Even After $100 Billion, Self-Driving Cars Are Going Nowhere

I'm not so sure new hardware is needed per se.  Maybe, but I'm not sure.  We have quite a bit of hardware both in terms of sensors and raw computing power.

The 'software' side clearly needs further evolution.

That said I'm simply not close enough to the technical elements to have an educated position either way.  That said I wouldn't be surprised if it turns out we're much closer to it sitting here in July of 2023 that you might suspect.

My reasoning for new hardware is the kinds of programming languages that I think could make a real AI driver - Idris for example - seem to be too slow for the kind of hard real time reactions driving requires.

As for how close we are, the shady way this tech is pushed onto the public and the complaints/issues are reflected in China as well. No one seems to be making enough headway and investor confidence has dropped. 

Maybe someone will have a breakthrough, but IMO the fundamental nature of machine learning leads me to suspect it's not possible. I'd also point out that those who would hack these machine learning drivers have not really stepped up yet but that too is inevitable, and machine learning - which [h]as produced these fatalities and accidents already - seems quite vulnerable in this regard.

My reasoning for new hardware is the kinds of programming languages that I think could make a real AI driver - Idris for example - seem to be too slow for the kind of hard real time reactions driving requires.

As for how close we are, the shady way this tech is pushed onto the public and the complaints/issues are reflected in China as well. No one seems to be making enough headway and investor confidence has dropped. 

Maybe someone will have a breakthrough, but IMO the fundamental nature of machine learning leads me to suspect it's not possible. I'd also point out that those who would hack these machine learning drivers have not really stepped up yet but that too is inevitable, and machine learning - which as produced these fatalities and accidents already - seems quite vulnerable in this regard.

I fact checked Idris on the Internet. There’s nothing about this programming language that seperate it from any other programming language in terms of hardware requirements. The only paradigm breaking kind of computation is quantum computing which requires yet to be invented hardware.

I fact checked Idris on the Internet. There’s nothing about this programming language that seperate it from any other programming language in terms of hardware requirements. The only paradigm breaking kind of computation is quantum computing which requires yet to be invented hardware.

It's slower than what you'd want for hard real time systems. That was my only point.

SF Says Driverless Car Problems ‘Skyrocketing,’ Techies Fume at Our Lack of Enthusiasm for the Glitches

Joe Kukura

The SFMTA says that dangerous incidents with self-driving robotaxis have increased exponentially in recent months, while Cruise, Waymo, and their associated venture capitalist types excoriate SF for not adequately loving the buggy self-driving cars.

But are SF city officials “publicly mischaracterizing” data they do not even have? The rub here appears to be “incidents” versus “collisions.” That is, public officials are concerned with blocked traffic, first responders being obstructed, or other road hazards. Cruise and Waymo would prefer we only consider collisions between two vehicles.

How out to lunch does Tan get on this rant? At one point he says, “There’s a reason why Star Trek placed Starfleet Command in San Francisco, and it’s because dreamers and doers are here to create things like self-driving cars.” I don’t know who needs to hear this, but Star Trek and Starfleet Command are fictional constructs that do not actually exist in reality.

Tan does not disclose in the video that his company Y Combinator absolutely has a financial stake in this game.

It's slower than what you'd want for hard real time systems. That was my only point.

This may be. But the software isn’t ready yet either. I still don’t understand that self-driving can be legal anywhere either.

Cruise blames Outside Lands for driverless car traffic fiasco in San Francisco

Laya Neelakandan, Ricardo Cano

UPDATE: Supervisor Aaron Peskin said Sunday that government agencies would ask the city attorney to file a petition requesting that the state revisit last week’s key approval expanding robotaxi service.

As many as 10 Cruise driverless cars stopped working in San Francisco’s North Beach on Friday night, causing traffic to back up and leaving some questioning the decision of state regulators a day earlier to approve the expanded use of robotaxis in the city.

The autonomous vehicles appeared to be stopped in the middle of Grant Avenue, according to social media posts, with hazard lights on, blocking other cars from moving. 

In a response to the incident, Cruise said the backup was caused by “wireless connectivity issues” that immobilized the driverless cars. San Francisco police confirmed that the cell connectivity issues were caused by the large number of people at the nearby Outside Lands music festival overtaxing the system.

“We are actively investigating and working on solutions to prevent this from happening again and apologize to those impacted,” Cruise said in a statement.

Cruise blames Outside Lands for driverless car traffic fiasco in San Francisco

Laya Neelakandan, Ricardo Cano

Cruise Agrees to Reduce Driverless Car Fleet in San Francisco After Crash

Yiwen Lu

On Friday, the California Department of Motor Vehicles, which regulates the safety of the driverless cars, asked Cruise to halve the number of vehicles it was operating in San Francisco. The Cruise vehicle’s collision with a fire truck the day before had injured a passenger in the driverless car. Earlier in the week, another Cruise vehicle got stuck in newly poured concrete on another city street.

I guess nobody had thought to include pictures of wet concrete in the images used to train the AI!

David

Well team:  I purchased a 2023 Model Y a few weeks ago.  Have to report that so far I simply love the car.

I did NOT purchase the "Full Self Driving" module but the car comes standard with a quasi-version of autopilot.  Basically, it has adaptive cruise control combined with auto-steering lane maintenance.  Its really wonderful on longer highway drives (which I do with some regularity).  Someone put it this way to me: "Teslas are technology with a car wrapped around it.  Other automakers make cars with technology wrapped around them."  That resonates with me.

I did get a loaner after having a couple of minor delivery items addressed.  The loaner (a Model 3) DID have FSD.  So, I told the car to drive me home.  It was pretty remarkable.  Changed lanes, stopped for lights, etc.  It basically did drive me home.   However, and its a BIG however, the car did get confused on a left turn (across traffic here in the states) at a traffic light.  The light went yellow and then red while the car was a teeny bit extended into the intersection.  That was well and good.  However, a couple seconds later after cross traffic was fully active it decided NOW was the time to make the left turn.  Needless to say I retook control of the vehicle and stopped its movement.  This would have clearly been an accident.

So, to Sci's point..... Tesla's FSD isn't quite ready for prime time.  That said it sure felt 'close' to me but I'll concede the proverbial last mile is probably the hardest part by a big factor.

So, to Sci's point..... Tesla's FSD isn't quite ready for prime time.  That said it sure felt 'close' to me but I'll concede the proverbial last mile is probably the hardest part by a big factor.

I'd say the challenge is that very few people - if anyone - can give you a clear measure of how close self-driving is to completion.

Personally I'd say it's like Russian Roulette, and it's a good thing you were paying attention and took control.