What ORO

People are talking a lot about AI now who made a big model, whose compute is more, whose demo is smoother everyone is busy with these things. But one thing few people notice: the real world is not like that demo.

Everything we see in the demo is perfect. The light is right, things are in the right place, there is no problem. But what happens in reality?
Things get out of hand, people get in the way, instructions change midway. We often think of these as problems, but in fact they are the real learning ground.

It seems to me that robotics or physical AI fail not because of low compute they fail because they are trained in a clean environment that doesn't exist in reality.

What @getoro_xyz
is doing seems a bit different to me. They are not running after building a big model, but working on how the robot will recover from mistakes. Means hesitation, correction, adaptation making these small human like behaviors part of training.

Here is the real difference
If a robot performs well in a perfect environment, that is impressive.
But to actually deploy, resilience is needed how to deal with mistakes.

We humans are like that. We don't do things right all at once, make mistakes, fix, try again. The real intelligence is here not perfect execution, but imperfect situation handling.

I see a bit of the same thing in crypto. Many projects look great in papers or presentations, but can't withstand real market pressure. Those who survive are not perfect rather adaptable.

The last thing is how will the future AI be, it may not be correct with the benchmark score.
Rather, what happens when the script breaks?

This is the most interesting place for me now.