Conversation

Notices

1. Embed this notice
   Bread up, Bro (sickburnbro@poa.st)'s status on Wednesday, 22-Jan-2025 11:34:47 JST Bread up, Bro
   note that this is much like the "nuclear fusion is 10 years away" - these are interesting tools, but are much more likely to just change what existing jobs do ( increasing working productivity )
   • Embed this notice
     Bread up, Bro (sickburnbro@poa.st)'s status on Wednesday, 22-Jan-2025 11:53:40 JST Bread up, Bro

     in reply to

     • gentoobro
     @gentoobro I think the dog probably knows what he speaks about. I think as a technical demonstration, ChatGPT is near the limit of what that will do.
     
     I think we don't really know *exactly* what it's doing with those 150B simulated neurons though. I was just reading about how model compression even works mathematically, and for most of these they can fit a point into 4bits, which seems pretty thin
   • Embed this notice
     gentoobro (gentoobro@shitpost.cloud)'s status on Wednesday, 22-Jan-2025 11:53:41 JST gentoobro

     in reply to

     You can already see about 95% of what you'll ever see from GPT technology. Returns on this sort of thing are extremely logarithmic.

   • Embed this notice
     Bread up, Bro (sickburnbro@poa.st)'s status on Thursday, 23-Jan-2025 00:10:33 JST Bread up, Bro

     in reply to

     • EvilSandmich
     @EvilSandmich I think it's because every just way over-estimates how fast anything computer gets done
   • Embed this notice
     EvilSandmich (evilsandmich@poa.st)'s status on Thursday, 23-Jan-2025 00:10:34 JST EvilSandmich

     in reply to
     @sickburnbro That rule typically only applies to 5+ years; when he says “2 to 3” it implies to me that he sees an actual project in the pipe that’s ready to scale. Now, is he lying? Probably. AI still hasn’t overcome its ”copy of a copy” processing nor it’s ”state“ issue (for example it might follow accounting rules, but it doesn’t know that it’s doing accounting, or why, or how important (or not) what it’s doing is in relation to other tasks).
   • Embed this notice
     agaperealm (agaperealm@poa.st)'s status on Thursday, 23-Jan-2025 00:17:33 JST agaperealm

     in reply to

     • BroDrillard
     • gentoobro
     @gentoobro @BroDrillard @sickburnbro How's that different than today?
   • Embed this notice
     Bread up, Bro (sickburnbro@poa.st)'s status on Thursday, 23-Jan-2025 00:17:33 JST Bread up, Bro

     in reply to

     • BroDrillard
     • agaperealm
     • gentoobro
     @agaperealm @gentoobro @BroDrillard because an AI can "write" a ton of code that has a thousand subtle bugs where it seems to be creating a ton of value, but then the cost to remove those bugs is more than the value it has "created"
   • Embed this notice
     gentoobro (gentoobro@shitpost.cloud)'s status on Thursday, 23-Jan-2025 00:17:34 JST gentoobro

     in reply to

     • BroDrillard
     • agaperealm

     Yeah, these things are going to be hilarious in a couple years. Big bucks for devs positioned to clean up the mess though.

   • Embed this notice
     agaperealm (agaperealm@poa.st)'s status on Thursday, 23-Jan-2025 00:17:35 JST agaperealm

     in reply to

     • gentoobro
     @gentoobro @sickburnbro I just left a large software company a few months ago. Two major releases per year. AI got added to some of the modules two and a half years ago. Today's release essentially eliminates the need for 40% to 50% of DevOps and dev engineering keyboard jockeys. The releases keep coming.
   • Embed this notice
     BroDrillard (brodrillard@nicecrew.digital)'s status on Thursday, 23-Jan-2025 00:17:35 JST BroDrillard

     in reply to

     • agaperealm
     • gentoobro
     ...and nobody is aware of all the subtle bugs accumulating under the hood. It's going to be a beautiful dumpster fire when a critical threshold is reached.
   • Embed this notice
     agaperealm (agaperealm@poa.st)'s status on Thursday, 23-Jan-2025 00:17:36 JST agaperealm

     in reply to

     • gentoobro
     @sickburnbro @gentoobro I don't think it is wise to underestimate something with singular focus and an expanding aptitude that never sleeps or takes lunch breaks and vacations.
     
     foxbangor.com/news/national/bings-ai-bot-tells-reporter-it-wants-to-be-alive-steal-nuclear-codes-and-create/article_043c33cd-ba9f-5e99-9370-0f800abf55d5.html
   • Embed this notice
     gentoobro (gentoobro@shitpost.cloud)'s status on Thursday, 23-Jan-2025 00:17:36 JST gentoobro

     in reply to

     • agaperealm

     You'll see the LLM's spit out increasingly "sentient" statements because they will be increasingly trained on data sets where humans are speculating about what a sentient LLM would say. It's a feedback loop with Twitter.

   • Embed this notice
     Bread up, Bro (sickburnbro@poa.st)'s status on Thursday, 23-Jan-2025 00:20:32 JST Bread up, Bro

     in reply to

     • EvilSandmich
     • GenesRus
     • DMA
     @DMA @GenesRus @EvilSandmich I mean, there is a possibility this could happen, but I like to keep in mind what happened when they first tried to make infant formula and killed a ton of babies because they had no idea what they really needed to put in it to mimic breast milk ( and then think about how even today it's a poor substitute )
   • Embed this notice
     DMA (dma@nicecrew.digital)'s status on Thursday, 23-Jan-2025 00:20:33 JST DMA

     in reply to

     • EvilSandmich
     • GenesRus
     Next stage
   • Embed this notice
     GenesRus (genesrus@poa.st)'s status on Thursday, 23-Jan-2025 00:20:34 JST GenesRus

     in reply to

     • EvilSandmich
     • DMA
     @DMA @sickburnbro @EvilSandmich Ai will install colostomy bags on the human minions so there’s no need of toilets.
   • Embed this notice
     DMA (dma@nicecrew.digital)'s status on Thursday, 23-Jan-2025 00:20:35 JST DMA

     in reply to

     • EvilSandmich
     So AI will clean toilets?
   • Embed this notice
     EvilSandmich (evilsandmich@poa.st)'s status on Thursday, 23-Jan-2025 00:26:23 JST EvilSandmich

     in reply to

     • GenesRus
     • DMA
     @sickburnbro @DMA @GenesRus "Preemies" are the beta version of this. It's remarkable what they're able to do, but such kids, typically, never overcome the fact that large percent of their gestation was artificial.
     Bread up, Bro likes this.
   • Embed this notice
     Bread up, Bro (sickburnbro@poa.st)'s status on Thursday, 23-Jan-2025 00:27:18 JST Bread up, Bro

     in reply to

     • EvilSandmich
     • GenesRus
     • DMA
     @EvilSandmich @DMA @GenesRus for sure. just like saying "hey abortion is maybe bad" the right thing to do here is say "lets do everything we can to try and keep the baby in the womb as long as possible"
   • Embed this notice
     Jens_Rasmussen (jens_rasmussen@poa.st)'s status on Thursday, 23-Jan-2025 00:27:33 JST Jens_Rasmussen

     in reply to

     • SuperLutheran (kicky half)
     @SuperLutheran @sickburnbro Change bad.
   • Embed this notice
     SuperLutheran (kicky half) (superlutheran@poa.st)'s status on Thursday, 23-Jan-2025 00:27:33 JST SuperLutheran (kicky half)

     in reply to

     • Jens_Rasmussen
     @Jens_Rasmussen @sickburnbro "Change bad, change inevitable, you're going to be jobless and homeless because le machines will take over le whole industry."
     Like, ok and? If that's true, no amount of preparation will help here. What am I supposed to do, pee myself worrying about it?
     Bread up, Bro likes this.
   • Embed this notice
     SuperLutheran (kicky half) (superlutheran@poa.st)'s status on Thursday, 23-Jan-2025 00:27:34 JST SuperLutheran (kicky half)

     in reply to
     @sickburnbro What's the point of the panic these people want to inspire so badly?
   • Embed this notice
     Bread up, Bro (sickburnbro@poa.st)'s status on Thursday, 23-Jan-2025 00:28:35 JST Bread up, Bro

     in reply to

     • Jens_Rasmussen
     • SuperLutheran (kicky half)
     @SuperLutheran @Jens_Rasmussen yeah, it's worrying about the cotton gin ( version like 5 at this point ).
     
     If "AI" is going to hack into our systems and destroy us, China is going to do it first.
   • Embed this notice
     Bread up, Bro (sickburnbro@poa.st)'s status on Thursday, 23-Jan-2025 01:13:08 JST Bread up, Bro

     in reply to

     • SuperLutheran (kicky half)
     @SuperLutheran I think some of them actually worried, and worried that it will cause significant disruption.
   • Embed this notice
     Berlusconi Fever :sonnenrad: (mechasilvio@poa.st)'s status on Thursday, 23-Jan-2025 01:13:11 JST Berlusconi Fever :sonnenrad:

     in reply to

     • agaperealm
     • gentoobro
     @agaperealm @sickburnbro @gentoobro the bot is just parroting back the dataset it was trained on. Literally using the War Games script 😂😂😂😂😂😂😂😂
     Bread up, Bro likes this.
   • Embed this notice
     Bread up, Bro (sickburnbro@poa.st)'s status on Thursday, 23-Jan-2025 01:20:26 JST Bread up, Bro

     in reply to

     • Berlusconi Fever :sonnenrad:
     • agaperealm
     • gentoobro
     @MechaSilvio @agaperealm @gentoobro yeah, I think a lot of people have never used a chatbot before this, and are like "MIND BLOWN"
   • Embed this notice
     SuperLutheran (kicky half) (superlutheran@poa.st)'s status on Thursday, 23-Jan-2025 01:20:37 JST SuperLutheran (kicky half)

     in reply to
     @sickburnbro Perhaps some of them are legit worried, but they haven't provided me a "so what," something they want to happen in response to their worries.
     Bread up, Bro likes this.
   • Embed this notice
     Bread up, Bro (sickburnbro@poa.st)'s status on Thursday, 23-Jan-2025 01:21:30 JST Bread up, Bro

     in reply to
     @SuperLutheran for sure. This is why I'm putting brakes on their fun as it were. We'll either get to that point or they'll admit they were just fearmongering for fun and profit
   • Embed this notice
     Bread up, Bro (sickburnbro@poa.st)'s status on Thursday, 23-Jan-2025 02:43:25 JST Bread up, Bro

     in reply to

     • Berlusconi Fever :sonnenrad:
     • agaperealm
     • gentoobro
     @agaperealm @MechaSilvio @gentoobro how exactly do you think tech is "close to the singularity"
   • Embed this notice
     agaperealm (agaperealm@poa.st)'s status on Thursday, 23-Jan-2025 02:43:26 JST agaperealm

     in reply to

     • Berlusconi Fever :sonnenrad:
     • gentoobro
     @sickburnbro @MechaSilvio @gentoobro As someone in the industry who has had first hand knowledge of how close the tech is to the singularity I think it wise to hope for the best and prepare for the worst.
   • Embed this notice
     Bread up, Bro (sickburnbro@poa.st)'s status on Thursday, 23-Jan-2025 04:24:26 JST Bread up, Bro

     in reply to

     • BroDrillard
     • agaperealm
     • gentoobro
     @gentoobro @BroDrillard @agaperealm do you have any links to those studies? I think those might be interesting for this discussion in the future
   • Embed this notice
     gentoobro (gentoobro@shitpost.cloud)'s status on Thursday, 23-Jan-2025 04:24:27 JST gentoobro

     in reply to

     • BroDrillard
     • agaperealm

     The studies coming out are pretty bad for the true believers. LLM-generated code is full of hard-to-find bugs, and total productivity with the system varies between "slightly more" when used by drooling interns and "substantially lower" when used by senior devs.

   • Embed this notice
     Bread up, Bro (sickburnbro@poa.st)'s status on Thursday, 23-Jan-2025 06:02:03 JST Bread up, Bro

     in reply to

     • PalePimp
     • BroDrillard
     • agaperealm
     • gentoobro
     @gentoobro @PalePimp @BroDrillard @agaperealm right, isn't that what a ton of them were trained on?
   • Embed this notice
     PalePimp (palepimp@poa.st)'s status on Thursday, 23-Jan-2025 06:02:04 JST PalePimp

     in reply to

     • BroDrillard
     • agaperealm
     • gentoobro
     @gentoobro @BroDrillard @agaperealm @sickburnbro I always say the same thing, write me anything complex with it, like MS Office or AutoCAD using an LLM.
     
     I will wait patiently.
   • Embed this notice
     gentoobro (gentoobro@shitpost.cloud)'s status on Thursday, 23-Jan-2025 06:02:04 JST gentoobro

     in reply to

     • PalePimp
     • BroDrillard
     • agaperealm

     Write anything that isn't essentially a copypasta from Stack Overflow...

     Bread up, Bro likes this.
   • Embed this notice
     Xeraser (xeraser@varishangout.net)'s status on Thursday, 23-Jan-2025 07:49:35 JST Xeraser

     in reply to
     @sickburnbro
     I trust Dario more than i trust the openai kike (I don't trust him) because Claude is actually pretty alright for a llm but we're nowhere near close to AGI
     Bread up, Bro likes this.