Slashdot Mirror
Nvidia Introduces a Computer For Level 5 Autonomous Cars (engadget.com)
From a report: At the center of many of the semi-autonomous cars currently on the road is NVIDIA hardware. Once automakers realized that GPUs could power their latest features, the chipmaker, best known for the graphics cards that make your games look outstanding, became the darling of the car world. But while automakers are still dropping level 2 and sometimes level 3 vehicles into the market, NVIDIA's first AI computer, the NVIDIA Drive PX Pegasus, is apparently capable of level 5 autonomy. That means no pedals, no steering wheel, no need for anyone to ever take control. The new computer delivers 320 trillion operations per second, 10 times more than its predecessor. Before you start squirreling away cash for your own self-driving car, though, NVIDIA's senior director of automotive, Danny Shapiro, notes that it's likely going to be robotaxis that drive us around. In fact, the company said that over 25 of its partners are already working on fully autonomous taxis. The goal with this smaller, more powerful computer is to remove the huge computer arrays that sit in the prototype vehicles of OEMs, startups and any other company that's trying to crack the autonomous car nut.
Why own a thing, when you can pay someone else an exorbitant fee to use theirs temporarily?
This will work out brilliantly for the 0.004% who currently own 80% of the wealth.
An enigma, wrapped in a riddle, shrouded in bacon and cheese
Except the last time I played a 3D game, there were a LOT of 3d objects passing through solid objects if you look very close. In a 3d world it's easy to get around that by making the set path of the moving object be the absolute guide. Not so easy in the real world.
Laws are rules for the court, but merely a bottom bar to hit for life. Think beyond laws in your actions always.
Comment removed based on user account deletion
Unlike a GPU where a memory error or an ALU or register bit flip might result in a 1 frame glitch, or at worst a frozen GPU, requiring a reboot, failures in this hardware will kill people.
I hope they have ECC on everything, and redundancy everywhere -- possibly a space-shuttle like voting system where multiple computers are fed the same input, and if they don't produce the same output, a majority wins approach is taken.
It should also have very detailed logging -- so every decision taken can be traced, so when there is an accident, a proper root cause analysis can be performed, and corrective measures instituted.
NVidia as a company has a great track record for being on the cutting edge of technology -- but no track record at all for making safety critical systems. That cutting edge will cause people to bleed if they don't get this right.
Ian Ameline
HAL!
I'm sorry AC. I'm afraid we can't do that.
I currently pay $250/month for a car payment, $160/month for insurance, and probably $120/month for gas. ($530)
I'll assume I need 3 trips per day (commute plus errand/grocery); for 1 month, that's 90 trips.
Simple math says if the robotaxi charges less than (530 / 90) = $5.89 per trip, then I should ditch my personal vehicle.
Let's see how hungry these transportation as a service companies are.
If you have a level 5 autonomous car, I suggest waiting before you download the latest drivers.
You are welcome on my lawn.
So, a few things:
1) Many video games don't compute paths in realtime. Rather, the set of paths is either precomputed or manually entered by the developers. The game then merely selects between one of the preset paths, without any ability to actually determine its own.
2) Even when they are able to determine their own paths, video game pathing algorithms generally have perfect knowledge. There's no need to do the heavy lifting of recognizing obstacles when you have a perfect awareness of every single one in the entire world.
3) Video games aren't necessarily bound by the rules. Most video games are tuned for fun rather than realism, particularly when it comes to their physics and recognition of the law. For instance, two vastly different cars may brake the same way in a game so that it's easier for the player to get the feel for cars in the game, but in real life, the very same car can brake completely differently based on weather, the condition of the tires, or how loaded with cargo it is, and it's important that an autonomous vehicle understand those differences so that it can drive safely. Likewise, video game cars can ignore traffic signals and the like with little concern for the law, but that's not the case in the real world.
4) Video games cheat. You'll frequently see vehicles in games clip through obstacles that would have caused an accident in the real world, take paths that would have destroyed a real vehicle, or have spontaneous boosts in their speeds as they go through boring parts of the world.
5) Video games are dumb. There are literally tens of thousands of YouTube videos of vehicles doing stupid things in games (e.g. driving themselves off cliffs, driving through the air, driving through walls, rolling over on gentle turns, mowing down pedestrians, etc.), so this is hardly a solved problem even in worlds that we have full control over.
All of which is to say, if video game vehicles are our standard for success, heaven help us all, 'cause we'll all be dead within a week.
Highly Autonomous Limousine?
4wdloop
The self driving cars/autonomous cars/driver assistance systems, I was involved with, run on 4 - 6 ARM Cortex, 160MHz and are mostly idle all the time.
You don't need such absurd computing power for a self driving car.
Cost free eBook I read (by iBook/Kobo/Amazon/ObookO/Gutenberg etc.): "The Green Odyssey" by Philip Jose Farmer.