Support NeoGAF

[Biff]

Uber abandons effort to develop own self-driving vehicle

Car-booking app transfers its technology at a marked-down price to rival Aurora

www.ft.com

Uber has abandoned efforts to develop its own self-driving car and will instead swap its operations for a minority stake in Aurora, a driverless vehicle start-up backed by Amazon and Sequoia, at a significantly marked-down valuation.

The deal brings to an end one of the most ambitious attempts to develop a fully autonomous vehicle, in which Uber leapfrogged its rivals in pursuit of getting a fleet of self-driving taxis on the road.

However, its efforts were marred by tragedy when a woman was killed in an accident involving one of the cars in Tempe, Arizona, in 2018, and Uber’s investors have pressed the company to focus on getting its core car-booking business to profitability.

Uber’s self-driving business was a significant cash drain for the company, but was valued at $7.25bn as recently as April 2019, when Toyota and SoftBank took minority stakes in the unit ahead of the group’s initial public offering.

Eric Meyhofer, the head of the self-driving unit, will not be joining Aurora and will leave Uber.

Aurora — founded by a trio of executives who played key roles founding the self-driving programmes at Google, Tesla and Uber — is known for its strong financial backing. But its strategy of working closely with top carmakers has faltered, as early partners Volkswagen, Hyundai and Fiat Chrysler all switched to backing technology from Aurora’s rivals.

In 2016, after Travis Kalanick launched the division, the Uber co-founder called the self-driving challenge “basically existential for us”. The fear was that Google could undermine its entire business with a cheaper, safer ride-hailing business.

Uber acted quickly and in 2016 invested $680m in driverless start-up Otto. It projected it would have 75,000 autonomous vehicles on roads by 2019.

But its early hopes faded and the unit instead became a public-relations nightmare after the accident in Tempe. This year the self-driving unit and “other technology programs” lost more than $300m on an adjusted ebitda basis.

I find this quite shocking given Uber's entire vision was based on using expensive cash-burning human drivers as a bridge until autonomous vehicles were ready, ushering in Uber's global domination of the transport industry.

Oops?

 

[MGHA]

That's uber disappointing.

 

[Pagusas]

ouch, wonder how much R&D $$$ were lost?

 

[iconmaster]

Yup, that’s pretty significant. I honestly didn’t think the fatality was going to hurt them as much as it should. Seems it did.

ouch, wonder how much R&D $$$ were lost?

From Wired:

“Uber ATG lost $303 million between January and September of this year, according to financial filings, and the company has spent more than $1 billion in its five years of existence.

 

[Majima Everywhere System]

I prefer the barely driving now over the self driving future.

 

[TheMan]

Not surprising. I think it’s probably relatively easy to make a car that can follow simple rules in controlled environment.

The real world? Idiots on their phones. Construction zones. Newly painted roads with missing lane markers or weird temporary lanes. Infinite variations and combinations of these issues and more. I just don’t think its going to be possible to make an AI that is capable of navigating these scenarios for a loooong time

 

[Cunth]

They were chasing a fantasy

 

[Super Mario]

One person dies two years ago vs how many distracted car accidents daily?

Keep on journalizing

 

[diffusionx]

The amount of money Uber spent on this is absolutely insane.

Remember when they partnered with Carnegie Mellon and then just hired away all the scientists and told CM to go screw?

I find this quite shocking given Uber's entire vision was based on using expensive cash-burning human drivers as a bridge until autonomous vehicles were ready, ushering in Uber's global domination of the transport industry.

I said this years ago - that I don't see how maintaining a worldwide fleet of robot cars would necessarily be cheaper in the end.

 

[cryptoadam]

Seems more like the self driving market is a tougher nut to crack.

Enhanced cruise control seems like something we got, but fully self driving cars is probably still many many years away.

 

[Grinchy]

So what happens when this new start-up company's driverless vehicle kills someone? How is taking a stake in another company supposed to fix that problem if it happens again?

The goal is always going to be driverless vehicles anyway. I just don't think that one random crash from 2018 is the real reason they made the move.

 

[Flying Toaster]

I was waiting for this to happen. They were literally burning through hundreds of millions since inception of self driving as the main goal. There was only a matter of time that the company was going to run out before the goal was realized. Self driving is so close but so far away I am curious who will get there first now.

 

[Loki]

Not surprising. I think it’s probably relatively easy to make a car that can follow simple rules in controlled environment.

The real world? Idiots on their phones. Construction zones. Newly painted roads with missing lane markers or weird temporary lanes. Infinite variations and combinations of these issues and more. I just don’t think its going to be possible to make an AI that is capable of navigating these scenarios for a loooong time

I’ve been saying the same for many years now. The notion that AI will be able to read and react appropriately to all possible situations which exist in urban environments within the next 5-10 years is a joke.

 

[Rentahamster]

I appreciate their effort. Technology progression is hampered without the risk taking R&D contributions from startups like Uber. Cutting their losses and pivoting to back those who are doing it better might be in their best interest now. The market at work.

 

[Slings and Arrows]

Making a self-driving car is a big waste of money.

It would be more cost-effective to 'drone' people to and from locations, safely and effectively. Flying in the air is easier to navigate than roads.

 

[diffusionx]

So what happens when this new start-up company's driverless vehicle kills someone? How is taking a stake in another company supposed to fix that problem if it happens again?

The goal is always going to be driverless vehicles anyway. I just don't think that one random crash from 2018 is the real reason they made the move.

These are all problems that people were talking about like... 8-10 years ago, when it was sort of common knowledge that we would have autonomous cars on the road by now.

I think the real issue is quite simple as others ahve pointed out, it is really fucking hard. It is easy to make big gains relatively quickly - hence why we have self-parking cars now, Tesla's system which works well in about 90% of cases (maybe too well, see the dipshits who don't pay attention to the road). Problem is, getting from 90% to 95% is 10x more work than going from 0-90%. Then getting from 95%-99% is 10x as hard as that. And so forth. And for this stuff to rolled out completely you need to get to 100%.

Part of the problem too is that for this stuff to really work it just needs tons of testing, to catch edge cases. But in this situation, when edge cases occur, people die. People criticized Apple when they mostly got out of this game a few years ago but it seems like they made the right call.