I honestly don't understand how anyone can actually believe that. Actual self-driving(not just a fancy cruise control) seems about 50 years away if not more. There will be "self driving" systems which work on nice american streets and motorways, but a general use self driving system that can be sold worldwide? Not a chance in the next few decades.
Tesla is trying hard too, but it goes with a big dose of bullshiting from Musk. They are a lot more dependent on AI break troughs in DL due to their vision only philosophy.
OpenAI research around scaling laws in DL suggest that naive 1000 increase in training compute over CLIP should get us to state of the art vision in zero shot settings i.e. without task specific training data. But they probably can cover 1/3 of that with algorithms improvements. Beyond that machines should really be above human vision across the board besides adversarial settings.
Quote https://arxiv.org/abs/2103.00020 :
> While scaling has so far steadily improved performance and suggests a route for continued improvement, we estimate around a 1000x increase in compute is required for zero-shot CLIP to reach overall state-of-the-art performance.
There is also the question of robust car control, but there should be good progress on it too due to RL research.
I think it will take years before Tesla FSD can drive from A to B without making catastrophic errors, but at least they're trying to solve the correct problem. And I think it's important to remember that it's OK if self-driving cars behave stupidly in ways that don't kill people, because the status quo has people dying every day because they're drunk/tired/texting, and self-driving only has to outperform the worst drivers in order to be a big leap forward.
They are trying to solve the “how to build FSD that works on existing Tesla cars because Musk has tricked a whole bunch of buyers into handing over money” problem.
Our heads constantly move around and have the worlds most advanced neural network trained over decades to understand objects and how they behave. So for us vision works. But because Tesla isn’t solving the right problem they insist on going down this road instead of LiDAR which is safe and works.
Sounds logical, doesn't necessarily mean it's true. I'm not an expert.
It's also interesting to think that we may have a lot of ideas like that. They are just waiting patiently for compute in some small communities.
And the compute quite consistently grows exponentially. We will make bigger absolute increases in compute in next 5 years than we did ever before.
So....I know this, and yet I still prefer to have my own car that I drive myself. What do self-driving cars solve here? A self driving car is literally like a taxi, minus a driver. Yet even if taxis were quarter of their price, I would still like to have my own car.
But sure, maybe it's just me. But I think a lot of people prefer to have their own vehicle rather than take taxis everywhere, even if it's logically cheaper thing to do.
It will only take a few fatalities for the govt to stop any autonomous vehicles on public roadways unless it is in a safe geo-fenced area with clear liability.
The life threatening situations are rather simple and reactive. It's a car colliding with something at a high speed.
No deep planing or human level negotiations necessary. Well designed LiDAR cars will almost never be at fault of killing someone.
I was predicting [0] in Nov 2018 that a next self driving car fatality will occur with 50% chance before March 2020.
80% before March 2021. Well, self driving cars are safer than I expected back then.
And Tesla should get there too sooner or later.
Also, the govt approach is proving to be quite risk accepting.
Tesla already killed and injured many people when on autopilot.
[0] https://www.metaculus.com/questions/776/when-will-the-next-f...
"automated system saved my life"
and
"automated system didn't kill me"
Like, my own car has several features which I believe are enough to prevent 90% of accidents in urban settings. But I find it hard to interpret when someone says "autopilot saved my life" - like, you mean "autopilot didn't kill me"? Because if it merely stopped hitting something then I really don't understand how you can be certain that
1) you wouldn't have avoided the collision by yourself
2) you wouldn't have avoided the collision with exisitng systems without going full autopilot
And I have been in a Tesla(few actually) and I personally think that they way Tesla does it is the worst possible implementation - it gives you so much confidence that you really stop paying attention(even though you absolutely should never ever do that) only to spontantiously decide QUICK! HUMAN, TAKE OVER! in the middle of a turn.
Like, if I owned a Tesla I'm sure I'd love the autopilot, but I also think it's far too dangerous to be forming any kind of statements like "we'll have full self driving rollout by 2025-2028" based on it.
"saving your life" - never been in such a situation before except once in my life when I encountered a parked tractor trailer at the side of a country road in the middle of the night without any working lights.
I saw it when the high beams hit the rear reflectors and reflective number plate and I swerved out slightly without a major correction otherwise it would been a rollover.
Multiple times ????.
Could you provide some data at least when you give any sort of predition on the future ? Otherwise it's meaningless. anyone can have a "feeling" or an "opinion"
Like, don't get me wrong. It's a great system. But it's just a very clever cruise control. It's nowhere near Level 5, not even Level 4 and I'd argue it's not even level 3 at the moment. And yet the OP says we'll see "massive self driving deployment by 2025-2028"??? Like...how? With what technology?
So? I can't BASE jump, but you should still heed my criticism if I tell you doing it from 30 feet up won't work.
Someone who can't cook can still identify bad food. Someone who can't code can point out that Facebook's UX is confusing. A non-driver can identify bad driving. A non-singer can identify someone being off-pitch. Dismissing criticism in this nature is a mistake.
No its just a meaningless opinion that is polluting the feed. Its just noise. In general in technology debates, especially here on hacker news, we prefer actual "insights" or "diffs" if you go read the site guidelines.
Just look at the state of Tesla's full self driving product. Yes, it's impressive in many situations yet it fails at things that are generally. What's the point of automation if I have to babysit it?