XaiJu
Some More News
Some More News

patreon


Dude, Where's My Self-Driving Car? – EARLY ACCESS, AD-FREE, AND UNCENSORED

Hi. In today's episode, we look at how autonomous vehicles (AVs) work, and also how they don't work, and also also how they're being tested on city streets without your consent to help enrich people like Elon Musk.

Dude, Where's My Self-Driving Car? – EARLY ACCESS, AD-FREE, AND UNCENSORED

Comments

I feel like Katie and Tom Cardy would have a lot to talk about. https://www.youtube.com/watch?v=--9kqhzQ-8Q

Eric Williams

In the 1970s there was an episode of the Rockford Files (The House on Willis Avenue, S04E21) that was about the collection of user data and how people may not know about it and legally there was nothing they could do. I would have hoped that more people back then would have fought against such collection but too many people seem to like the free stuff they can get by allowing this tracking (loyalty programs, etc.) On another note, I saw an accident (at least the after effects) on the weekend after driving home for over 4 hours. There was 2 cars that were damaged, a Mazda or Toyota with damage at the rear end, and a white Tesla (why are they always white?) with front end damage. I cannot tell for sure but I would not be surprised in the least to find out the driver was using Auotpilot, or even just adaptive cruise control, and the Tesla decided not to stop when the rest of the traffic did.

Rob Tapp

@Shawn that's not *quite* true. Two things the human brain excels at are building extensive heuristics and generalization, both things computers cannot do. Models have to be trained to specific situations and can only understand input based on that training, whereas the human brain can generalize previous experiences in a given moment to situations previously experienced, situations a model would view as unrelated to the data it was trained on. Humans can also learn by simply watching other drivers, again, where computers cannot. Maybe some day we'll be there, on the second count, but an equivalent of the former I don't think we'll ever see (but, how knows, maybe we will).

boldfield

I love SMN, but this video was frustrating for me. It's possible that I'm too close to this issue, but I do agree with most of the issues pointed out here. There were just a couple of things that felt like they had an incomplete perspective. One is that the video treated autonomous vehicles like a monolith, which they VERY much are not. As was pointed out, a lot is left up to companies to self-certify and attest, so that essentially means that whatever they think is good enough gets launched (which absolutely should horrify you). Companies have very different views on what that standard is, and from my knowledge of how the industry has evolved, Google seems to be the only one taking it seriously. I no longer work at Google because there were too many horrible things that I couldn't ignore, and I very much agree that I want to extricate Google from my life wherever possible, but to lump Waymo's AV system with the likes of Tesla's feels either dishonest or uninformed. Make no mistake: I don't think Tesla should be allowed to call their product "self-driving" at all, since it requires a human to be monitoring it at all times. That's dangerously misleading, as evidenced by all the accidents when people weren't paying attention. The other point goes to that uninformed bit: there were several times when the video seemed to be indicating that the software is programmed with how to handle lots of different scenarios individually. Maybe you know something I don't about horrible things that Tesla or Cruise are doing, but this is not generally the approach used for these kinds of computing problems. They're certainly tested with simulations of real life scenarios, but they aren't programmed directly in this way. Don't imagine a programmer writing a bunch of conditional logic by hand. Modern AI has more in common with a human brain than it does with that. And whenever questions of when the technology will be feasible end with "if ever", it feels like it's missing this fundamental difference. There is absolutely no reason to think that driving is beyond the scope of what synthetic brains can do, other than that they haven't achieved it *yet*. The human brain is amazing, but it's also just an organ. From a scientific perspective, there's no reason to believe it's anything more than the hardware that hosts our minds (as opposed to being the mind itself). I see far more reason to be skeptical that our brains are special and have some fundamental superiority, rather than being skeptical that they're not. All of that being said, I love you guys and everything you do. Except Warmbo. God, what did the world do to deserve him?

Stabby

Current cars NOT advertised as autonomous include LOTS of “assistive” technologies that isn’t necessarily even implemented at time of sale. Cameras are built into every outside facet of the vehicle; cameras “helping” you avoid collisions are actually un-helpfully over-sensitive yet also illiterate to a variety of road conditions. Beeping, alarm sounds and auto-braking all happen when you approach a tuft of grass, a drooping branch or a “please drive forward" drop gate in a car-wash…. Those can be disabled by the factory, but the tech is already installed without proper testing: Tech that an autonomous car would use to make decisions about what you can safely approach is actually interfering with the human driver’s much greater perception, processing speed, and contextual awareness. The “assist” is no better than a nagging, nervous, foot-stomping passenger frightened of every unexpected road condition. So we ALREADY have tech in our regular cars that makes driving less safe than regular human driving.

Carl P Crossgrove

Crazy timing of this episode. After an auto extrication course last week, my partner told me that the Canadian government is planning to make all new cars sold by 2040 autonomous cars. I got really mad and then felt bad because he was just the messenger. Of course, let's just slap a capitalistic money-making bandaid on the systemic issue rather than address its root cause.

Aednik K

Your show needs a tip jar so that I can drop $100 in every time you call JKR out on her TERF-itude BS.

The Wyverns

Sooo .. Katie... how's your butthole feeling today?

Jax Uzpen

Can we make an AI that drivers car noises?

Uthor

What gives me pause about self driving cars is an issue that has come up due to autopilot on airplanes. With the computer flying MOST of the time, the pilots zone out and lose practice. There have been incidents where something bad was happening and the pilot didn't have the skill a regular pilot should have to take care of the relatively simple problem. Crashes have happened because of this. When cars get to the point where they drive fine 99. 999% of the time, but a driver will have to be ready for an emergency (while doing nothing for days or years), bad stuff will come out of it.

Uthor

Fun fact: after this episode finished, Cody's "Vroom, Vroom" lead into an older Some More News episode "Palins War on Christmas" which starts with Cody going "Vroom" at the start. So the lesson is, the YouTube algorithm likes car noises. Food for thought and I had no where else to post this.

Bryan Newton

That rant at the end was ::chefs kiss:: It's funny cause I said, essentially, the same thing about the "metaverse" and comparisons being made about it would be like "The Matrix" or "Ready Player One." Like who saw "Ready Player One" and thought, "yes, I would like to live in a world where our houses are RVs welded together in the least safe fashion you can think of and our only escape from our real-life hell is a virtual world filled with nothing but nostalgia, ads, and mega-corps." Or "The Matrix," where the virtual world sucks just as bad as it does now, AND we live out our real lives in goo-filled pods while robots harvest our meat-suits for power! Where the fuck do I sign up?!

Matt Helton


More Creators