Getting some air, Atlas? [video] | Hacker News Reader

Getting some air, Atlas? [video]

youtube.com

66 pointstlb29 comments

Open articleView on HN

Atlas is free!

https://www.youtube.com/watch?v=lAmfJgqw3oU

"No Disassemble! Number 5 is alive!"

That would have been a very different movie with Atlas in the starring role.

I wonder what proportion of the development efforts go into securing these things.

Imagine what would be happen, if these robots start getting hacked. A virus would no longer be just sitting inside a computer, it would manifest itself in a physical form.

Hardcore viruses/trojans/worms/whatever can create physical implications: disrupting power grids, sabotaging nuclear programs[0], etc.

But, even the basic forms of viruses which hackers will start out with, can cause serious consequences because of the physicality of these robots. And not just advanced robots like BostonDynamics', even basic autonomous robots have a high potential to do physical harm, if hacked.

I somehow feel, people should focus more on securing these things vs empowering them with increased intelligence, agility and sheer strength.

I would like to know, if security of these systems rely on the current state of computers, or do they have a different set of guidelines to abide by, considering that the attack vectors can be vastly different.

With the current computer systems, most attacks are done via the internet. But with these, physical access can exploit a whole different range of vulnerabilities, which can be difficult for the bot the detect. Like, how the old PS/2 hardware keyloggers turned out to be quite difficult to detect[1]. What if someone just sticks a different kind of lens in front of the camera(s)/sensor(s) these systems rely on, which can just break all the math it is doing around visuals.

Does the current state of the art of securing these systems, handle these cases ?

[0] Quora -> http://qr.ae/TUTPJK

[1] https://security.stackexchange.com/questions/79124/detecting...

I imagine what some creative blackhats could do if they rooted Tesla's systems, and could now control a few hundred thousand cars.

The prospect of an OTA update hijacking your car seems to get closer and closer every day.

Daemon by Daniel Suarez is about connected nodes (including automated cars) being taken over by a rogue program.

https://www.amazon.com/DAEMON-Daniel-Suarez/dp/0451228731

I loved this book. The destruction a few thousand remotely controlled cars could wreck is terrifying.

the finale of the first season of Westworld (and the second season of it so far) have your answer.

On one end, we have a super hacker, on the other side, we have Dave. https://imgur.com/gallery/OgWs3

To be fair, if someone switched out the lenses on my glasses I would have a hard time too.

I've rewatched this video maybe 10 times since it came out a few days ago. The jogging part -- it's so hard to believe, still seems like CG but looking forward to seeing this in person. I love this stuff -- so much world-revolutionizing potential captured in this video!

The jogging seems to be much slower than a human would jog (for now), which contributes to the unreal look (I think).

As Elon Musk quipped, in a few years this thing will move so fast you’ll need a strobe to see it. It’s not subject to the human reaction time or mechanical constraints.

It is subject to mechanical constraints AKA laws of physics.

Acceleration is limited by the grip on the ground under its feet. It could slowly accelerate to amazing speeds though.

It could use spikes and/or go on all fours in “fast” mode. It’s whatever you design. My main point though is not mechanics, but the significantly lower reaction time. Given enough compute, the world that seems “real time” to us could appear to be slow as molasses to a robot. I think the natural thing to do would be to make it move at the speed that provides a good balance between accomplishing objectives and energy consumption. I think that this speed is much faster than what we’re used to with humans. For eg military robots that balance could be further tweaked in favor of quickly accomplishing objectives.

Compressed-air rockets. Or H2O2 with a platinum catalyst; "decomposition ... into steam and oxygen" https://en.wikipedia.org/wiki/Hydrogen_peroxide#Propellant

We're spit-balling hero-bots, not nasty kill-bots, riiiight?

Potato, potato.

But then it's not robotics, it's rockets!!! Yay rockets!

Let's invent a new Godwin's law reached when any discussion devolves (upgrades?) into rocket talks.

Me: mutters under his breath, "rockets are robots..." ;-)

I'm kinda curious which Boston Dynamics exec has robots running around in the backyard of his summer home in the Berkshires

Exciting and scary! We live in the time where this if finally a real thing after almost a century of discussion of humanoid robots. I wonder when / if this will become a real commercial thing.

Very soon! https://news.ycombinator.com/item?id=17064664

For me the exciting part died out with the understanding of how locomotion actually works. Now it's almost a solved problem, only requiring some hardware advantages and a lot of (excruciatingly boring) man-hours to implement.

So I'm only left with the scary part. This will hit the streets in less than 5 years. Things are going the exact opposite direction of where they should be for the deployment of such robots to be beneficial to the masses. I don't like this, and I'm not alone.

> Things are going the exact opposite direction of where they should be for the deployment of such robots to be beneficial to the masses

Just going to toss that out with no explanation?

The speed at which these robots are advancing is simply astonishing.

Any predictions on how long until we see a bi-pedal robot in a manufacturing environment?

This was from 2013:

https://www.youtube.com/watch?v=pgaEE27nsQw

I'm wondering what challenges we have solved since then, beside the mechanical implementation.

I think it's a bit disingenuous comparing a simple simulation of a 3d model with dubious physics with the problems involved in designing a real robot.

I'm sorry but clearly you have no idea what you're talking about.

>beside the mechanical implementation.

You say that like that was the easy part.

Reply on news.ycombinator.com