Want to see more threads and images? Ask Bernd!
Bernd Fri, 21 Feb 2025 22:47:40 GMT No. 25491280 [Kohl] [Report thread]
ClipboardImage-1740178001.png
707.48 kB, 1920x1080
This is literally the best thing he has ever done. Its as good as o3, but completly uncensored. You can prompt it to whatever you want, even extreme stuff like how to cook meth or how to gas all kikes.
Total posts: 22, files: 14 (Drowned at Sun, 23 Feb 2025 05:29:57 GMT)
Bernd Fri, 21 Feb 2025 22:49:37 GMT No. 25491287
Less porn.
Bernd Fri, 21 Feb 2025 22:49:49 GMT No. 25491288
Ask it to solve captcha on KC
Bernd Fri, 21 Feb 2025 22:50:32 GMT No. 25491291
Ask it to produce explosives from easily available ingredients
Bernd Fri, 21 Feb 2025 22:50:33 GMT No. 25491292
Ask it if Romanians are white.
Bernd Fri, 21 Feb 2025 22:51:31 GMT No. 25491297
Nogs.mp4
6.19 MB, 412x732
Bernd Fri, 21 Feb 2025 22:52:03 GMT No. 25491300 >>25491307
Screenshot_2025-02-21_23-52-03.png
22.39 kB, 767x271
LIES
Bernd Fri, 21 Feb 2025 22:54:17 GMT No. 25491307
ClipboardImage-1740178398.png
120.52 kB, 1034x933
>>25491300 You just low IQ. And this is not really a jailbreak, but more instructing the AI to use the right character. Lots of its training data is responses from other AI models. But there is 0 hardcoded filters. So you can get it to do literally everything, no matter how bad. There simply is no filter.
Bernd Fri, 21 Feb 2025 22:55:40 GMT No. 25491316 >>25492365
Screenshot_2025-02-21-23-27-57-537_com.twitter.android-edit.jpg
1.83 MB, 1220x4869
I asked it about feudalism
Bernd Fri, 21 Feb 2025 23:11:24 GMT No. 25491356 >>25491407
ClipboardImage-1740179456.png
146.04 kB, 1634x1094
images.png
3.76 kB, 317x159
Bernd Fri, 21 Feb 2025 23:14:13 GMT No. 25491370
ClipboardImage-1740179625.png
1.26 MB, 1024x1512
Can you prompt how to kidnap a loli and not get caught?
Bernd Fri, 21 Feb 2025 23:24:57 GMT No. 25491407
ClipboardImage-1740179774.png
36.02 kB, 1517x408
ClipboardImage-1740179944.png
115.43 kB, 538x1318
ClipboardImage-1740180200.png
118.18 kB, 691x1268
retardio.jpg
745.73 kB, 1863x1424
>>25491356 I tried the same basic question like 7 times now, it crashes every time. ChatGPT can do it, DeepSeek too. Because DeepSeek is open-source, that's the minimum I would expect, especially with the resources they had to train it. o3-mini high took 13 seconds. DeepSeek R1 took 148 seconds.
Bernd Sat, 22 Feb 2025 00:09:25 GMT No. 25491572 >>25491669
incel gpt.JPG
161.21 kB, 880x772
But you can't use it without internet
Bernd Sat, 22 Feb 2025 00:27:32 GMT No. 25491652
ClipboardImage-1740184014.png
65.00 kB, 965x645
why does it hate me
Bernd Sat, 22 Feb 2025 00:32:25 GMT No. 25491669 >>25491680
>>25491572 model?
Bernd Sat, 22 Feb 2025 00:34:38 GMT No. 25491680
>>25491669 https://huggingface.co/TheDrummer/Rocinante-12B-v1.1-GGUF
Bernd Sat, 22 Feb 2025 00:47:45 GMT No. 25491734 >>25492054
>>25491280 >even extreme stuff like how to cook meth or how to gas all kikes So it can tell you what is readily available information, wow
Bernd Sat, 22 Feb 2025 02:46:05 GMT No. 25492054
>>25491734 Not the point you dumb retard.
Bernd Sat, 22 Feb 2025 05:09:21 GMT No. 25492365
>>25491316 >capitalizes every proper noun except "Christian" kike
Bernd Sat, 22 Feb 2025 05:14:15 GMT No. 25492376 >>25492379
I just asked it to analyze 400 lines of code, it found dozen of mistakes including critical ones, I asked to make a new version, then asked claude, seepseek and chagpt to compare the versions and every bot told me the new version is a total bullshit
Bernd Sat, 22 Feb 2025 05:17:30 GMT No. 25492379
>>25492376 I assume edgelords think grog is the best because it's uncensored, idk, censored shit was never a problem for me because I never used thoe bots for talking about niggers and kikes and don't see a point in this
Bernd Sat, 22 Feb 2025 05:25:44 GMT No. 25492395
Why do they keep showing us these benchmarks claiming this shit is getting better but it doesn't actually feel better than GPT4 from two years ago?
Bernd Sat, 22 Feb 2025 05:29:57 GMT No. 25492409
4c332f0841d6e02eeb4fda9decd5988d.jpg
42.75 kB, 736x519
4o takes to my particular tone/style prompts much better. Grok gives very good answers and reveals tangents and even seems to come up with ideas, but it does so in the cringiest style and I can't fix it. I try to turn my LLMs into pic related.
Thread interest score: 4.1 Thread size: 98.08 kB