Unverified50% confidenceFactExact time
Physical trial and error in robotics introduces incompressible physical constraints (grasping failures, collisions, force feedback), which are information that symbolic systems cannot reason out from nothing.
1
Sources
50%
Confidence
Long-term
Relevance
10/16/2026
First Seen
Sources
The Bottleneck of Intelligence Is Not Intelligence, But Continuous Contact with Reality
twitterfchollet10/14/2026
Related Entities
Related Claims
UnverifiedIn robot demonstration collection, incorrect hand reconstruction will directly mislead the downstream imitation learning policy, causing the robot to learn actions that humans never performed, and errors in this kind of data itself will not be clearly exposed in the training loss.74% similarUnverified机器人任务的分布外（OOD）失败会直接引发物理后果，如损坏物体、伤害周围人员或造成设备故障，这与图像分类等纯感知任务不同73% similarUnverifiedAI-generated animation has limitations including temporal consistency issues, physical realism defects (such as incorrect finger counts, objects suddenly disappearing or deforming), insufficient artistic control precision, and lack of material realism.73% similarUnverifiedScale transfer from miniature robots to full-size humanoid robots remains an unsolved problem, because mechanical properties change nonlinearly with scale.72% similarUnverifiedRobotic manipulation of flexible objects is a recognized challenge, as they possess infinite-dimensional deformation degrees of freedom and continuously changing shapes, making traditional model-based algorithms nearly inapplicable.72% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/1044403API
curl https://kongchang.com/api/v1/knowledge/claims/1044403MCP
get_claim(id=1044403)