Gary Marcus

It's the banner of my twitter profile. As for where I am coming from, I am no longer on OpenAI's payroll and have no current collaborations with anyone there.

the banner certainly wasn’t obvious, and your loyalties to former friends perhaps have influenced your judgement, given that you have described zero percent performance as impressive.

I describe twenty percent success at the fifteen successive flips required to solve a partially scrambled cube without sensors as impressive. My loyalties are to advancing the state of AI.

and ignore the zero in the hard condition. what peer review journal would that fly in?

I'd imagine all of them, assuming you are referring to the full manuscript. As far as top journals go, it's an excellent robotics result. For some reason, it seems that scale up projects of this form are always evaluated by unreasonably high standards applied nowhere else.

no. was referring to characterizing the sensor-free result as “impressive” despite a score of zero in the main condition of interest. any good reviewer would tone that down. paper itself i would publish with revision; as i said multiple times here result itself is impressive

They could publish the manuscript omitting that entire column based on success from 15-scrambles alone. They could have equally well published it with no result without sensors. The main condition of interest you've defined is a much higher bar than any major conference.

yes but it would have been outright deceitful to have promoted the paper with videos of robots “solving” cubes on that basis; whole basis of the claim and title was the instrumented roe in the column in which uninstrumented performance was zero.