301–310 of 379 posts
For years I've been saying my next car will be a Tesla, because I wanted an electric car with autopilot. A few weeks ago I got a Comma2, which runs OpenPilot, and installed it on my Honda Odyssey minivan. I've driven a Model X with autopilot, and I have to say, my minivan is just as good, if not in some cases better, than the Tesla. I'm now much less excited about getting a Tesla. Knowing that I can upgrade basically…
Earlier quoted context omitted.
I'm not talking about running general purpose CPUs. I'm talking about the most high end specialized silicon we have. Throw the best TPUs ever made at the problem and you still won't get anywhere even remotely close to what the squishy pink meat can do. The power gap is so huge, our software simply must be inadequate. We're talking about a power gap wider than the gap between my toaster oven and the Saturn V.
I rechecked some random internet articles, and the brain contains about 10^15 synapses. They operate at about 10Hz (~10 PFLOPS). If we arbitrarily assume that each synapse operation is equivalent to 1 floating point operation on a computer, then this is about 3 orders of magnitude more than a good GPU (~10 TFLOPS). Which is actually a lot less than I thought. Still, the brain's basic structures have been optimized ov…
Earlier quoted context omitted.
I believe it follows the car, as you can buy used Teslas with FSD.
It follows the car if you sell directly to another person or through a third party dealership. If Tesla gets the car back, they'll wipe the upgrade and charge for it again.
Earlier quoted context omitted.
There's also the very common argument (even here on HN) that it doesn't matter because "humans have plenty of accidents too". I think it's not a good rebuttal because self-driving cars will have to be orders of magnitude safer than humans if order to be successful. If I'm drunk and I decide to drive and I end up crashing into a pedestrian, then it's my own damn fault and I'll face the consequences. If a self-driving…
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…
Earlier quoted context omitted.
I'm not saying that FSD was done in 2017. But, How long are you going to hold on to that one? In 2024 when FSD is working really well are you still going to say, "Yeah I mean it work now, but what about that tweet from 2016"
If 2024 comes and Musk says it’s coming in 2 weeks will you still believe him?
Earlier quoted context omitted.
I think we still must be missing something very fundamental with the way we’re doing AI. When you look at nature and see a tiny bee or fly navigating perfectly around all sorts of obstacles, including the newspaper I’m trying to swat it with, using just that teeny tiny brain running of mili-watts of power with just compound eyes to guide it. And in 3D space not a nice flat surface. But we still can’t make a car not d…
Organisms tuned by years of evolution can fail when confronted by novel environments. Think of a bee that wanders indoors and keeps hitting itself on a window pane trying to escape. A moth attracted to a light bulb. A cat chasing a laser pointer. Deer paralyzed by headlights.
If driverless cars were crashing because all of a sudden a giant sheet of glass appeared in the middle of the road, or a giant alien death ray shot up the tarmac, I think most people would have some sympathy.
Earlier quoted context omitted.
There will always be an edge case that could result in a crash leading to a loss of life - don't think 99% is good enough that is why we won't have aircraft flying on autopilot all the time.
Humans are hardly perfect drivers either.
Earlier quoted context omitted.
I think we still must be missing something very fundamental with the way we’re doing AI. When you look at nature and see a tiny bee or fly navigating perfectly around all sorts of obstacles, including the newspaper I’m trying to swat it with, using just that teeny tiny brain running of mili-watts of power with just compound eyes to guide it. And in 3D space not a nice flat surface. But we still can’t make a car not d…
I think your comment is interesting and generally agree, but to play devils advocate, there are certainly times when a fly might crash right into a window or something and then just gets up and keeps flying so there is also an aspect of resiliency built into a fly that we haven't (or maybe can't) built into cars.
Earlier quoted context omitted.
I think we still must be missing something very fundamental with the way we’re doing AI. When you look at nature and see a tiny bee or fly navigating perfectly around all sorts of obstacles, including the newspaper I’m trying to swat it with, using just that teeny tiny brain running of mili-watts of power with just compound eyes to guide it. And in 3D space not a nice flat surface. But we still can’t make a car not d…
These are self inflicted problems by only using cameras. These are entirely avoidable problems using the sensors and methods every other company besides Tesla are using
Sensors might be an alternate strategy though and a way to at least take vision processing out of the list of hard problems.
The full self-driving capabilities at $10,000 seem to be a poor value when you compare it to the fact that you could just... pay a person to drive you around. I mean, hopefully no one is financing or leasing these cars when you're purchasing FSD, right? It only makes sense to me if you buy the car in cash. You could literally just set aside the $10,000 to occasionally have a professional driver chauffeur you, provide…