Xingxin on Bug I paused my literature review this week and started designing my experiments. Mostly, I dealt with logistics, ordering materials, tools and other necessary supplies.
One thing has been quite surprising to me lately. I hate to say this since I haven’t published a paper myself yet, but some papers claim to achieve zero-shot sim2real transfer, yet their demo videos have clearly been edited. For example, in one frame a bottle is simply near the gripper, and in the very next frame, it is magically grasped…😂
Especially in the era of AI agent, generating texts is no longer constrained by time or human effort. It’s actually a bit scary. This reminds me of a post on I saw on X that went something like this:
Authors use AI to write papers, and reviewer uses AI to review papers.
This is a critical issue, but I think it is a trend we can no longer avoid. We either join in or we get left behind. However, exactly how to best leverage AI to help us do genuine research still remains a big question for me.
A few interesting readings from this week: