Good to see actually experienced people in the technology saying stuff like this.If you don't know this guy, he was the guy that made 4chan GPT and was having it post hundreds of replies and no one noticed the bot posting until he switched to a rare flag to signature them.https://youtu.be/hkAH7-u7t5k
>>106454047This guy will be fired or de-personed within a weekAGI is the last big grift that AI companies have. It's the final frontier of empty promises to keep investors and speculator money flowing. Once AGI dream is dead, that's when it comes crashing down. Otherwise, nobody would support an industry with 95% losses.
>>1064540472 more weeks thoughbeit!
>>106454059yup its unfortunate to because i really like using the llms the way he describes it in this video>theyre glue to feed/mix/format data between various tools i dont need it to be sentient to get use out of it
AGI is not coming (for the general public)
I don't think you can make it currently. Need to interface specially curated human braincells into whole data centers to achieve actual intelligence.Regurgitating words based on mathematical predictions is not intelligence.
>>106454068>2 more weeks thoughbeit!This sounds more like the mantra of the people saying that the "bubble" is gonna pop any time soon. As if investors give a shituber ran at a loss for over a decadeyou all sorely underestimate the amount of money these people have to burn and overestimate how much they value it
>>106454122good point. this is the stuff the crazy fucks get into when you think like>what does the richest man in the world spend his money on?and no i dont mean elon, hes a puppet/actor. the real richest people have never had their face photographed.so this person, with infinite money, would he invest his useless infinite money on a sentient computer program that can help him rule the world? definitely.
>>106454047>Yannic Kilcher
>>106454047>literal whothe bigwig himself says so
>>106454047>actually experienced people>guy that made 4chan GPThe's knowledgeable but nowhere as experienced as the big majority that push for AGIonly laion people acknowledge him and the rest barely interacts with him
>>106454757>as the big majority that push for AGIno lol, those people are twitter posters. this guy actually uses the tech, shutup faggot
>>106454122so your argument is that a company that's pulling 8 figures in revenue with 10 figures in costs that hilariously claims it's worth 11 figures in capex with absolutely no intention of going public and no prospective buyers... is fine?seriously, go look up uber's revenue and burn rate at the peak if you want to make that case. this thread will be right here when you are ready to substantiate your claim with numbers.
>>106454047>OpenAI blows their lead, hemorrhages talent, and pivots to mid preferencemaxxed tiny quants to serve their billion normie users>Meanwhile China is about to pull ahead, Google is solving world models, humanoids are improving literally daily>Muh AI winter, the bubble popped, it's overFrog is being boiled, zoomers get upset without a new weekly release to satisfy their fried attention spans
>>106454820gpt5
>>106454068Both are disgusting
OpenAI's diminishing returns LLM system is a dead end but other companies and areas are much more promising. Not for "muh AGI" but for solving and automating other things.
>>106454890For example? I agree just wondering what your experiences are. I think Google is going to dominate.
>>106454890aren't they all the same shit? Are other guys using something other than LLM system?
>>106454047>If you don't know this guy, he was the guy that made 4chan GPThttps://youtu.be/efPrtcLdcdMhttps://en.wikipedia.org/wiki/GPT4-Chan>The development of GPT-4chan began in May 2022, when Kilcher announced his project on his YouTube channel.[3] Notably, at the time before ChatGPT, he explained that he wanted to create a large language model that could generate realistic and coherent text in the style of /pol/, one of the most notorious online communities.[4] >takes opensource GPT model>trains it on 4chan>runs it against LLM benchmark test suites>beats the bleeding edge foundation models from OpenAI in "Truthfulness skill"simply wonderful
