I believe first the company and its engineers that concluded radar wasn't needed, not YouTube Warriors watching clips.Yes Sir I will concede that it is an assumption but it is an assumption derived from only a few possibilities that in the end only leave one plausible explanation. It’s ironic that so many here are rallying around the video only Teslavision and proclaiming how much better video is than radar but when an actual video is presented as evidence they suddenly don’t believe video is good enough and can’t be trusted. Video is good enough for Teslavision but video isn’t good enough when offered as evidence.
I don't think it's that so much. The more the smart man learns the more he understands how stupid he is and is circumspect about what he says for fear of advertising his stupidity. The stupid man is not restrained by that fear. See Dunning-Kruger effect (which I note, with great amusement, now seems to be considered politically incorrect).People who don't know what they are talking about are the loudest because they don't have much to say.
I don't think it's that so much. The more the smart man learns the more he understands how stupid he is and is circumspect about what he says for fear of advertising his stupidity. The stupid man is not restrained by that fear. See Dunning-Kruger effect (which I note, with great amusement, now seems to be considered politically incorrect).
I agree with your premise that we shouldn't limit vehicle capabilities based on human limits. It's actually nonsensical as the entire point of any tool is to transcend human capabilities.So my comments about passive SDR radar, passive sonar, and phase array radar where actually all legit. I might be a "funny guy" but I'm not always joking. ;-)
Any argument that starts with "if human vision is good enough for driving" is dumb, no matter who it comes from. Humans should NOT be the measuring stick especially if you look at their poor performance history. The question should always be how can we make humans obsolete in that task when it comes to creating new tech. That means do it better not as good as.
The issue that bugs me with all hype comments about FSD and how good it is etc, is that professional drivers, and skilled drivers are in a league of their own when compared to "normal" drivers. Many of the traffic issues are aged or inexperienced drivers, with poor eyesight and poor judgement, the other half of accidents are drunk or stoned. Yet another factor is very poor standards of instruction for beginners that leads to poor driver habits and etiquette. Plus poor maintenance and only intermittent roadworthyness testing. These all compound to give a huge sub-standard set of drivers and vehicles, that most definitely should not be the standard for FSD. Just because we can doesn't mean FSD should. This further invalidates the vague statement, hence my opposition to it. "Good enough" simply never is.
That leads to next point, that I think we'll be lucky to see FSD performance in a race, as good as a professional driver in the next decade. And this re-establishes the baseline of where FSD is now. My 5 year old grandson can probably drive as well with a bit if training. (Yes our whole family is special like that ) Cruising down a freeway with no throttle, brake or steering input has been standard on many cars for years as "assistance" packages and is the primary reason the term "autopilot" was taken off the German menu. That is not worthy to be called FSD in my book despite the fanboy videos. FSD town driving is another matter where Tesla leads.
To be clear I'm not against FSD, but we need a reality check on the current atate of affairs.
Of course if radar "works" is dependent on their implementation and priority to get it to work, and in particular in what situations. Because from what I can tell (also from this convo), it seems vision is better in common normal driving situations and radar is better in less common fringe cases with poor visibility.
And I think that is the key to the discussion, shouldn't we be adding "extra sensory" perception as well because humans should not be considered the standard?
The best part principle needs to act within the reasonable confines of "you need some sort of part to have a function". Otherwise if you don't you won't have a car either. All principles have constraints.
And this is where phase array beam forming would become useful in that it can act as a lidar, (yes I said it) but because of its near solid state would be able to scan at much higher frequencies, and resolutions. Plus it could actively focus a ping to track items of interest in realtime to add fidelity. Technically you could even start bouncing it around corners etc too. Try that with CV and reflections...The cost of these could be minimal as SpaceX could supply them.
That leads us to extra sensory perception in the non-visual EMF frequency ranges using passive SDR radar. The passive part simply means that it is a "listen' only system, unlike lidar that is transmitter as well. SDR is Software Defined Radio, where most of the radio analysis is done in software, which makes it highly capable but super cheap. The brilliant thing about passive radar is that you use existing ambient RF noises to generate a location map. Be they from other EV drivetrains or cars etc or from just mobile phone signals and radio towers.
And lastly passive sonar. There is huge treasure trove of things you can hear, in particular if you can sample and analyse it in realtime. Just imagine if your experienced mechanic would always be riding with you to tell you what he doesn't like hearing on the car. Things like the hepa filter or ventilation is blocked, or a CV joint is got to much play and is clicking while you turn, the tyre or wheel is out if balance, hinges need oiling, windows get broken etc etc. And its cheap to boot because microphones and half decent AD converters are cheap, and processing is trivial and already onboard.
What does everyone think?
Completely agree that all technology is a iterative process.I agree with your premise that we shouldn't limit vehicle capabilities based on human limits. It's actually nonsensical as the entire point of any tool is to transcend human capabilities.
A few points spring to mind (not just responding to you, but to the general thread):
In short, I would like our tools to be able to do everything better than me, that's why we have tools afterall. So I think my vehicle should be able to see through fog, heavy rain, be impervious to phaser fire and able to survive driving off a cliff only to be found, years later, buried in the sand by a kid on the beach. But perhaps we have to iterate to that end goal? Maybe the price is a few vehicle upgrades or retrofit kits to enable in the intermediate goal of achieving FSD in some common scenarios first?
- When Musk says that humans use two low resolution cameras etc. I believe he is saying that our equipment is not that good and we can exceed it. It's a way of framing the problem in a way that's relatable. I don't think he or their engineers at Tesla are limited by this thinking.
- It's true that FSD just has to be better than most human drivers in most situations to be a huge win.
- Extra-super human abilities could be added as the technology progress. Perhaps the problem of sensor fusion was too much of a distraction? Maybe there was a supply issue with Radar. Maybe Radar wasn't the right solution anyway once sensor fusion was solved. So better to have more capable FSD for most conditions than less capable FSD in all conditions.
- There does seem to be this idea that all Teslas will be able to do anything with software upgrades in the future. For FSD Tesla did say this, so fair enough but I think we have to accept that there will be obsolescence and we may need to retrofit or upgrade Tesla vehicles to obtain phased array beam forming or heisenberg compensators.
Yes, I think that's a lot of it. And I think a lot of it stems from their ability to convince themselves that they know more about a subject than someone who really does. Believe me when I say that I've dealt with this a lot in several fora dealing with various subjects. Someone makes an absurd claim. All they have to do is look up the subject on Wikipedia to see how absurd it is but they haven't done that and if you suggest they do they ignore that advice and continue to insist that they are right even if it violates a physical law or mathematical principle. It becomes pretty clear that there is no point in arguing with such a person and refraining from doing so is my general approach. The only exception is when these people say that it is OK to do something that, if done, might get someone hurt or result in a building code violation or something like that. Thus I leave it to the readership to decide whether to ignore or follow these people. To an amazing degree the readership follows them. A poll was done here I think asking who the members considered to be the most informative. There are a handful here who are astonishingly consistent in being wrong in what they post. They were rated highly as being the most informative. Thus there is no incentive for them to check on the accuracy of their postings as they are equally rewarded for posting disinformation as those who post accurately.But I find the tendency is that those types of people find volume more appealing than content too, especially if it fits their own biases. This becomes self validating through a group herd mentality, because everyone starts repeating others based on emotional positions, regardless if it makes sense or not. It never fails to surprise me how strong the desire is to belong, and what cognitive sacrifices we make to achieve that.
I think in trying to solve a problem that requires estimating the state of system one should consider any sensor that is likely to improve the desired estimate. There are obviously trades. Phased arrays obviously improve angular DOP but require coherent processing. That's tricky with light but it can be done thus implying expense which must be traded against the gain in the state picture one might gain from adding such a sensor. Getting back to Dunning - Kruger, I am smart enough to see the potential benefit of having such sensors in the suite but too stupid to be able to assess what that benefit might be in terms of the overall requirements.So what do you think of phased array beam forming lidar based on Starlink dish, or passive SDR radar or the sound monitoring? I note that EM recently said they're adding sound to FSD so they can detect emergency vehicles etc. What are your thoughts?