Live data from Hacker News

Tesla loses U.S. designation for some advanced safety features

reuters.com

361–370 of 379 posts

Re: Tesla loses U.S. designation for some advanced safety features

#361

Earlier quoted context omitted.

Human drive in fog just fine with pure vision, why can’t AI do the same? Keep in mind that vision is needed in all cases for lane keeping at the very minimum. So if vision is not able to see in fog, it won’t be able to self drive even radar is working.

> Human drive in fog just fine with pure vision, why can’t AI do the same? Humans also use real intelligence and higher-order thought when driving. Humans also have stereo vision. Unless you’re telling me that Tesla has invented AGI, the comparison is absurd. In the meantime, Tesla can’t even get my auto wipers right.

Is there a formal proof that FSD requires the same level of intelligence of AGI?

Stereo vision can be built from cameras as well, even with a single moving camera. This is exactly what tesla is doing with their new DNN but with multiple cameras.

I don’t disagree with you that Tesla is still far away from FSD, but using what’s not possible today to draw the conclusion that a goal is not achievable in the future is the opposite of first principle thinking. 10 years ago i could also say go AI can’t even beat semi-pro so we will never be able to solve Go with AI.

Re: Tesla loses U.S. designation for some advanced safety features

#362
post #224

Earlier quoted context omitted.

Human drive in fog just fine with pure vision, why can’t AI do the same? Keep in mind that vision is needed in all cases for lane keeping at the very minimum. So if vision is not able to see in fog, it won’t be able to self drive even radar is working.

False. Here's dozens of cases of pile-ups caused by fog: https://www.google.com/search?q=fog+highway+pile+up

Note that self driving doesn’t mean 0 accidents, it just need to be better than avg human. People get into accidents when there is no fog as well.

Re: Tesla loses U.S. designation for some advanced safety features

#363
post #260

Earlier quoted context omitted.

> Or, better yet, make a database and have cities provide feeds of changes/permits/closures. "Sorry, I know you crashed, your car's on fire, and you're bleeding out. But I need to get this road closure form submitted before I can help, or a self driving car might ignore the stationary accident and make things worse. Ugh, 5G's not working today... hold on..." The cars must be able to handle random, unexpected events h…

If both are 98%, and you combine them...

On an annual basis, you "self drive" a staggeringly large number of cars into accident scenes, road construction, traffic jams, police, and just about everything else that could possibly be on the road.

Start talking about 99.998% or so, OK, now it's worth sitting down with the math. But 98% accuracy at avoiding construction/existing crashes/etc is horrible.

Re: Tesla loses U.S. designation for some advanced safety features

#364
post #96

Earlier quoted context omitted.

More people need to see your comment because there seems to be a belief that full self driving is just around the corner. However, IMO, it will never happen because it needs a full artificial general intelligence. In your case, the new lane was clearly marked, which just shows how weak the system really is. Imagine if there was snow covering the lane markings. Or sunlight reflecting off of the road confuses the visio…

I think self driving will catch on for special roads or lanes that are automatic only. Mostly on long highways that are mostly straight.

That's the reality that Tesla owners already live in. Last week I visited my family for my grandmother's birthday. Of the 700+ miles traveled, over 98% were on autopilot. I entered my destination and hit "Navigate on Autopilot". It drove from onramp to exit, obeying speed limits and switching lanes to get around slow vehicles. It even handled the interchanges.

Re: Tesla loses U.S. designation for some advanced safety features

#365
post #276

Earlier quoted context omitted.

It doesn't make any sense to look at a situation and say "well, I could save 5 lives at no cost ... but nah". It really doesn't matter who is responsible when a self driving car kills someone. We're going to respond by making the self-driving module of the car safer whether it is responsible or not. We're putting a driver on the road who can still learn after suffering a fatal accident! This is huge, it means all par…

What makes you think cars will ever drive more safely than humans? What if they don’t? This always seems to just be a given, but like GP, I think “self driving safer than humans” requires GAI unless maybe you make massive changes to the road infra.

There’s no need for GAI, a driver doesn’t need to solve logic puzzles or crosswords or diagnose medical conditions.

A driver needs to identify lanes, understand road markings and signs, model the physics of moving objects and predict the immediate future behaviour of other objects based on immediate past behaviour.

Also basing an evaluation of FSD on current AP behaviour is like assuming a housekeeper will be as effective as a Roomba.

Re: Tesla loses U.S. designation for some advanced safety features

#366
post #335

Earlier quoted context omitted.

I didn't imply that you said anything about the look, I mentioned it as any example of something that is objectively different. So you mentioned actual argument, repair parts availability. So you mention one thing and the declare the CB to be DOA. You must release how incredibly dumb that is right? How would you react if a Tesla fan said 'Cybertruck is faster to 60mph therefore the F-150 is DOA'. Its just disingenuou…

> I assume you have some sort of dislike for Tesla. This is a theme in your posts. You assume that everyone critical of Tesla's choices must be a hater. Just because you see Tesla as some sort of personal reflection doesn't mean the rest of us do.

I assume that because he makes a single not very good argument and then draws wide ranging conclusions from it that simply do not follow from his argument.

Re: Tesla loses U.S. designation for some advanced safety features

#367
post #333

Earlier quoted context omitted.

Go calculated the battery need for these 400k/year and try to understand that industry. And Tesla is building a factory that is equally designed to build that many trucks and the vertically integrate it with a huge battery factory likely even cathode manufacturing and lithium hydroxide production. The last Ford CEO (not that long ago) still believed Ford didn't even need a battery partnership. And the new CEO has bee…

Ford has quite wisely chosen to let the battery experts make the batteries for them, and is using scale to drive down the costs of their batteries below what Tesla can achieve through vertical integration. Moreover, by not competing with the battery makers, Ford can always choose the best battery available when it comes time to sign new contracts, and isn't stuck with whatever technology their CEO is promoting on Twi…

> Ford has quite wisely chosen to let the battery experts make the batteries for them, and is using scale to drive down the costs of their batteries below what Tesla can achieve through vertical integration.

So did Tesla for 15 years and in that time they build a huge amount of knowledge and worked together with Panasonic on improving their cells. They learned a huge amount, hired many experts, bought a number of technology companies and are bringing things to production that literally nobody else has managed yet. They actually did the hard work of bringing university ideas into mass production.

Your claim that Tesla simply can not achieve the lower prices is your assumption that is based on nothing. Not having to pay for the profit of the cell maker alone is a huge part of the cost.

Ford has to buy on the very contested open market because their former CEO failed to set up a partnership. So in fact its very likely that they pay some of the highest amount by any of the major car manufactures. And of course so far they are a tiny part of the market while Tesla is by far the market leader.

Tesla simply buys far, far higher volume then Ford and will continue to do so.

> Moreover, by not competing with the battery makers, Ford can always choose the best battery available when it comes time to sign new contracts, and isn't stuck with whatever technology their CEO is promoting on Twitter.

Tesla is continuing to buy lots of cells from battery makers and they will continue to increase that. Tesla is the largest costumers of 2 of the 3 largest battery makers in the world.

The idea that Ford who has tiny volume will somehow get better prices then Tesla is nonsense. Tesla also has long term supply contracts for the raw materials that they locked in years ago while Ford is fully exposed to the raising market in raw materials.

> I'm part of the crowd of people who would use it. And I talk to people, online and off, in the hiking, trail running, mountain biking, and aquatics, and off-roading communities. The interest in the Lightning in these groups is huge. People even cancelled their orders for Subarus when the Lightning dropped. In contrast, the Cybertruck is viewed as a joke: a truck that can't be used for hauling, can't be used for work, and can't be used for any outdoor activities; in short, a truck that can't be used as a truck unless you think a truck just sits in a driveway or a parking spot.

So here is my problem with your argument. You simply say thing, based on literally wrong facts. Tesla Cybertruck has significantly better performance for hauling, loading and range at the same price. And a far, far higher top level specs. If you seriously want to haul something long distance the top range CB is far beyond anything the F-150 can do.

Its absolutely simply wrong to claim the CB can be used for outdoor activities or work. Like, this is just an opinion that you have that you have absolutely no factual bases for.

Talking to people who are not equally not informed and have formed their opinion on pictures and maybe a few confused news articles is simply not a valid source of information.

Re: Tesla loses U.S. designation for some advanced safety features

#368

Earlier quoted context omitted.

We've already seen several instances of Waymo's approach not working. I agree that ultimately they need a significant amount of local training data. But I believe they would be better off having a human drive a car around to map things out, run their 'AI' on the results, and then manually review/correct lower confidence results. Or, better yet, make a database and have cities provide feeds of changes/permits/closures…

Depending on what "not working" means to you, I'm ok with that. I think the point of the Waymo approach is to not work very frequently so they can flush out the issues.

I’m not interested in walking down streets that Waymo is using for their beta tests.

Re: Tesla loses U.S. designation for some advanced safety features

#369

Earlier quoted context omitted.

If those cameras were hooked up to an AR helmet and a human had 360 view of the car through them, I bet they could drive just fine. The data is there. Sensors might be an alternate strategy though and a way to at least take vision processing out of the list of hard problems.

Radar and Lidar would make the car safer than if a human was driving?

those are sensors. "safer than a human" is broad and not really applicable to talking about just sensors. but as an example, lidar has orders of magnitude faster detection of an object than human perception. so in that case, sure it is much safer.

Re: Tesla loses U.S. designation for some advanced safety features

#370

Earlier quoted context omitted.

again, nothing high danger. That is not a neural net, it could have easily been programmed to not try and maneuver around and just stay stopped, so sounds like a behavior that needs some additional improvement. It didn't lose sight of the cones. It was an issue with generating an appropriate path around the cones. Notice how it didn't hit the cones or anything else That is fundamentally different than trusting a neur…

> again, nothing high danger. Not having control of the car is high danger. Even when they recognized it needed rescuing they still didn't have control of it.

that sounds like a bug(albeit severe), but that is in no way descriptive of current technology or some sort of cap on performance
Post reply on HN