GarryP

Was supposed to be this year, but might be pushed back. Reportedly apple is now looking at it too, but Samsung have head start.

https://www.tomsguide.com/us/samsung...ews-26842.html

ozzieeagle

Bugga, I figure that could well be the final death knell of everyday mail, make a huge impact at least.... Shame it's so far off.

We don't seem to be moving as fast on many fronts as I hoped..... including driverless.

GarryP

Interesting move from Ford, at least in the US market.

They are going to effectively dump normal cars, cutting back to just the Mustang and a crossover. Everything else will be trucks/SUVs/etc. And they are going to move hard to hybrids, and to electric vehicles, with a performance utility in 2020 and 16 battery vehicles by 2022.

And then there's the robotaxis, which obviously work with batteries better than conventional cars. I'd guess many of those 16 vehicles will be targeted at that market - the robotaxi is likely to be classified as a truck. More minivan than sedan.

Things will change dramatically over the next 5 years - and if you didn't think they were serious, they are.

https://techcrunch.com/2018/04/25/fo...-focus-active/

ozzieeagle

Well, I'm waiting and feel like i've been waiting for some kind of progress for the past 3 years or so now.

Slight digression and one I applaud wholeheartedly, Amazon are now delivering directly to parked cars as a proper address now, which is a great idea...not sure how they manage to get them into them automatically and have it lodged as securely delivered. Obviously, they've found a (software) way and it takes away the parcel locker issue for households somewhat.

Beoz

No mention of Robotaxi's but it makes perfect sense to head in the direction they are.

Transportation of goods is a massively growing industry and if they can put fuel efficiency at the forefront of consumer choice, like the airline industry has, then its going to be very interesting for the oil market.

GarryP

It's in the original press release, but not explicitly spelled out. Talk of autonomous vehicles and 16 electric vehicles - but no conventional cars.

Just think of the plants they are going to close in the US. If they have only SUV/crossovers and Mustangs then there are whole classes of engine, drivetrain, etc. that they don't need.

The thing that I think they've realised is that a robotaxi doesn't look like the hatchback or sedan of today - and they are shifting now to be ready for where they expect to be.

It helps that they haven't been particularly profitable on the conventional cars they have been selling. They can get away with higher margins in the SUV market.

It also means that when they do launch the robotaxies, they won't be hurting their own sales as much, which means they can go harder on it without the shareholders screaming.

GarryP

I tended to consider that more rural vehicles (and those that claim to be for appearance sake) would be later to the automation/robotaxi game. The lack of good roads and roadmarkings seemed to need the manual controls etc.

However, research by MIT suggests that there are solutions for the yokels that can make automation a general thing, and probably EVs with it.

https://motherboard.vice.com/en_us/a...ds-without-map

Beoz

Given where Lidar and Photogammetry are today it amazes me that this stuff is still being messed around with at universities.

This crowd from Adelaide are flying the Australian cities on a regular basis. No reason why they couldn't be doing the country side too. There are many other companies flying too. The cameras they are using aren't basics either.

Photogammetry for the existing context. Lidar for the just in time ground level needs.

http://aero3dpro.com.au/Melbourne2017

GarryP

Google I/O kicked off overnight, with the star of the show probably being Google Duplex - an AI that does certain constrained tasks for you, over the phone :


you can see how this is an extension of the Google Home type AI, but if they don't apply it to get rid of the other end of the call, the call centre, I'll be shocked.

They also had an update on Waymo, with video of Google's car, which is driving members of the public around, without a 'safety' driver, and with the people in the back seats....

Shard

Nice work, "Lisa" !

I wonder how much processing power the Waymo vehicles require to run deep nets? Presumably the AI is on board (in the vehicle) and not simply linked by a data network?

GarryP

A lot.

Waymo use intel chips, and there is talk of needing 2.5kW to power the processing. Obviously as they refine algorithms and only deploy processing power when needed; and with new purpose designed chips, that will come down. Probably will end up at the GPU level of processing power/electrical power.

The nVidia chip aimed at this market does 30 trillion operations per second, for 30W of power. Final version will need more, but optimisation will cut the load and Moore's law will eat into the power requirements.

DeadVim

Shard

Pretty cool.

Beoz

Ha ..... well Jason .... you've changed your tune.

https://www.news.com.au/technology/innovation/inventions/australia-needs-to-embrace-automation-or-risk-missing-a-22-trillion-boom/news-story/23b2608dec515e3749601d46bac6143d

I have known Jason for many years and we have had this discussion many times. He never took my point of view. Looks like he's seen the light. Must give him a call over the weekend.

Shard

The report doesn't seem to grasp the concept of recursive learning. Still if Oz can re-train an army of redundant office staff, retail workers and and drivers into AI engineers, more power to them.