Large language models (LLMs) face issues with censorship, lack of randomness, and stochastic behavior that hinder their effectiveness.

Title: P2: Generate customer review task [2024-03-02 Sat] - LLM stochastic behaviour problem - LLM intelligence hiding - censorship during posting - not enough randomness I see 3 ways to fight censorship: 🦾 - Self-hosted fine-tuned models with unique characteristics as a person - Chains of advanced and constantly changing prompt engineering techniques to query corporate LLMs, possible with NN, constantly seeking new holes. #dailyreport #llm #ai #gpt #yandex #unemployed #jq #json

Title: P3: Generate customer review task [2024-03-02 Sat] - Generative adversarial network (GAN) that simulate corporate censorship 'to strike back/first' To hide intelligence and add randomness a careful prompt engineering should be enough for practical usage. Stochastic behaviour problem is advanced problem binded with limitations of current LLMs and hardware. ¯_( =| )_/¯ #dailyreport #llm #ai #gpt #yandex #unemployed #jq #json

Title: P2: P1: Generate customer review task [2024-03-02 Sat] texts and forums, that is it. This s*it is in my *ss actually, can not do anything bad in my country, but at others... Here is my big IMHO :) for this task problems: - LLM censorship at generating #dailyreport #llm #ai #gpt #yandex #unemployed #jq #json

Title: P3: Generate customer review task [2024-03-02 Sat] - Generative adversarial network (GAN) that simulate corporate censorship 'to strike back/first' To hide intelligence and add randomness a careful prompt engineering should be enough for practical usage. Stochastic behaviour problem is advanced problem binded with limitations of current LLMs and hardware. ¯_( =| )_/¯ #dailyreport #llm #ai #gpt #yandex #unemployed #jq #json

Title: P2: Generate customer review task [2024-03-02 Sat] - LLM stochastic behaviour problem - LLM intelligence hiding - censorship during posting - not enough randomness I see 3 ways to fight censorship: 🦾 - Self-hosted fine-tuned models with unique characteristics as a person - Chains of advanced and constantly changing prompt engineering techniques to query corporate LLMs, possible with NN, constantly seeking new holes. #dailyreport #llm #ai #gpt #yandex #unemployed #jq #json

Title: P2: P1: Generate customer review task [2024-03-02 Sat] texts and forums, that is it. This s*it is in my *ss actually, can not do anything bad in my country, but at others... Here is my big IMHO :) for this task problems: - LLM censorship at generating #dailyreport #llm #ai #gpt #yandex #unemployed #jq #json