Pinned
Auki
2,996 posts
Auki
@Auki
Making the physical world accessible to AI by building the real world web for robots and intelligent machines.
Links: linktr.ee/AukiNetwork
- Auki reposted
A strong gathering for Physical AI in SF. Thanks to everyone who joined the embodied AI meetup with @RichardRobinson. Special thanks to @GEODNET’s @mikeahorton and @AukiNetwork’s Nils Pihl (@broodsugar) for helping drive the conversation.
- Auki reposted
. @Auki picked up one of the first handful of Galbot S1 units to ever leave China. The Galbot S1 is a 300+ kg semi-humanoid robot, 20kg lift per arm, described by the team as currently the strongest semi-humanoid on the market. They're working with Galbot and one more partner
00:00 - When I close a very exclusive conference by being on stage with Dave McClure and Richard Robinson (the top-rated speaker), it feels surreal. I spent hours with Richard last night, and the dude is lit. Sometimes you meet people who are just alive, and Richard is one of those.
- - Galbot S1 arrival! - Strongest semi-humanoid available - Working with partner on stealth use case - We demo'd anomaly detection + live robot feed + ability to send robot anywhere - Community data improving our ESL detection - Will test Galbot in replica of target environmentAuki Update June 19 x.com/i/broadcasts/1…
- Auki Update June 19 x.com/i/broadcasts/1…
- Thanks for visiting and sharing, @TheHumanoidHub! Our plan: 1. Capture territory with immediate ROI on perception tasks. 2. Add manipulation capabilities, or let others do so. 3. Capture more territory & expand ecosystem of builders using our robots as infra. 4. Repeat step 3.Nils Pihl, founder of Auki Labs, gives me a tour of the Hong Kong lab. Instead of waiting for manipulation to work, Auki deploys robots that perceive: in this use case, they roam stores at night, catch empty shelves and wrong price tags, and leave the tasks in a shared
00:00 - Newest Auki crew member! @GalbotRobotics G1's heavyweight sister, the S1, has entered the building. She can lift up to 50 kg (110 lbs). 🦾
00:00 - Preliminary results from training our own price tag detection model on ~850 community-annotated shelf images: F1 score of 90.1% vs 51.7% on the open source model we were using before (7x fewer misses, 3x fewer false positives). Many thanks for helping us improve so much! 🤍
Reminder: you can upload images of shelves from supermarkets, convenience stores, or DIY stores and earn 30 $AUKI per price tag or empty space annotation. 🛒 auki.com/community/news… - With data collection costs plummeting alongside hardware costs, your data may not be the moat you think it is.
