are bullshit generators in the technical term, they have no sense of correctness.But for the co-pilot programming role, you have a post-bullshit correctness filter in the compiler that explicitly removes a huge amount of the problems.
       
 (DIR) Post #AaIYhb8RmM1Som7zfM by ncweaver@thecooltable.wtf
       2023-09-30T16:24:57Z
       
       0 likes, 0 repeats
       
       @simon Also you are using LLMs in a tight human-and-checking computer coupled feedback loop.That ends up explicitly attacking the error/bullshit problem on the lossy decompression that is an LLM's execution of spicy autocomplete.But this also means that LLMs are a remarkably limited and weak tool, and provably can't go beyond that without some notion of "truth" being added on outside of the LLM itself.
       
 (DIR) Post #AaKfheSiUtw58DV9No by ncweaver@thecooltable.wtf
       2023-09-30T17:45:56Z
       
       0 likes, 0 repeats
       
       @simon @jernej__s @axleyjc @interpipes @briankrebs But that is the same reason you worry about copyright: if you can get it to spit out trade secret it can spit out copyrighted data, so the output is derived from the copyrighted data.
       
 (DIR) Post #AabCIii4KfIORdssFc by ncweaver@thecooltable.wtf
       2023-04-30T00:12:34Z
       
       0 likes, 1 repeats
       
       @scalzi
       
 (DIR) Post #AabCPBsCoefgdCi8Dw by ncweaver@thecooltable.wtf
       2023-09-01T01:55:04Z
       
       0 likes, 1 repeats
       
       The cryptocurrency space has always been a tire fire, but this is taking it a bit TOO literallyhttps://www.theguardian.com/us-news/2023/aug/31/bitcoin-mining-plan-pennsylvania-tire-burning
       
 (DIR) Post #AabCPLAcFZkHSuOSfY by ncweaver@thecooltable.wtf
       2023-09-30T00:08:25Z
       
       0 likes, 0 repeats
       
       @ai6yr Its a DC3.  "The only replacement for a DC3 is another DC3"Those planes are almost nigh-indestructible.
       
 (DIR) Post #AabCPQ1wBGm4XRN6Ku by ncweaver@thecooltable.wtf
       2023-09-30T01:26:35Z
       
       0 likes, 1 repeats
       
       @douglasvb @davidnewman @ai6yr Plus, now we have the State Bird of California:the VLAT
       
 (DIR) Post #AbGTrznqHRTIryEAXg by ncweaver@thecooltable.wtf
       2023-10-29T14:20:29Z
       
       0 likes, 0 repeats
       
       @benjaminwittes Yes it is.Unfortunately I'm on the wrong coast.
       
 (DIR) Post #AcVRC7K4NpUw93w5bs by ncweaver@thecooltable.wtf
       2023-12-05T17:21:57Z
       
       0 likes, 0 repeats
       
       @clive Probably the best truck for most people on sale today is the Ford Maverick.  A short-cab/long bed variant would be the ultimate work truck for so many users, too.As a bonus, it doesn't break the bank.
       
 (DIR) Post #Aco9SLR2T6mxyXGPFg by ncweaver@thecooltable.wtf
       2023-12-14T18:02:01Z
       
       0 likes, 0 repeats
       
       @simon One addition, it does NOT say your data will not be used to train Dropbox's customized model, just that it won't be used to train OpenAI's model.The NSF sent out a warning that reviewers MUST not allow proposals to be used as training data for a LLM.
       
 (DIR) Post #Ad6auf1iHYrUuhe16O by ncweaver@thecooltable.wtf
       2023-12-23T15:34:19Z
       
       0 likes, 0 repeats
       
       @mjg59 That the ocean isn't a giant block of Ice IX
       
 (DIR) Post #AdL5yLiYMeb96ub1JA by ncweaver@thecooltable.wtf
       2023-12-30T15:26:20Z
       
       0 likes, 0 repeats
       
       @mjg59 @cstross Plus there is a too big tendency to say "That asshole-bomb did good work" as if that somehow forgives the damage they've done.  You can do the good work without the damage.
       
 (DIR) Post #Ai5RprHF6O4gUoZzHc by ncweaver@thecooltable.wtf
       2024-01-29T19:37:06Z
       
       0 likes, 0 repeats
       
       Today I learned:import java.security.MessageDigestval String.sha256: Stringget() {   val bytes = MessageDigest.getInstance("SHA256").digest(this.toByteArray())   return bytes.joinToString("") { "%02x".format(it) }   }is valid Kotlin that now gives every string an on-demand sha256 'field'
       
 (DIR) Post #Ai68DlltQ4kMRg04PY by ncweaver@thecooltable.wtf
       2024-02-12T18:42:54Z
       
       0 likes, 0 repeats
       
       Compare and contrast: The same problem (Dijkstra's traversal on a graph) in Kotlin and actually "safe" C++.The amount of tedious shit C++ needs to not be a memory unsafe nightmare is "why bother?"https://github.com/nweaver/ucsc_teaching_demo/blob/main/kotlin/src/main/kotlin/Graph.kthttps://github.com/nweaver/ucsc_teaching_demo/blob/main/C%2B%2B/Graph.h