Fei-Fei Li is Solving the Hardest Problem in Robotics | World Labs with a16z

Comments · 68

• TH

  @ThuyThai-o5g · 1 month ago

  Amazing from start to finish.

• SA

  @Samanbeachhikkaduwa · 2 months ago

  Great...<br>Thanks so much for conversation, explaining, focusing and sharing this conference...<br>🌐🙋‍♂️☘☘☘☘☘

• AI

  @AITECH-k1h · 2 months ago

  Every technology that changed our lives started as an impossible idea.

  5

• DA

  @davidhoracek6758 · 2 months ago

  Yeah, that&apos;s how we finally use AI to change the world - not to create code, but to create things. I want to see one of those &quot;alone on an island&quot; reality shows where a guy shows up with a rowboat&apos;s worth of gear, including solar panels, robots, Starlink antenna, and accessories. Then the guy and the robot team up to transform the physical environment for a year, and see how much they can accomplish together! Like, how comfortable can they make the island for the guy to live on?

  45

• CA

  @calvinlin8274 · 2 months ago

  Good interview, thanks

• NO

  @normalPlayerSeverEarth · 1 month ago

  Great talk. But how do you mitigate the differences between simulation and real world? The simulator sometimes generate scenarios that don&apos;t exist in a real world. And sometimes it is simplified for programming. For example, an uncontrolled high torch of a gripper could break the gripper itself in real world, but in simulation they are configured as a articulated joint and only move in their range no matter the force, so it will never break in simulation.

  1

• RE

  @Read-le4ev · 1 month ago

  Thank god the questions were curious type

  1

• LO

  @louisliu6488 · 2 months ago

  What you mentioned is our DAP Digital Backbone. DAP Digital Backbone would be the Digital Twins of international trade and project management, based on the AI engines as the fundamental power.

• CH

  @ChristianGalindo1118 · 1 month ago

  Who is Colombian teacher that mentioned in video Min <a href="https://www.youtube.com/watch?v=-tabaM5l3s0&amp;t=462">7:42</a> ?

• 林金

  @林金根-e4s · 1 month ago

  Marble and the Real-to-Sim-to-Real pipeline brilliantly establish System 1 data scaling for 3D spatial perception and simulation. However, visual consistency is not physical rigidity; the Sim-to-Real gap is fundamentally physical state dispersion ($\sigma^2 &gt; 0$). To build unshakeable robotic assets, you must integrate discrete conservation laws and DTT-style Kernel verification (System 2/3). Compressing simulation variance down to &quot;Zero Dispersion&quot; is the only key to transforming generative geometry into 100% reliable physical intelligence!

  1

• UM

  @umeshpatelwithpriyankapate4873 · 1 month ago

  My 5-minute break turned into watching the entire video.

  2

• DE

  @DERAJEL · 2 months ago

  :face-green-smiling::face-pink-drinking-tea: