The requested article has expired, and is no longer available. Any related articles, and user comments are shown below.
© Copyright 2022 The Associated Press. All rights reserved. This material may not be published, broadcast, rewritten or redistributed without permission.China's Baidu races Waymo, GM to develop self-driving cars
By JOE McDONALD BEIJING©2024 GPlusMedia Inc.
12 Comments
Login to comment
Jim
The good news about this is that since China has taken a big leap in self driving vehicles, other countries such as Japan and The US will make more efforts to break through this industry just to compete with China purely out of jealousy! Then we will see some actual developments as far as the self driving vehicles are concerned!
Sh1mon M4sada
No one can catchup to Tesla. The only reason China is pouring money (wasting really) into self driving is to spy using the many sensors on the car.
Eventually every one will give up throwing money into the air and negotiate with Tesla.
Tesla's AI, machine learning, environment learning are just too advanced and are protected with patents that there really is no point for others to be developing second best tech.
Samit Basu
@Jim
Japan is not a viable self-driving competitor.
Desert Tortoise
If you have ever driven in Shanghai you will get this question, but did Baidu program their cars to make a left turn the very instant the traffic light turns green to beat the oncoming traffic into the intersection? That seems to be standard practice in Shanghai.
Desert Tortoise
Well, at least not when it comes to running into emergency vehicles.
https://www.motorbiscuit.com/teslas-keep-crashing-emergency-vehicles/
Desert Tortoise
Tesla's may be the least advanced. Tesla relies exclusively on cameras. Other systems use LiDAR. Tesla's neural system relies on self learning what pixel's represent. The problems arise when it encounters something unfamiliar and doesn't recognize it as a collision risk. It may not even recognize there is an object in front of it and crash. This has happened already in a fatal collision between a Tesla and a disabled truck in Florida.
Sh1mon M4sada
Tesla started out with radar (lidar) + camera. They just decided lidar often gave false or inconsistent data compared to camera. There were 2 teams in Tesla auto pilot, Tesla just sacked a lot of the lidar team. This is something other brands have not gotten to 'YET', because they don't have AI/neural net.
Other brands are programming every possible situation as a path to self driving. Tesla has AI deciding, on the fly.
Have a view of the Tesla Owners club interview with Elon Musk, all layed out there in minute details, especially how neural net progress was delayed by lidar.
Yrral
Most driverless cars cannot pass a test in America,their are certain turn a car has to make while flowing through traffic is such as right on red
Desert Tortoise
@Sh1mon M4sada, I would argue that Tesla's choices are proving themselves inadequate in use. Because the software has not be taught all the possible situations it may encounter on the road Tesla's AI is proving unable to recognize dangers it has not previously encountered. It is not in fact learning enough to keep cars out of collisions. It may turn out that self driving cars only work on well controlled guideways that present driverless cars with a limited number of knowable situations to deal with. The open road with merging traffic, cross traffic, cars passing on two lane roads, rocks, deer, bunnies, etc., it too complex and uncertain for AI to deal with. What happens to your AI managed car when a big truck blows a tire and a big chunk of the tread comes at you? Real life, been there done that, on a motorcycle on an LA freeway. Better think fast and make your move. Same thing when a wheel came off a car on a transition road. I caught it out of the corner of my eye and spiked the brakes just before it could hit the driver side of the car. Instead the wheel skimmed over top of my hood and there was no damage. Or how about when the idiot if front of you on a rainy night looses control and spins out? Will your AI recognize the danger and react appropriately? If you can model every situation and program the vehicle to respond to each situation you might have a chance. Relying on the car's AI to learn on the go might end up as bad as turning a driver loose on an LA freeway with no driver training.
albaleo
My thoughts exactly. In some ways, turning roads into railway lines.
Kyo wa heiwa dayo ne
Well obviously China is competitive with Tesla so that ends that silly announcement
Desert Tortoise
It all feels the same, right up to the moment of collision.