Tesla FSD is rapidly improving and seems to be creating level 4 self driving before our eyes. It is something that is difficult for many to believe and accept. AI systems have a history of not working for a long time and then having improvement that seems sudden to human observers. Once an AI gets into the pretty good relative to humans then they tend to very rapidly get to as good or better than the best human.
They are now releasing FSD 12.3.4 to all Tesla buyers in the North America. It is available to almost 2 million new drivers. FSD 12.4 should be out this month (April) and it is reported to be substantially better. FSD 12.3.4 should be able to handle reverse. There are only a few features that need to be added to get to a feature complete system.
There were reports two weeks ago from some Tesla engineers who believed that FSD would reach robotaxi level by July 2024.
Tesla will be revealing its robotaxi vehicle on August 8, 2024. We will learn on that day what Tesla is planning and possible timelines.
The initial review for Tesla FSD 12.3.4 are good.
Brian Wang is a Futurist Thought Leader and a popular Science blogger with 1 million readers per month. His blog Nextbigfuture.com is ranked #1 Science News Blog. It covers many disruptive technology and trends including Space, Robotics, Artificial Intelligence, Medicine, Anti-aging Biotechnology, and Nanotechnology.
Known for identifying cutting edge technologies, he is currently a Co-Founder of a startup and fundraiser for high potential early-stage companies. He is the Head of Research for Allocations for deep technology investments and an Angel Investor at Space Angels.
A frequent speaker at corporations, he has been a TEDx speaker, a Singularity University speaker and guest at numerous interviews for radio and podcasts. He is open to public speaking and advising engagements.
Fact: Many long uncut vids outthere that show how good FSD 12 already is. That people that complain either have bad luck with their area / situations that FSD needs to learn first, or they just don’t like Tesla and have never been using FSD 12 at all (even when FSD is perfect, they will complain).
I do “study” the subject dude, and that’s why I ask questions. And no, I haven’t sold any Tesla products. And what does this has to do w/answering my question?
Tesla allowed me a trial of fsd_supervised and I used it to drive from Scranton PA to upstate NY to see the eclipse. I couldn’t believe how it was driving.
At one point it stopped for an orange light where there was no traffic coming my way but it stopped so far past the white line that it couldn’t see the traffic light anymore.
During the whole trip there where many occasions where it was hesitating so much about what to do I had to intervene because it would just jiggle the steering wheel.
It’s not even close to being ready.
I love my new Tesla and the FSD is amazing. Does any other company have a comparable system? I don’t understand all the critical feedback it receives. It must be unreasonable expectations set by media and Musk himself.
I am not convinced that Tesla will be the first company to solve autonomous driving. Musk has a consistent tendency to overpromise and underdeliver.
However, even under the assumption that they are the first to truly solve autopilot, in my opinion, Tesla will be doomed, and the issue is twofold, from manufacturer’s side and from the human side
Solving self-driving means solving legal/regulatory issues about accident liabilities that will misalign the interest of the manufacturer and the car owners. And the manufacturer will likely have to put them in black and white on their user license/ownership contracts:
-is it ok for your car to choose a trajectory that will kill you but will save a child who jumped in front of the car? What about if it is an elderly man who stumbled?
-On the other hand, is it ok to choose to save the car occupant no matter the collateral cost, even though this will increase the insurance costs because you might end up killing and maiming more people?
-What about the differential survival rate inside the car?
These look like absurd questions, and an autopilot with superhuman driving skills (that does not exist at all at the moment) might reduce the total amounts of accidents, but accidents will still happen (due to malfunctions, for example) and the manufacturer’s legal office and software engineers will have to have the answers, because in a full self-driving scenario, the company will be the one driving. This issue is true for all the manufacturers, but different manufacturers might choose different solutions and might compete for customers (one claiming that their model will save you no matter what, but will be more expensive due to insurance costs…)
However, Tesla has specific issues:
because Tesla cars are considered cool cars and are consumer models, not a specific kind of vehicle model that is not consumer-grade (like the ugly London minicabs, or bosch’ experimental pods for example).
If Tesla robotaxis becomes real, since there is a private market for Tesla cars, criminals will summon them somewhere to steal them. It is not going to be particularly difficult:
You summon a Tesla somewhere, and then some accomplices park in front and behind the car close enough to block it there (you will probably be able to use traffic cones, too) while someone hacks the system.
I assume that the software will be patched regularly, but there will be a persistent push to find new exploits (as there is a persistent push to hack/jailbreak iPhones). Furthermore, it is becoming clearer and clearer that it is not easy to patch neural networks effectively.
-please note that hardware exploits will be much more difficult to patch as cars (especially if Teslas will be as reliable as Elon claims) will outlive most consumer-grade electronics lifecycle so that you will have old models with expired warranties traveling around and probably not eligible for cheap hardware patching.
Whatever the strategy, the issue is that there will be a market for stolen cars and spare parts. If Tesla tells the car owner, ” You activated the Robotaxi service, and they stole your car/ vandalized it. We are sorry, but it is your problem,” then the Robotaxi service is as good as dead. If Tesla somehow compensates the owners, then Tesla’s costs will skyrocket (and, nevertheless, Tesla will expose itself to fraud).
Please also note that as long as there is the possibility to summon the car, even without the Robotaxi service, there will be a strong criminal incentive to break the Tesla app, steal Tesla owners’ phones, and so on to steal the car.
If you have crappy and sturdy cars that nobody wants to steal and that are not meant for the consumer-grade market, you might have an autonomous self-driving service, if you have cool cars, in my humble opinion, you will have problems.
A similar discussion might be done for autonomous self driving trucks.
Truck drivers do not only drive, but are there making sure that nobody is stealing the payload (nor the truck). If you have millions of autonomous trucks on the streets you have a multibillion dollar incentive in learning how to hijack the truck to steal the payload and deliver it the temporary shade warehouse of your choice.
I do not see these issues ever discussed when I read about the trillion dollars self driving market, but I think they should not be underestimated when projecting tesla’s stock (or any other car manufacturer deeply involved in autonomous vehicles) evaluation
OK, stupid (but IMO reasonable) question: What was Level 1 threw 3, and why did I hear so little about this in open media? Just how does this technology work, and why should we trust it? Corporate confidentiality goes out the window when it affects the public at large. So I ask again, how does this technology work, and why should WE trust it? Like nuclear powered ships, you have ONE chance to get it right, or no one will accept such things for decades.
Get it right now, and take your time to do it right. If you DO IT RIGHT, you won’t have to convince anyone, we’ll just “get it” . By the way, what was level 1 threw 3? I really want to know…
If you haven’t sold Tesla yet then you should study the subject and wise up.
I do “study” the subject dude, and that’s why I ask questions. And no, I haven’t sold any Tesla products. And what does this has to do w/answering my question?
Nonsense! FSD 12.3.4 on my brand new M3+ has many issues, just like the 11.x.x releases preceding it; and others in the Denver area Tesla Facebook group do, too, as well. Get real!
I’m just gonna leave this here: https://gizmodo.com/elon-musks-worst-predictions-and-broken-promises-1851398745
The same headline would make sense whether you have written it in 2018, 2020, 2022 or 2024
Key words here are “seems to be creating level 4 self driving”
seems to be creating…
Regulators are deciding what level is FSD, not Musk, not Brian Wang. Musk will reveal on August that his new robotaxi vehicle will be built in 2026 and be available for 2027, reaching level 4 in 3 years, not in 3 months.
I will take that bet. You name the amount.
Until Tesla and regulators solve the liability issues, there will never be level 4, no matter how infrequent accidents are. To date, Tesla has never accepted liability for any accidents while in FSD mode, at least not publicly.
Waymo is driverless in city driving here in phoenix. I own an 2016 model S. The hype lead me to believe that it someday soon would be selfdriving, like the Waymo is now. The 2016 S will never be self driven due to hardware incompatibility. The hype lead me to believe that autopilot would be autodriven. No-one knew at the time the difference- the hype lead me to believe that all Teslas could be autotaxis.
They don’t need to. You accept liability in terms of use. At the same time how does something that is significantly less accidents when used Require such scrutiny around liability.
Not true. Tesla also do insurance!