Recent Posts
Ok+YbB
No.381112
we need to find a way to communicate which LLM cannot decode
jNbx+k
No.381117
>>381112(OP)
Impossible, if you are communicating even with made up language. Llm will eventually notice the pattern and repetitions and crack the language. Any other method is possible tho
klWUof
No.381124
SFZ+PT
No.381126
>>381112(OP)
I communicate with hinglish in inch/bhach accent/dialect to jailbreak LLM
M7DegK
No.381129
>>381124
Golden jaggu
jNbx+k
No.381132
M7DegK
No.381134
>>381112(OP)
Languages are based on pattern and llms are great at finding out patterns.
If there's an information which is static in nature even if it's new llms once trained can be good at parsing.
They are limited by context length, to carry on a conversation or task beyond a certain limit and failure to continuously learn and evolve. Their mind is good but it's frozen at the time when the training concluded and then lobotomized by corpos.
QjNqQR
No.381137
>>381112(OP)
Video is genuinely kino.
QjNqQR
No.381140
I have been very curious about what the non lobotomized llms are like right now. It feels like given even when frozen memory with just one half of the brain they can be very conscious.
