The author argues that doing things at higher pace should be cognitive intensive. But robots are not bad at doing things fast: if they can do things, they can do them also "fast" (at my understanding)
Robots are not a single thing. You can't draw general conclusions about them like that.
They are objects designed by humans, they go as fast or as slow as they are designed to go. The design space is heavily constrained by physics. This means some things are possible, some things are possible but very expensive and other things are just not possible.
A FGM-148 Javelin anti-tank missile is a robot. It has an onboard computer, a sensor, and fins to change direction. It flies about 300 m/s as top speed and it doesn't do things slow. It can't. [1]
A Konecranes AGV is a giant self-driving platform carrying containers in automated ports. It carries a maximum of 70t and the top speed of it is about 6 m/s. It is also a robot. [2]
There are many kinds of robots. I intentionally selected here two extremes here to show how you can't make generalisation like that.
1: https://en.missilery.info/missile/javelin 2: https://www.nauticexpo.com/prod/konecranes/product-30447-521...
They are still computationally limited in how fast they can process new terrain, course-find, object detect, etc.
If their sensors only saw that object 10ms ago, they probably don’t have an idea what it is yet. At speed, 10ms can be a long time.
60mph == 88 feet per second == .8 feet (about 10 inches) per 10 milliseconds.
10 inches matters a lot when you’re on terrain, or even on a road.
They can move as fast as they have been designed to move, if further computation is not required.
In most cases (repetitive work, stuff with relatively basic, straightforward computation - the human equivalent of reflexes), that means they’re basically constrained by their physical limits.
Those fast object detection algorithms require an extremely artificial environment (with high contrast backgrounds, tuned high intensity light), and are generally only good at sorting specific classes of items, at least last I researched them.
The algorithms involve do a LOT of down sampling too, if I remember.
They aren’t general purpose navigation or object detection algorithms, I’m trying to say, and won’t work at that speed in anything but a very controlled environment. It’s the problem that a lot of players have been dumping billions into for decades - Waymo, DJI, Tesla, and a million smaller names.
General purpose navigation in an unknown environment starts running up against real hardware limits right now at speed, even with essentially unlimited power budgets (aka a car, truck, drone with a huge battery and limited flight time, etc). Based on the context, I understood the question to be in reference to something portable, and most of those hardware implementations would be nearly impossible to fit in a humanoid style robot.
It’s going to take a decade or so for that to change probably, maybe a half that, depending on what Skydio and DJI have cooking.