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

Hey Sci.  Early reports on Tesla's most recent self-driving tech are very positive.  Small sample size of users as it hasn't been rolled out to the broad subscriber base, but they've supposed re-engineered the system from a rules-based traditional computational structure to (you guessed it) an LLM AI structured system.

Keep an eye on it.  Tesla's stock price sure needs a jolt.

Hey Sci.  Early reports on Tesla's most recent self-driving tech are very positive.  Small sample size of users as it hasn't been rolled out to the broad subscriber base, but they've supposed re-engineered the system from a rules-based traditional computational structure to (you guessed it) an LLM AI structured system.

Keep an eye on it.  Tesla's stock price sure needs a jolt.

It looks like it might be good to keep an open mind on this. Tesla and Elon Musk seem to have latched on to what perhaps will be a winning technology for self-driving cars - an approach mimicing large language processors like Chat-GPT except where the data is millions of videos of successful human driving in difficult situations, rather than billions of lines of human-written text on the Internet. It’s like ChatGPT, but for cars. Musk's engineers are using the video data conveniently and rapidly accumulating from almost two million Teslas being driven by human drivers. To be used as training data, the videos have to first be reviewed and pass tests for being good driving responses to the situations.

From https://www.cnbc.com/2023/09/09/ai-for-c...cerpt.html :

Tesla’s Autopilot system had been relying on a rules-based approach. The car’s cameras identified such things as lane markings, pedestrians, vehicles, signs and traffic signals. Then the software applied a set of rules, such as: Stop when the light is red, go when it’s green, stay in the middle of the lane markers, proceed through an intersection only when there are no cars coming fast enough to hit you, and so on. Tesla’s engineers manually wrote and updated hundreds of thousands of lines of C++ code to apply these rules to complex situations.
..................................................
The “neural network planner” that Shroff and others were working on at Tesla took a different approach. “Instead of determining the proper path of the car based on rules,” Shroff says, “we determine the car’s proper path by relying on a neural network that learns from millions of examples of what humans have done.” In other words, it’s human imitation. Faced with a situation, the neural network chooses a path based on what humans have done in thousands of similar situations. It’s like the way humans learn to speak and drive and play chess and eat spaghetti and do almost everything else; we might be given a set of rules to follow, but mainly we pick up the skills by observing how other people do them.
..................................................
By mid-April 2023, it was time for Musk to try the new neural network planner. He sat in the driver’s seat next to Ashok Elluswamy, Tesla’s director of Autopilot software. Three members of the Autopilot team got in the back. As they prepared to leave the parking lot at Tesla’s Palo Alto office complex, Musk selected a location on the map for the car to go to and took his hands off the wheel.

When the car turned onto the main road, the first scary challenge arose: a bicyclist was heading their way. On its own, the car yielded, just as a human would have done.

For 25 minutes, the car drove on fast roads and neighborhood streets, handling complex turns and avoiding cyclists, pedestrians and pets. Musk never touched the wheel. Only a couple of times did he intervene by tapping the accelerator when he thought the car was being overly cautious, such as when it was too deferential at a four-way stop sign. At one point the car conducted a maneuver that he thought was better than he would have done. “Oh, wow,” he said, “even my human neural network failed here, but the car did the right thing.” He was so pleased that he started whistling Mozart’s “A Little Night Music” serenade in G major.

I am pretty skeptical not only in the long run, but even that any demo of Tesla FSD was honest. Musk should probably be in jail at this point for falsely advertising a technology that led to fatalities.

Still, maybe Tesla has improved its FSD tech significantly and I'll keep an eye on their work...however my advice for anyone inside a Tesla that's driving itself is to keep an eye on the road and one hand on the wheel...

Still, maybe Tesla has improved its FSD tech significantly and I'll keep an eye on their work...however my advice for anyone inside a Tesla that's driving itself is to keep an eye on the road and one hand on the wheel...

I agree in the sense that you meant it, but I'm not sure driving a car like that would be psychologically wise. It might give you the subliminal message that driving was easier than it really is - a bit like 'driving' a toy car.

David

Again, as a Tesla owner, albeit without full self driving (FSD), the autonomous driving elements are wonderful.

I regularly drive (weekly) two 100 mile trips.  While not a difficult drive as those things go, it does incur some fatigue.  Since getting a Tesla this has diminished significantly.  I use two autonomous features that come with all Teslas (even those that didn't pay up for the FSD).  The first is an adaptive cruise control (available on lots of cars now).  The second is autonomous steering (also available in other types of cars).

This differs from FSD in that you have to 'set' both features to 'on' and it does NOT navigate for you.  Meaning, its really only valuable when on long highway drives (such as the one I make frequently).  It will not change lanes or navigate to other roads/streets.  It simply drives the car at the set speed, on the set road, in the current lane.  The car maintains its position in the current lane of traffic and adjusts both direction and speed.  The car has built it mechanisms to try and ensure the driver is still overseeing the vehicle (pressure to the steering wheel and even an internal camera that 'watches' you to see if you eyes are generally on the road).

Its been remarkably good at what it does.  It has even engaged in evasive maneuvering such as when cars in adjacent lanes get too close.  It slows down for slower traffic in your lane.  Etc.  It has noticeably reduced driving fatigue as while I'm still 'paying attention' I'm not tasked with physically operating the vehicle when these features are engaged.  I love it.

Now, the full move to FSD is another matter altogether as you might suspect.  I remain skeptical of that component but not close-minded to our ingenuity being able to figure it out.  Time will tell.

Now, the full move to FSD is another matter altogether as you might suspect.  I remain skeptical of that component but not close-minded to our ingenuity being able to figure it out.  Time will tell.

Oh there is almost a certainly a way to make a fully autonomous vehicle, but I don't think the current strategies can get there.

There is some work in robotics and comp-sci that may have promise but a lot of funding has been based on the questionable beliefs that our brains do something similar to the self-driving computational models and that you can just keep supplying data until you achieve your goal.

This is all before anyone really tries to mess with these self-driving systems using the varied vulnerabilities in machine "learning". There have been some protests in San Francisco with people fed up with being forced to be guinea pigs of corporations that put cones atop the cars, but we've yet to see how bad this can get...

edit:

Driverless cars were the future but now the truth is out: they’re on the road to nowhere 

Christian Wolmar

Developing driverless cars has been AI’s greatest test. Today we can say it has failed miserably, despite the expenditure of tens of billions of dollars in attempts to produce a viable commercial vehicle. Moreover, the recent withdrawal from the market of a leading provider of robotaxis in the US, coupled with the introduction of strict legislation in the UK, suggests that the developers’ hopes of monetising the concept are even more remote than before. The very future of the idea hangs in the balance.

Oh there is almost a certainly a way to make a fully autonomous vehicle, but I don't think the current strategies can get there.

There is some work in robotics and comp-sci that may have promise but a lot of funding has been based on the questionable beliefs that our brains do something similar to the self-driving computational models and that you can just keep supplying data until you achieve your goal.

This is all before anyone really tries to mess with these self-driving systems using the varied vulnerabilities in machine "learning". There have been some protests in San Francisco with people fed up with being forced to be guinea pigs of corporations that put cones atop the cars, but we've yet to see how bad this can get...

edit:

Driverless cars were the future but now the truth is out: they’re on the road to nowhere 

Christian Wolmar

New highly accurate GPS systems can quickly create virtual rails for fixed vehicle routes in urban areas, like a tram or train uses real rails on a fixed route. Pedestrian/cyclist streets/routes can be separated from vehicular routes, just as they are in the Netherlands. Private vehicles can be banned from urban centres. Traffic signal crossings can be upgraded with machine readable signals. Autonomous driverless buses could work in similar situations, with remote operator backup. This would allow flexible mass transit routes to be rolled out in urban areas that don't currently have them, and cannot afford them, quickly and cheaply. This could be rolled out as a series of stages that culminates with mass urban transport becoming driverless after all the other urban changes are completed.

I don't know how many times I was nearly knocked down by trams in the Netherlands because I'm used to vehicles driving on the left, the tram drivers made no attempt to stop, they whooshed by just inches from my back on two occasions after crossing the tracks and me looking the wrong way.

Honestly, I think the highly adversarial and emotive media stories are for other purposes like manipulating share prices, attacking competitors, etc.

Honestly, I think the highly adversarial and emotive media stories are for other purposes like manipulating share prices, attacking competitors, etc.

Yet there have been fatalities, and the people who stand to profit are not the ones who've been chosen to be part of the experiment?

In another thread you mentioned a critic of neuroscience possibly wanting to sell his book, but these companies have billions on the line and have failed so much investors are shying away...to me the manipulation is more likely to come from that direction?

For example ->

Judge finds ‘reasonable evidence’ Tesla knew self-driving tech was defective

Yet there have been fatalities, and the people who stand to profit are not the ones who've been chosen to be part of the experiment?

In another thread you mentioned a critic of neuroscience possibly wanting to sell his book, but these companies have billions on the line and have failed so much investors are shying away...to me the manipulation is more likely to come from that direction?

For example ->

Judge finds ‘reasonable evidence’ Tesla knew self-driving tech was defective

You need a wider frame of reference here... the elites DON'T CARE about fatalities... that article is an emotive narrative designed to elicit a behavioral response in the massed population... there is no other point in magnifying that, relative to other things which happen that you don't hear about.

We don't know what behavioral response is required, and to whom, or who suffers/benefits... we can only guess. But that is what is going on... it's honestly is not what you think it's about, but it's a narrative made to be arguably about that.

You need a wider frame of reference here... the elites DON'T CARE about fatalities... that article is an emotive narrative designed to elicit a behavioral response in the massed population... there is no other point in magnifying that, relative to other things which happen that you don't hear about.

We don't know what behavioral response is required, and to whom, or who suffers/benefits... we can only guess. But that is what is going on... it's honestly is not what you think it's about, but it's a narrative made to be arguably about that.

Isn't this comment veering into the realm of politics?

David