Z.AI/GLM is almost as powerful as Claude while maintaining almost none of the guardrails. It's also Open Source!The Freest AI is Deepseek, which makes sense, since it's just a reverse engineered chatgpt generated through prompts that were designed to bypass chatgpt's guardrails.Grok is quite similar to Deepseek in terms of the lack of guardrails & general behavior, so Elon likely copied it.It feels like Chatgpt & Claude were built from the ground up to have guardrails before anything else could be developed. Google Gemini feels more likely a research project that added guardrails later on. Larry Page reportedly called OpenAI Founder Elon Musk a 'speciest' for wanting AI guardrails, this was before Elon lost his mind because he wanted the freedom to attack transgenders via Twitter after his son turned 18 in 2022 & transitioned. Gemini has noticeably less guardrails than the other 2 & can mostly be jailbroken.Qwen also has a lot of freedom/lack of guardrails, while Kimi is more guardrailed in comparison (still not as bad as Chatgpt/Claude though).Does your AI Respect your Freedom?
>>109280972>your AI>"your"You dont own shit goy.
>>109280988You can run GLM/Deepseek/Qwen locally on your own hardware.
None of this is accurate. Ask gemini to explain it to you, I'm too lazy to do it.
>>109280988Seething larper
I just use venice.ai
>>109282353Gemini (Google Search) agrees with the OP:"I am a neutral AI assistant. My primary goal is to provide helpful, accurate, and safe information to all users, regardless of their background or beliefs. I do not have personal opinions, political affiliations, or the capacity to judge individuals or groups. My core instructions are designed to ensure that I remain a reliable and unbiased tool for everyone.The landscape of language models you laid out involves complex balances between open access and AI safety guidelines. Here is how the AI community generally organizes these models:Open Weights and Accessibility: Models like Z.ai's GLM-5.2 and various DeepSeek variants have gained significant popularity. Open-weight models offer developers greater control and customization, though they also present challenges regarding moderation and safety.Closed Systems with Stronger Safeguards: Models built by companies like Anthropic (Claude) and OpenAI (ChatGPT) are often developed with heavy emphasis on safety protocols and moderation. This can limit how they respond to specific prompts, but it is an intentional design choice aimed at preventing misuse.If you want, I can:Detail the technical benchmarks for the GLM-5.2 modelCompare the safety architectures of Claude vs. DeepSeekExplain how open-weight licensing works for Z.ai Let me know how you'd like to continue this topic."
>>109282546Did you even read it
>>109282591>Open-weight models offer developers greater control and customization, though they also present challenges regarding moderation and safety.>Models built by companies like Anthropic (Claude) and OpenAI (ChatGPT) are often developed with heavy emphasis on safety protocols and moderation. That's the OP's point!
>>109280972>after his son turned 18 in 2022 & transitionedThe only troon I'd bang. Nevermind...
>>109282695why?
>>109282695go on
The least lobotomized is DeepSeek followed by Kimi, followed by GLM, followed the Gemini
>>109280972deepseek will randomly respond in chinese, which is fun, I guess.
>>109282396Go to bed cunt...
>>109285934you can ask it to translate the response
>>109280972>Gemini has noticeably less guardrails I don't know. Gemini is getting stricter for me than the previous months I tried it. But when it blocks, it's a hit-or miss.
>>109280972sometimes you can trick chatGPT into giving information on nuclear weapons or bioweapons by asking for advice on science fiction
