HN user

hacoo

165 karma
Posts0
Comments69
View on HN
No posts found.

Yep, some incidents may be caused by human drivers being extra aggressive around the waymo. I work in this industry and this definitely happens. That said, weird / unexpected braking is a frequent problem for self driving cars too.

It is common for practitioners of Muai Thai and other kickboxing styles to harden their shins by whacking them with hard objects, or rolling something hard like a bottle across the bone. Sparring and hitting the heavy bag also harden your bones over time.

Kicks are normally blocked with the lower leg in Muai Thai. This often results in shin-on-shin contact which can be very painful for both fighters. So the point isn't so much to prevent full bone breaks (which are very rare) as to toughen the bone enough to handle this abuse... and make sure it hurts more for the other guy than it does for you.

Some background here: in the autonomous vehicles world, it's common to split things up into 'perception' (detecting things around the car and creating a world model), 'planning' (deciding what to do), and 'controls' (giving gas, brake, steering etc so the car smoothly follows the plan). Saying telsa is using AI 'controls' here is probably a misnomer, Musk's statement indicates they are moving the entire planning stack to a NN.

'End to end' is also a bit misleading, as this usually indicates that perception and planning are fused into one network. They aren't doing this (which is probably a good idea), it's only 'end to end' from the planning and possibly controls side.

True but, angle grinders make a lot if noise and very conspicuous sparks. Of course thieves will still do it if they can get away with it, but it’s not a very quick or stealthy method.

Picking the lock, in the other hand, is pretty fast and won’t attract much attention. Thief just needs to be sophisticated enough to have the correct picking tool and practice with it a bit.

The idea is that remote operators help get the car unstuck, but the car should always be able to avoid collisions on its own. I think these cars are supposed to pull over as well instead of blocking traffic.

Key word here being “supposed to”, of course problems occur all the time. Also once a car gets stuck, pulling over is hard because… it’s stuck.

It’s not hard, use a tow truck. But getting a tow truck to every incident (especially in the bay area, not know for functional towing services) is the hard part.

Since there’s no driver present and the controls are presumably locked, you cant push the car to the side of the road which is what would normally happen with a stalled car.

Technically I don’t think it’s resold, as google hasn’t bought the data from anyone. But I think what the author meant is that Google also uses the data for ads, etc and then sells it.

It’s actually much easier to remotely pilot a drone than a car for a few reasons:

- Cars operate very close to obstacles, at any given time, you are only seconds from hitting something. This means you need extremely high reliability and low latency. Planes have a lot more time to recover from communication loss before hitting the ground.

- Planes can circle in the air safely if they lose communications. Cars have to stop and then block traffic.

- In the air there are no buildings, mountains, tunnels, etc to block signal.

- The military has resources to set up and fully control its own communications system. Rolling out such a system on all public roads would be a huge undertaking. Existing networks (LTE) are not designed for remote vehicle operation.

So basically, while you CAN drive a car over LTE, there are a ton of edge cases that make it pretty unsafe. As mentioned above, these companies do have remote operators to unstuck cars, but they don’t drive them directly.

Waymo and Cruise both have remote operators who can remotely unstuck cars.

AFAIK they can’t drive the cars directly, but the they can give override commands, reset software, or make the car follow a manually set path. They probably can’t make it reverse at full tilt though ;)

Directly driving cars is considered very sketchy as you could have network latency spikes or dropouts at any time.

That said, it’s possible the cone still prevents the car from moving at all because either:

a. It creates a large enough sensor blockage to be deemed ‘unsafe’ for any movement

Or b. The car thinks its been in a collision, or thinks it can’t move without ‘colliding’ with the cone further.

Not a metallurgist but — my understanding is that pure iron weapons are quite inferior to bronze. Pure iron is very soft and ductile. It would not hold an edge well and is easily bent - to the point that some sources say they would need to be repeatedly un-bent during battle.

Another advantage of bronze weapons is they could be cast in a mold instead of forged.

Iron was much more plentiful than copper and tin however (especially as they were mined out), which was the main reason bronze fell out of use.

Good steel was superior to both, but much harder to make.

Legged-robot technology is still very immature, even more so when the rover was designed. Wheels work well on relatively flat Martian terrain and are a lot less likely to break than robot legs.

Interestingly the latest Mars rover also includes a small helicopter, another technology which requires spinning something on a bearing and does not commonly exist in nature.

I agree with this. Screen sharing is a very effective way to transfer knowledge — I have had a lot of success by hopping onto a screen share and walking others through things. There is lots of room for questions and they can record/review the meeting too.

However, this requires a culture where impromptu calls are expected and normalized. The more friction there is from scheduling, waiting for people to respond, etc the less likely these kind of fast information transfer sessions are to happen.

Tldr, I think people can learn very effectively in a remote environment, it just requires that people put the same kind of time into communicating that they would in an office.

[dead] 3 years ago

How did they control for non-racial differences in the bots? In the thumbnail the Black bot looks more trustworthy (to me) because the white bot looks like a creepy baby.