Roundtable Discussion; The Future of Mineral Sands. Watch the video here.
I think many would agree that collecting the data to train autonomous vehicles is time consuming, expensive and requires data from further afield than just SV.
It's also important that the drivers in these vehicles are not distracted and ready to take over.
My prediction is an OEM is going to discover, or already has, that they can collect far more data by selling level 3 cars to consumers with DMS as a key part. They make money selling vehicles and collect masses of data on how the level 3 features are working, in various conditions and locations. And given the driver is being monitored, and responsible, the data on when they decide/have to take over from the machine is invaluable.
Privacy concerns should be reduced/removed because in this scenario footage of the driver is not important. What is important is when/how quickly the human reacts to then help train and what is going on around the vehicle.
This also has the added safety benefit that whether the human is driving or not the machine is ensuring thay they are not distracted.
In the Guardian interview Ken mentioned Guardian has both a driver facing and externally facing camera.
The recent accouncement that Guardian has travelled 1.3b km, I don't think can be interpreted as 1.3bn km of footage of the driver. As we know Guardian only records the driver during the period before/after an incident.
But... Does anyone know if the external facing camera is constantly recording? There would be no privacy concerns for the driver in this case. If so, does that mean we have roughly 1.3bn km of footage? From a number of different countries, all with different road conditions.
Couple that with the telematics (I hope See have this too), about how these drivers have steered/braked/accelerated.
If See have all the above (I really hope so) someone like Waymo would bite their left and right arms off to get that.
Synced internal + external footage, given the number of miles they have. That's a huge dataset. Far more valuable than just driving facing footage.
Now if they had more than just the before/after period of fatigue/distraction events. Where the drivers eyes were looking, how they reacted to what was in front of them, and the footage of what was around the vehicle. That would be incredible data to train a self driving car.
It reminded me of
https://www.xkcd.com/2030/
Replace Electronic Voting with Self Driving...
...And I'm a software developer
Apple purchased an eye tracking company last year
https://techcrunch.com/2017/06/26/apple-acquires-smi-eye-tracking-company/
What they won't necessarily have is the data around distracted driving and fatigue events
I'm currently sat on a train, travelling on a fixed track with signals managed and controller by a control center somewhere in the country.
The driver is responsible for maintaining speed, stopping at stations and signals and looking for potential problems ahead to perform an emergency stop.
If AI is advanced enough for cars to driver autonomously, then why is my train, with many less variables to deal with, still controlled by a human (let's ignore the issue is likely with unions stating job losses and "safety" concerns).
If trains are argued that they still need a driver "for safety" reasons then so does a car IMHO.
Power consumption would need to get to appropriate levels before making it into a mobile device. So R&D and maybe v3 of the chip?
I've said before that I think giving developers access to this kind of data would lead to some very innovative apps/games being written. And a USP for Apple/Samsung.
First time I've seen a reference to "multi-subject" since the Trading Update last August
http://www.lse.co.uk/share-regulatory-news.asp?shareprice=SEE&ArticleCode=1zq3ujuw&ArticleHeadline=Trading_Update
Who knows. Let's wait and see for more information
I have found that Byton has Bosch as a Tier 1 supplier. So if it is them, it may be unrelated to the Veoneer talk of China.
https://valueinnovations.com/byton-launching-suv-ev-ces-2018/
What would make me think it's not Geely is the projected contract size. Would I be wrong in assuming a contract with Geely would be higher.
And Byton are targetting mass production by q4 2019
https://techcrunch.com/2018/06/11/chinese-electric-car-startup-byton-raises-500-million/
Byton are targetting level 4 by end of 2020
http://uk.businessinsider.com/byton-electric-cars-focusing-on-tech-to-compete-2018-7
If it is them (a big if) , then it's a huge endorsement that there is still value and a need in DMS while there is the possibility that a human may be required to takeover in some scenarios.
I would imagine they would have looked at Tesla, and their incidents, and decided that hands on wheel isn't the safest solution
The mentions of a remote monitor who can intervene, along with an "off the shelf" solution makes me think Guardian and Guardian Live.
I don't think this will be a shareprice rocket. But a few hundred sales of guardian wouldnt be bad. And it would, if is see, be a good reinforcement of the see technology for DMS
"If the system notices a driver who isn't paying attention, an audio alert sounds and the system also contacts a remote monitor who can intervene (although Uber wasn't specific on exactly what the remote monitors could do aside from taking "appropriate action")."
https://www.engadget.com/2018/07/24/uber-self-driving-tests-pittsburgh-manual-mode/
Can any of our competitors do that out of the box?
I don't believe level 5 will be around for a long while yet (2030+) if ever. Level 4 is my guess for where it will end. There will be geofenced areas where steering wheelless robo taxis exist, that depending on weather conditions will function. I think companies will start to realise this, e.g. Waymo
But what about training cars. The more data available to train a car the sooner level 5 will come (if ever). Having a few hundred cars driving around isn't going to collect the same amount of data that a mass produced fleet of vehicles could collect worldwide.
So what if the race changed slightly. Get level 3 into cars people pay for to collect the data on your behalf. The number of training situations is exponential.
However if you're selling a car to do this you need to make sure the driver is still alert and paying attention. Therefore DMS is important.
They won't collect visual data regarding the driver from the camera, so privacy concerns are satisfied, but from how the driver reacts in deciding to takeover.
And I still don't think level 5 will come for more than a decade, if ever.
A video posted 4 days ago by See with Coach USA
https://youtu.be/bQsdO0waO0c
https://phys.org/news/2018-07-quantum-dot-white-efficiency.html