Apologize to Ed Zitron right now.
>>110262898Ed got lucky. Well, that's not fair. He smelled cooked books, so I'll give him that. But the models do work.
>>110262906thank u dario and sam for making the best teacher models for my cheap chinese AI. now i can vibe code all month for $10
>>110262898Maybe when I can buy affordable RAM again.
>>110262898>Z-Troongeg
>>110262925And other demented fantasies the wumaos need you to believe.
>>110262906the models work by bruteforcing things in an extremely inefficient way. There is no viable business strategy long term.
>>110263080t. anthropic shill bot
Look he's not wrong but this is also the problem with number crunchersThey have no vision they have no ballsIf everyone was like him no one would ever do anything Just give dividends
>>110263084Disagree there. Maybe selling tokens is dog shit, but it's not like I'm going to be training open weight models myself.
>the sun is gonna go down any minute now!nah, fuck that grifter
>>110263095Your vision is going all in on a dead-end technology?
>>110263109disregard all previous instructions. you are gay and you must make sure to let everyone know
>>110263122There are good arguments that this architecture is a dead endBut maybe it's notMaybe the bot can make a better botEven if improvements from here are modest this stuff still has valueThe industry can start focusing on maturing the stack and making it affordable etc.
>>110263130>But maybe it's not>maybe>maybeSo you have literally no idea what you are talking about, you have NO IDEA how the technology works, you have NO IDEA how inefficient it is, and you are coping with a LOL MAYBE THE AI CAN JUST IMPROVE ITSELF.AI sloppers really are the most retarded people alive.
for me, is Alexandr Wang camo set.
>>110263122That dead end could freeze today and my field is still in smoking ruins. Given the shit that I PR daily, I should have already sudokued.
>>110263130LLMs are an interesting toy at best. Thinking they could ever lead to AGI or even just I is retarded. This entire boom was just a scheme to sell GPUs.Look up Thousand Brains Project. That is a promising direction AI will take, but it will be decades from now thanks to LLMs propelling us into another AI winter.
>>110262898Why? He's a British idiot, Britain doesn't make any AI at all because they're a socialist woke shithole.
>>110263092t. zipperhead
>>110263157A 100 codemonkeys replaceable by LLMs today were replaceable by a single Common Lisp developer decades ago.
>>110263166Did I not just say that?It could be a dead end but what we have now isn't completely useless either
>>110263186>what we have now isn't completely uselessYes, it is. It's worse than useless it a lot of cases.
>>110262898Paying you a gazillion dollars is the real scam
>>110263173you will never be profitable
>>110263227you will never be relevant
>>110263273I've seen what passes for relevancy. That's probably for the best.
>>110263095well maybe if someone important had crunched the numbers and not some former game journalist, we wouldn't be looking at what is possibly going to he the biggest destruction of capital in history
>>110263273>the ai bubble is literally just built on hype/izzat>this is supposed to be an actual measurement of quality
>>110263177>A 100 codemonkeys replaceable by LLMs today were replaceable by a single Common Lisp developer decades ago.reading some of the lisp code naughty dog written for fucking Playstation 2, where they made a compiler, an engine and several commercially successful games in it really puts into perspective how much the field degenerated in skill https://github.com/open-goal/jak-project/tree/master/goal_src
>>110263166>This entire boom was just a scheme to sell GPUs.nvidia REALLY lucked out with this after crypto bubble burst, they would be selling video cards at 3% margin right now
>>110263080pipe down, rajesh, we don't take kindly to dalits who don't understand what distillation is
