mete
@VaderResearch • 44,234 subscribers
first principles | @imperialcollege alum
Shorts
Videos
.Andrew Kang compares humanoids costs to human labor "A humanoid will cost $2 per hour A human costs $35 per hour in the US Humanoids can do everything humans can do They don’t need to rest and they don’t quit A humanoid will replace 3 humans Even in low-cost labor countries Humanoids will become a cheaper option"
mete22,064 views • 4 months ago
Baldur's Gate 3 didn't take any VC money Took over 6 years and $200m to develop Sold 2.5m early access units in 2020 at $60 each - generating $150m gross from gamers who had no financial expectations Crowdfunding at its finest The game ended up being a huge hit, generating $200m gross only 15 days after its launch Those 2.5m early access buyers aren't involved in any decision-making via DAOs. Nor are they getting any financial upside from the success of the game. And they're completely fine with it. But what if... The studio behind Baldur's Gate 3 launched a token, pledged 10% of revenue to buy back this token regularly and allocated 100% of the token to the 2.5m early access buyers?
mete91,694 views • 3 years ago
Most robot demos are scripted. Generalist's GEN-1 is not. > GEN-1 was doing a task > Mid-task, an extra object was thrown into the bin > GEN-1 quickly adapted to the new environment > Still completed the task smoothly This is the difference between a scripted policy (seen in viral launch videos) and a real foundation model. GEN-1 was trained on 500,000 hours of proprietary, real-world dexterous manipulation data. It has NEVER seen that exact scenario before. This was the coolest demo at Automate Show Generalist Flexiv Robotics
mete11,651 views • 3 months ago
Egocentric Videos for Robot Training 🧵 Researchers at NYU and UC Berkeley published research where they developed a system called EgoZero that trains robots using human demonstration videos recorded with smart glasses. 👓 The system converts first-person human actions into 3D point-based state-action representations. The policies executed on a gripper-equipped robot, achieved a 70% zero-shot success rate across seven manipulation tasks, with only 20 minutes of human data per task. EgoZero stands as one of the solid empirical proofs that egocentric smart-glass video collected from everyday human behavior can serve as powerful, scalable training data for real robot learning. Vincent Liu Ademi Adeniji
VaderResearch23,055 views • 1 year ago
Short update video on what we’ve been building! TLDR: > Been working at Eastworld Labs w/ a Unitree G1 > Focusing on commercial deployment in manufacturing > Shipped a new tooling to train VLA & WAM models > Shipped SimPlay; do tasks with a G1 in sim env to earn > Shipped new version of EgoPlay and litepaper
mete12,276 views • 5 months ago
No more content to load