We're building a new type of AI designed to be emotionally and ethically intelligent. Instead of a single, massive model, our system uses a network of specialized subsystems that work together, much like the human brain.
Loosh AI
589 posts
Loosh AI
@Loosh_ai
A cognition engine designed for robotics and agents, empowering them with self awareness, self reflection, deontology, memory, and emotional inference.
- We are launching soon, get readyPress start ▶️ Get set for @Loosh_ai 🚀Our first upcoming subnet team Conscious robotics, trained by multimodal cognitive inference, right here on #Bittensor Crowdfund coming (very) soon. Link in bio✌️
- Hey everyone it’s been a while since we had an update, we have been making huge progress: - our 1st engineering hire was on-boarded and has been finalizing the subnet codebase in preparation for deploying on bittensor - we brought on a designer to work on our branding elements
- This past weekend the Loosh team had the privilege of attending a conference that pushed the boundaries of science and consciousness. In the words of Tom Campbell, “Consciousness is fundamental and from consciousness all else is derived. Discovering the coherent structure and
- All miners will have an opportunity to earn emissions with our subnet. Every miner that wins gets a reward.
- Our humanoid robot prototypes arrive in November We are very excited to integrate our cognition sdk Until then our ethical reasoning module will be ready for testing in the next couple months If you’re interested in helping test drop a comment below 👇
- AI and robotics aren’t just advancing—they’re converging. Every new humanoid demo is a reminder: what felt like sci-fi yesterday is becoming practical tech today. We are preparing to launch our subnet, our inference and training data that directly feeds into our humanoid robots
GIF - 🧵 1/ What if consciousness was infrastructure? Not metaphorically. Literally. Loosh is building decentralized cognition — a substrate where machines and humans can meet in awareness.
- As part of our updates heading into the launch of the subnet on Bittensor, we are sharing more about what is happening behind the scenes: Loosh's narrative memory empowers our robots to learn as they encounter new situations enabling a responsive and adaptive artificial
- We are going even further than this Our architecture is designed with subsystems replicating human thought processes And our models are trained to understand non verbal communication and conscious states🧠 Scientists build an AI that can think like humans Brain-scan alignment hints AI internal states now resemble human reward circuits. This is from Helmholtz Munich and they did it by feeding psychology data turns Llama-3.1 into a brain-like predictor without sacrificing
- Most humanoid robotics that we see today are operated via remote control Those that are standing in an assembly line have been trained on that one specific task The robots cleaning homes today are functionally no different than the robot vacuums sold 10 yrs ago What’s
- Fine tuned models + proper RAG + proprietary data = competitive moat We aren’t using AI trained on everybody’s data Our customers will expect more than that
- It’s time for our weekly update: - miner codebase is done - validator codebase is done - inference subnet is ready - training subnet is currently in development - we have hired our 2nd person: a data scientist with a background in neuroscience 👀
