roblab
Active Member
So much excitement because company A is a few months more advanced than company B. Y'know, at 74, it just doesn't make all that much difference.
-
Want to remove ads? Register an account and login to see fewer ads, and become a Supporting Member to remove almost all ads.
Mobileye's Self Driving Car is unbelievably good
- Thread starter Bladerskb
- Start date
anticitizen13.7
Not posting at TMC after 9/17/2018
In the full context of all your posts in this thread, is how. Your original post says:
Tesla is not the only competition for Mobileye, not by a long shot.
As a former MBLY shareholder, I had and still have very high regard for their technology. However, that demo does not necessarily make them vastly more ahead than Waymo or Tesla.
You seem to be hard over and trying to get a rise. I’m not putting down what Mobileye showed but I’m not dismissing understanding intersections either. I found a Mobileye presentation from a couple of years ago much more impressive. In particular they recognized some complex traffic light configurations.
It will be interesting to see how the specialized processor approach of Mobileye plays out compared to the more general neural network approach of Tesla. Neural nets initially develop slowly but have the potential to perform better in the end.
incorrect. Tesla doesn't use a more general neural network. This is again another fable and myth that surrounds Tesla. Everyone uses deep learning, everyone uses convulutional neural net. Tesla system is basically a copy of what everyone else is doing. At one point Tesla was literally using GoogleNet model from 2012. The only one actually doing something different is Mobileye who are using reinforcement learning quite heavily and its REM mapping.
Tesla reality is quite different than the fiction.
Look at the Mobileye chip architecture and tell me they are general purpose and neural net based. The last time I looked at their die it was segmented into very targeted processing units.
anticitizen13.7
Not posting at TMC after 9/17/2018
Who cares?
What this is really about is you wanting people to believe that you are “right”. It has always been about this.
Oh, i think they have a fit for purpose neural net capacity.
https://www.eetindia.co.in/news/article/a-peek-inside-mobileyes-eyeq5-part-2
between their programmable macro array and their SIMD, i suspect Mobileye has very potent hardware, in need of very potent software.
will they deliver? time will tell
Except it isn't a product.
It's just another company-curated demo, just like the one Tesla did back in 2016 and Waymo did a year later.
Provided that you can restrict the orbit of the vehicles within a pre-scanned area (i.e. a taxi service) it should work fine.
But it would seem that there can be unforeseen conflicts even then.
You could well be right about this, but not just Tesla's; everyones.
israndy
Supercharger Hunter
Like we all are saying NONE of this matters until you have a product, and no product will be of any use until you have legislation. I don't know how far legislation will get if the car cannot communicate with its surroundings. I saw the MobilEye guys say their car uses aggressive driving to communicate. Sure, that's a strategy. If, however, you are at a crosswalk I worry about that communication being lost on a dog.
I liked the signs bolted to the DriveAI cars for communicating intentions. I wonder what Tesla will do to indicate "No, you go ahead" to those around it. Also, how are they going to hear a siren w/o ears? Seems we ALL have a long way to go, and if the MP3 market showed us anything, it was that after 10+ manufacturers come to market with their solutions, Apple will release the ultimate gadget and take over the market.
-Randy
I liked the signs bolted to the DriveAI cars for communicating intentions. I wonder what Tesla will do to indicate "No, you go ahead" to those around it. Also, how are they going to hear a siren w/o ears? Seems we ALL have a long way to go, and if the MP3 market showed us anything, it was that after 10+ manufacturers come to market with their solutions, Apple will release the ultimate gadget and take over the market.
-Randy
In the ongoing us vs them.
When it comes to LIDAR vs RADAR and cameras, what you are referring to is object detection, identification, velocity, and positioning. This means that you are detecting a bicycle at 45 feet heading at 235 degrees at 5 mph.
That's what I think that Elon indicates when he talks about LIDAR vs cameras.
After you find out what's around you, the LIDAR debate no matter enters the equation. Now it is up to the computers to determine if the object is going to cause the vehicle to deviate from current path or even where the current path is. That's the real "teaching the car how to drive" part. That's generally what you have to teach the 16 year old new driver. Yes, you look both ways before crossing the street, but when you look, what are you looking for?
Think about it, as you drive to work, take a 1 second mental picture and then think about what you had to think about to analyze the scene. So much of it comes naturally to us. Simple things like what is the road and what isn't comes so naturally, but it pretty intensive when you try to compute it. And then think about so many construction site where you really have to study the picture.
When it comes to LIDAR vs RADAR and cameras, what you are referring to is object detection, identification, velocity, and positioning. This means that you are detecting a bicycle at 45 feet heading at 235 degrees at 5 mph.
That's what I think that Elon indicates when he talks about LIDAR vs cameras.
After you find out what's around you, the LIDAR debate no matter enters the equation. Now it is up to the computers to determine if the object is going to cause the vehicle to deviate from current path or even where the current path is. That's the real "teaching the car how to drive" part. That's generally what you have to teach the 16 year old new driver. Yes, you look both ways before crossing the street, but when you look, what are you looking for?
Think about it, as you drive to work, take a 1 second mental picture and then think about what you had to think about to analyze the scene. So much of it comes naturally to us. Simple things like what is the road and what isn't comes so naturally, but it pretty intensive when you try to compute it. And then think about so many construction site where you really have to study the picture.
Anyone else getting a little drained out on demos?
It was fun 1-2 years ago, but it feels like that's all there is.
That we're just stuck in this L2 land where nothing changes. Anything autonomous is fleet vehicles in extremely locked down routes/environments.
It's become one of those hyped things, and we're supposedly in the AI winter.
We had a brief amount of hope from Audi with their L3 A8, but then they decided against introducing it in the US for the foreseeable future.
It was fun 1-2 years ago, but it feels like that's all there is.
That we're just stuck in this L2 land where nothing changes. Anything autonomous is fleet vehicles in extremely locked down routes/environments.
It's become one of those hyped things, and we're supposedly in the AI winter.
We had a brief amount of hope from Audi with their L3 A8, but then they decided against introducing it in the US for the foreseeable future.
Because it really sucks and is the most brittle and restricted ADAS I've read about. It seems we are on the cusp of some pretty exciting stuff thanks to Tesla. I actually now have heard things that indicate Tesla might actually be readying themselves to deliver something much better than the current offerings (EAP -- here we come (finally))
Similar threads
