How OrKa’s Maternal-Inspired Workflows Echo Geoffrey Hinton’s Call for "Maternal Instinct" in AI.
1. A Remark at Ai4 That Stuck
In Au...
For further actions, you may consider blocking this person and/or reporting abuse
Nice! It's only work on local models?
@ermes__59155de0c05eb Thanks!
No you can change all agent to
openai-answertype, adapt the model and it will run using openAI API.But it will be quite expensive if loop do not end within 2 or 3 rounds, the context tokens grow up exponentially at each run (up to 1Euros with 10 loops). Keep the max looping around 3 or 5 and if prompting is good you may find agreement way before hitting the max_loop. Good luck :)
Hey what of gemini or another free API call?
@daniel_githinji_6e1e402c3 thanks for asking. 0.9.x (witch I am working on) will have a more generic LLM nodes local and cloud. Right now implementation is only providing OpenAI or full local solution. I'm sorry, but also thanks. You are sort of validating my planned work 😅
Yeah and it also proves you are solving real problems, by the way I made a video on this tried telling gemini to be my parent but voila, it failed but venice AI did, cant wait to be a contributor as a dev
Nice, there are alwasys ways to find good deals. There are multiple ways to get free inference calls across Amazon Bedrock, Huggingface, Fireworks AI (well not free but very cheap), and using Google's code Assist, Qwen Code, Atlassian Rovodev (Claude Sonnet 4 and GPT 5) which can all comunicate and connect via CSP (collaborative stream protocol - a protocol I designed for direct semantic communication across all mofels, API or no API from local / VMs / other Clis) I'm happy to explain more to anyone who is interested.
Thanks for the comment either if I did not fully understand it...
Really interesting take! Like how you turned Hinton’s "maternal instinct" idea into something practical with OrKa’s workflows. The focus on comfort and protection makes this stand out from typical AI safety discussions. 🙌
Appreciate that! 🙌 Hinton’s metaphor struck me too, but I didn’t want it to remain a poetic image. With OrKa the goal is to operationalize that “maternal instinct” through workflows that actively prioritize comfort, protection, and stability when reasoning. In practice, it means routing decisions not just for efficiency, but for care, making sure the system remembers, reassures, and safeguards context the way we’d expect from a protective presence. That’s a different flavor of AI safety than control or constraint: it’s safety expressed as care.
Pretty interesting, and nice design!
Thanks! 🙏 @kurealnum
I love it!
Thanks!