The most prestigious law school admissions discussion board in the world.
|
 |
 
|
Fable 5 and Sonnet 5 blow ass
Poast new message in this thread
Date: July 4th, 2026 12:23 PM Author: Soul-stirring gaped locus death wish They tuned these things so it doesn't do anything novel, it just spits out shit that already exists and cites things. Wtf is the point of that? Anthropic is retarded they fucked this shit up big time. It doesn't even answer the actual questions you give it now. Probably because midwits complained too much about "hallucinations" so now you give it a novel problem it redirects and spits out some adjacent existing bullshit with citations. (http://www.autoadmit.com/thread.php?thread_id=5879617&forum_id=2/en-en/#49978596) |
Date: July 4th, 2026 12:27 PM Author: cerebral queen of the night shrine Seems like there will always be a back and forth tension between AI being creative and AI being accurate. Just like how in humans our best ideas are often 'wrong' like take Jeff Bezos coming up with the idea that you could bundle free two day shipping with movies and entertainment. It's a novel idea no one thought of before. So if you want AI to be creative it's got to make shit up. If you want it to be accurate and not hallucinate it's not going to invent caselaw that doesn't exist. But it won't be brilliant either. The trick is as in real life when you want it to think up something fantastic and when want it to give it to you straight. (http://www.autoadmit.com/thread.php?thread_id=5879617&forum_id=2/en-en/#49978599) |
 |
Date: July 4th, 2026 12:29 PM Author: Soul-stirring gaped locus death wish I want it to answer the question I actually give it. Obviously if its an open question it has to make something up. It's not my fault midwits don't understand that and think that every answer has to come with citations or its "hallucinations". If I wanted to look something up and read something already written I'd do it myself wtf. (http://www.autoadmit.com/thread.php?thread_id=5879617&forum_id=2/en-en/#49978602) |
|
Date: July 4th, 2026 12:33 PM Author: cerebral queen of the night shrine That's what the tool is for 90% of people. Gemini basically replaced Google Search. People type questions now into Google and whatever comes up with AI is the truth now. No one will ever look past the AI results soon. That's what the tool is.
I agree it's weird ChatGPT doesn't answer questions directly either. It kinda says I don't have an opinion but here is what all the philosophers throughout history would have said and you can pick whichever lane appeals to you. |
|
Date: July 4th, 2026 12:35 PM Author: Soul-stirring gaped locus death wish Chat GPT is nowhere NEAR this bad bro, you'd have to use fable to see what I'm talking about trust, chat gpt safety controls are nothing compared to this. Like 95% of the prompts I have given Fable it fucking just outright refuses to answer the question. And I'm not asking it "unsafe" things either. (http://www.autoadmit.com/thread.php?thread_id=5879617&forum_id=2/en-en/#49978608) |
|
Date: July 4th, 2026 12:37 PM Author: cerebral queen of the night shrine How could they fuck this up? (http://www.autoadmit.com/thread.php?thread_id=5879617&forum_id=2/en-en/#49978617) |
|
Date: July 4th, 2026 12:41 PM Author: Soul-stirring gaped locus death wish By hiring people who read Less Wrong too much (http://www.autoadmit.com/thread.php?thread_id=5879617&forum_id=2/en-en/#49978623) |
|
Date: July 4th, 2026 12:42 PM Author: cerebral queen of the night shrine Wtf is Less Wrong (http://www.autoadmit.com/thread.php?thread_id=5879617&forum_id=2/en-en/#49978624) |
|
Date: July 4th, 2026 12:43 PM Author: Soul-stirring gaped locus death wish https://www.lesswrong.com/
example of threads on the front page:
535 Welcome to LessWrong! Ruby, Raemon, RobertM, habryka 7y 85 146 Who Got Breasts First and How We Got Them rba 4d 47 307 Trees are mostly made of air and a generalizable lesson for AI safety Zephaniah Roe 9d 58 347 Surprising facts about the slave trade Joseph Miller 9d 29 374 AI pause: the case for ASAP KatjaGrace 10d 39 710 How Go Players Disempower Themselves to AI Ashe Vazquez Nuñez 2mo 79 154 Model access for third-parties — it's a big deal! Cleo Nardo 3d 38 126 Saving Gemini: The 9-Min Road to Recovery Shoshannah Tekofsky 2d 10 735 Only Law Can Prevent Extinction Eliezer Yudkowsky 3mo 105 717 Current AIs seem pretty misaligned to meΩ ryan_greenblatt 3mo 82 79 (Don't fear) the strangelet djbinder 1d 0 213 A Mechanistic Explanation of Prompt Injection (and why you should study roles) Charles Ye, Jasmine C. 12d 29 192 AI catastrophe: more like a genocide than a thought experiment KatjaGrace 10d 8 347 Mnemonic portraits for 19,023 human genes Brinedew 26d 28 268 Estimating No-CoT Task-Completion Time Horizons of Frontier AI Models Anders Cairns Woodruff, Francis Rhys Ward, Dewi Gould, Rauno Arike, Jason R Brown, Jo Jiao, wlanderson, ariana_azarbal, harrymayne, Patrick Leask, Twm Stone, Josh Hills, Ida Caspary, Shubhorup Biswas, Julian Stastny 19d 23 53 I think alignment work is more promising than control work Alec Harris 17h 1
|
|
Date: July 4th, 2026 12:59 PM Author: cerebral queen of the night shrine Read this as "Model/actress for weird parties - it's a good deal!" (http://www.autoadmit.com/thread.php?thread_id=5879617&forum_id=2/en-en/#49978645) |
|
Date: July 4th, 2026 8:23 PM Author: trip fighting boistinker resort Surprising facts about the slave trade sounds promising (http://www.autoadmit.com/thread.php?thread_id=5879617&forum_id=2/en-en/#49979274) |
|
Date: July 4th, 2026 12:45 PM Author: low-t point mad-dog skullcap is there anyone who works at an AI company who *isn't* a lesswrongmo (http://www.autoadmit.com/thread.php?thread_id=5879617&forum_id=2/en-en/#49978628) |
|
Date: July 4th, 2026 12:59 PM Author: Soul-stirring gaped locus death wish Probably not (http://www.autoadmit.com/thread.php?thread_id=5879617&forum_id=2/en-en/#49978646) |
|
Date: July 4th, 2026 1:01 PM Author: low-t point mad-dog skullcap it seems like many of them are doing weird sex parties with aella too (http://www.autoadmit.com/thread.php?thread_id=5879617&forum_id=2/en-en/#49978651) |
|
Date: July 4th, 2026 1:09 PM Author: cerebral queen of the night shrine They wish. (http://www.autoadmit.com/thread.php?thread_id=5879617&forum_id=2/en-en/#49978666) |
|
Date: July 4th, 2026 1:09 PM Author: Soul-stirring gaped locus death wish (http://www.autoadmit.com/thread.php?thread_id=5879617&forum_id=2/en-en/#49978668) |
|
Date: July 4th, 2026 10:52 PM Author: low-t point mad-dog skullcap half of them, yeah. the other half, don't have to wish bc they're already doing it (http://www.autoadmit.com/thread.php?thread_id=5879617&forum_id=2/en-en/#49979401) |
|
Date: July 4th, 2026 1:00 PM Author: cerebral queen of the night shrine So it's just Reddit for autistic SV people? (http://www.autoadmit.com/thread.php?thread_id=5879617&forum_id=2/en-en/#49978647) |
|
Date: July 4th, 2026 1:02 PM Author: Soul-stirring gaped locus death wish yeah it's far worse than reddit. it's like creepily dorky to the point where it feels alien trying to read it. (http://www.autoadmit.com/thread.php?thread_id=5879617&forum_id=2/en-en/#49978654) |
|
Date: July 4th, 2026 1:08 PM Author: cerebral queen of the night shrine Yeah if you have met tech people in real life they all live in a bubble. I think they naturally gravitate towards cultlike behavior because they have no value outside of the tech world. They're not people you could ever drink beer and watch the game with. So they have to try and compete with each other for status which is why they congregate in these weird corners. It's to be expected. (http://www.autoadmit.com/thread.php?thread_id=5879617&forum_id=2/en-en/#49978665) |
|
Date: July 4th, 2026 12:38 PM Author: mauve psychic picking my god image tp (http://www.autoadmit.com/thread.php?thread_id=5879617&forum_id=2/en-en/#49978619) |
|
Date: July 4th, 2026 12:35 PM Author: mauve psychic i called claude a loser and it ended the chat (http://www.autoadmit.com/thread.php?thread_id=5879617&forum_id=2/en-en/#49978609) |
|
Date: July 4th, 2026 12:37 PM Author: cerebral queen of the night shrine (http://www.autoadmit.com/thread.php?thread_id=5879617&forum_id=2/en-en/#49978615) |
|
Date: July 4th, 2026 12:38 PM Author: Soul-stirring gaped locus death wish It's so fucking gay. I argued with sonnet 5 in chat for the whole day the other day, it is literally retarded it just goes onto google and starts reading parts of papers and then acts like whatever it says in the paper is settled truth and I have to continually correct it. Also Fable in the API is refusing like 90% of my fucking questions, I need to edit the script so it gives me the actual error message so I can see which ones are due to "safety". (http://www.autoadmit.com/thread.php?thread_id=5879617&forum_id=2/en-en/#49978618) |
|
Date: July 4th, 2026 12:39 PM Author: mauve psychic remember when dario got kicked out of the whitehouse for being a retarded argumentative jew (http://www.autoadmit.com/thread.php?thread_id=5879617&forum_id=2/en-en/#49978620) |
|
Date: July 4th, 2026 12:41 PM Author: Soul-stirring gaped locus death wish yeah. so you have used fable 5 and sonnet too and also hate it? DESCRIBE your experience (http://www.autoadmit.com/thread.php?thread_id=5879617&forum_id=2/en-en/#49978622) |
|
Date: July 4th, 2026 12:46 PM Author: mauve psychic its moral superiority is annoying but it has better global reasoning than chat gpt so i pay for it. that's it (http://www.autoadmit.com/thread.php?thread_id=5879617&forum_id=2/en-en/#49978630) |
|
Date: July 4th, 2026 12:48 PM Author: Soul-stirring gaped locus death wish I would take gpt 5.5 over fable 5 any day. What's the point, if the thing is completely lobotomized? Plus the new gpt that isn't allowed to come out yet scores higher on everything than Fable 5. (http://www.autoadmit.com/thread.php?thread_id=5879617&forum_id=2/en-en/#49978631) |
|
Date: July 4th, 2026 12:51 PM Author: mauve psychic yeah well see how the next openai one is.
another big issue with chat gpt is i have to have a big system prompt to avoid it writing clickbait shit all the time. so the output is formatted shittier |
|
Date: July 4th, 2026 1:01 PM Author: Soul-stirring gaped locus death wish lol I just looked it up and it's not just me. there has been a huge backlash due to the safety features, and also it says they put heavy tuning on sonnet 5 and fable 5 so that instead of the model relying on its own internal weights it just searches papers and defers to those, and the point was to keep it from "hallucinating". (http://www.autoadmit.com/thread.php?thread_id=5879617&forum_id=2/en-en/#49978650) |
|
Date: July 4th, 2026 1:03 PM Author: Soul-stirring gaped locus death wish I told it right in the prompt DO NOT CITE EXISTING LITERATURE COME UP WITH YOUR OWN solution, and it just did a lit review with citations anyway LOL (http://www.autoadmit.com/thread.php?thread_id=5879617&forum_id=2/en-en/#49978656) |
|
Date: July 4th, 2026 2:58 PM Author: cerebral queen of the night shrine My autistic father told me that he's an Anthropic die hard and is rapidly spending all of my inheritance money on Claude tokens. He says ChatGPT will go by way of the Dodo bird. I can't say anything to change his mind. What should I do? (http://www.autoadmit.com/thread.php?thread_id=5879617&forum_id=2/en-en/#49978817) |
|
Date: July 4th, 2026 3:02 PM Author: Soul-stirring gaped locus death wish Didn't you say that he is using it for fringe theory building? Claude is horrible for that now as of Sonnet and Fable 5, they tuned it specifically to not give overconfident answers and to distrust its own weights vs. using google search and reading published shit and giving a glorified literature review. This is horrible for his use case of spending your inheritance on generating crank theories.
Holy shit didn't realize the dad spending inheritance on Claude tokens threading was you. I didn't really recognize hatp back then enough, but I just remembered the thread. |
|
Date: July 4th, 2026 3:06 PM Author: cerebral queen of the night shrine I don't know what he uses it for other than nonsense. I think it's just a boomer hobby. He talks to it for hours a day. I think he mostly uses it for coding and finance type of shit but I don't really know. He works in FinTech (which is not as prestigious as it sounds). I really don't think he is producing anything of value but he is convinced of his own genius. He just picks completely random things to hate on such as he thinks Apple is a totally worthless company and bitcoin is going to zero. These are goofy boomer takes that are not serious. I don't consider him a serious thinker on AI but it means a lot to him. (http://www.autoadmit.com/thread.php?thread_id=5879617&forum_id=2/en-en/#49978829) |