Hacker News Reader

Comment by naasking | Hacker News Reader

I don't think that will happen nearly as soon as there AI automation. Navigating intelligently through 3D space with arbitrary obstacles has always proven the more challenging of AI tasks.

View on HN

Disagree. Boston Dynamics has had locomotion solved for a decade. Vision systems have advanced significantly, as well as the human language interface. Both seem to have held back industrial robotics use since widespread deployment of the technology on auto assembly lines. These days, asking the computer to recognize a coffee mug on a dining table is no big deal.

> Boston Dynamics has had locomotion solved for a decade

Locomotion is not navigation, as you can see from the persistent failures of the self driving cars.

It's funny, where you see persistent failure, I see the engineering process at work and making previously unfathomable progress. Amazing what people take for granted.

Being consistently wrong in the estimates of how long it would take, despite the history of failures, does not make the projections reliable. It shows the problem is poorly understood, even now.

> Being consistently wrong in the estimates of how long it would take

I don't know if you've ever organized a group of people... If you want everyone to be together and ready to go by 8:00am, you will have to tell them to be ready by 6:00am. Else you will be late waiting for someone. The consistently wrong estimates you point to are exactly that. Else they wouldn't be consistent.

Again, where you see failure, I see effective management practice. Why do you think that is?

Yeah, good luck getting a robot to plumb or wire an entire house anytime soon.

Great example of https://en.wikipedia.org/wiki/Normalcy_bias

Reply on news.ycombinator.com