Roundtable Discussion; The Future of Mineral Sands. Watch the video here.
Article 11
Specific requirements relating to automated vehicles
1. In addition to the other requirements of this Regulation and of the delegated acts adopted
pursuant to it that are applicable to vehicles of the respective categories, automated vehicles
shall comply with the requirements set out in the delegated acts adopted under paragraph 2
relating to:
(a) systems to replace the driver’s control of the vehicle, including steering, accelerating
and braking;
(b) systems to provide the vehicle with real-time information on the state of the vehicle
and the surrounding area;
(c) driver readiness monitoring systems;
"Don't worry that rural areas will be left out. A vehicle could be parked in every village waiting for your order to come."
What I don't get about this 'on demand autonomous car' future is that there would still need to be 1 vehicle for every commuter who currently has a car so they can all get to work.
Two solutions I can think of to this - ride sharing and surcharging. Ride sharing requires a willingness to accept a longer journey and/or a pre-booked departure time. I call this a bus.
At the end we're still left with tens of thousands of cars sat around doing nothing all day whilst everyone works.
I don't think the BBC journalist has stepped outside a big city. I'd agree these 'hailing' services makes sense in cities, since the public transport tends to be mostly there - cars would be used for other purposes than commuting. But in the towns and villages, where commuting doesn't have the public transport infrastructure, i'm a skeptic. Even if the autonomous technology was there.
I can't make this out. Alot of noise has been coming from SEE recently around Guardian, but from my understanding SEE won't be focusing on or selling the G2 units in the next year?
And today they tweeted a new video about Guardian
Check out @seeingmachines’s Tweet: https://twitter.com/seeingmachines/status/1044931446761295872?s=09
Theories?
Bosch has recently made a strategic investment into Ceres for their fuel cells.
STAR have recently partnered with Bosch and the LSE board is hopeful a similar strategic investment will come their way.
Semicast calls Bosch out as a potential to take on the Fleet aftermarket. Could this lead them to become a strategic investor in SEE as part of any deal?
The Volvo order starts in 2019
https://www.engadget.com/2017/11/20/uber-order-24000-volvos-self-driving/
So if Uber the order so far would be for around 200 units, but a decent increase from 2019 through 2021
We discussed this back in July when Techcrunch/Engadget etc first reported on Uber using driver monitoring in their new testing.
Back in May they had 200 cars on the road. With 24000 volvos ordered. How many do they have now?
https://www.techradar.com/uk/news/uber-self-driving-cars
Complete self-driving hardware and software development kit available Oct. 1.
https://blogs.nvidia.com/blog/2018/09/12/nvidia-drive-agx-devkit-autonomous-driving/?linkId=100000003496682
These advanced software capabilities can also be utilized inside the car. Using a driver-facing camera, applications built on the DRIVE IX SDK can track the driver’s facial expression to know whether they’re drowsy or paying attention to the road.
Is there more reward for SEE going after Fovio being built in to newly produced fleet vehicles, possibly required due to mandation of DMS for these vehicles, instead of continuing their focusing on the after market?
Fovio wouldn't need network capabilities. Somewhere along the line the data collected from Fovio will need to be made available to the OEMs software via an Api. Similar to Seeing Machines old Faceapi*.
This will provide to the rest of the car software with where the driver is looking (to help control HUDs, windscreen wipers and whatever the OEMs come up with), but also whether the Fovio chip believes the driver is fatigued/microsleep/distracted (this magic secret algorithm).
I would imagine SEE and the OEMs would have some kind of licensing agreement on what else, and where else, they can use the data. Giving the OEM the ability to pay more to use the data, or more features of the chip, for other purposes whilst protecting SEEs moat.
Apple, for example, in building a car will be looking at it from their experience in building consumer electronics. The car is a computer, and Apple like Google, like data. They'll be wanting to, and be more willing/comfortable in, collecting as much data from the car as possible than perhaps Ford etc...
Apple will want to innovate with what a car is. This will involve a huge amount of data collection from both inside and outside the vehicle. They will want to monetize their car via their AppStore and this will involve exposing features of the car, including any driver facing camera, to developers.
Google/Waymo will probably go this route too. They'll soon realise level5 is a dream for a more distant future and instead they can make more money in the meantime, whilst collecting masses of useful data to help get there, by focusing on the software/experience of the car. Designing in partnership with, or acquiring, the existing OEMs to build the vehicles.
Privacy will be a concern for some, but these SV tech companies know how to make people willing to give their data away. And consumers will tick yes to the EULA, without reading it, as part their car purchase or when they're prompted to use certain software.
I'm in no doubt these vehicles will be mass data collection machines, and would expect them to collect where the driver has their focus and attention along with the external state of the vehicle. I don't think this will include the actual video footage of the driver or surroundings. Regulators won't allow that, and that would relieve the biggest privacy concern with having a camera in the cab**. So in this regard Cenkos is right and Guardian/Fleet will continue to be the only source of the most detailed training data available.
* http://www.ekstremmakina.com/EKSTREM/product/faceapi/downloads/index.html
** what's been the reaction to the camera in the new Tesla?
- knowing they're training autonomous vehicles
- insurance reductions because they've opted to have their data recorded, in the event of an accident that data is available to them
- self improvement with feedback through an app at the end of their journey on how well they drove and paid attention.
- sharing data with a family member (parents wanting to keep a tab on their 17 year old with a new license)
- creating "AI" journey directions of which roads are going fast/slow at that moment based off telematics of other cars on those roads.
- an in car "app" that detects potholes and reports them automatically to the relevant authority
All require collecting various data from either inside/outside the car that provide value to the consumer in some way. Once that data is consented to be collected and stored in the cloud its available to Google/Apple to help train autonomous vehicles.
Cenkos is right, the data that is collected by Guardian (video footage of the driver) will not be accept by consumers.
But with regulation to prevent that, which I feel is important in an auto DMS case, I don't think the same concerns will be there in the collection of what would effectively be a set of coordinates of eye and head position.
Likewise actual footage of around the car isn't necessary.
Of course people will opt out, but if there is value to the consumer by opting in people will.
"no private driver will agree to their data being shared."
I disagree with this. Everyone walks around with a mobile phone. Google/Apple/Microsoft/Facebook know more about us than we realise, whether for our good or bad.
People need to get some value back from sharing this data though otherwise they won't.
But people will share it.
I don't disagree that there wouldn't be privacy concerns, but it would be down to regulation to ensure internal footage of the driver doesn't get recorded or leave the vehicle. This is just as important for camera based DMS though, ignoring what I mentioned previously.
At some point cars will be collecting masses of data to send back, and I think it will include where eyes are looking without footage of the driver, with consent required in some form.
You're right, it would be daft to say you can't have feature x unless you hand over data.
I don't believe GDPR would come into play if all that was being collected was coordinates and direction of focus that the driver was looking. Assuming this wasn't combined with any unique identifier of that vehicle, there is no sensitive or personally identifiable information being recorded.
Saying that consent may be required as part of the terms of using the autonomous functionality of the vehicle.
Privacy issues come around a camera being placed in the car, whether used in a scenario to help train an autonomous vehicle or purely for DMS.
Regulations (government or industry) stating that footage of the driver is not stored or leaves the vehicle need to be made to alleviate privacy concerns.
Collecting data of where the driver is looking can be collected anonymously. Footage of the driver isn't necessary. Just the positioning of their eyes and direction of focus.
Basically just the data that drives the overlay we see in the Guardian demos. There's no need to record real imagery of the driver. Saving on both local storage costs and data transmission
Imagine Toyota selling 10m vehicles every year with some form of autonomous functionality. Collecting how well its performing by combining how the monitored and responsible human reacts/trusts its decisions.
How many miles per year, in various situations worldwide, would these 10m vehicles do while in an autonomous mode. How much data would they collect vs. a small fleet of Waymo or Uber cars in comparison?