Tesla FSD is changing radically with version 12.3 having three times the miles between interventions vs version 11. FSD 12.4.X could have 5-10 times the miles between interventions. New teen human drivers are three times more likely to have accidents than average human drivers. Removing the human driver completely to achieve robotaxi will involve government approval. However, increasing Tesla FSD safety by 3-10 times for each major release every 1-3 months means that Tesla is providing entirely different driving products every 1-3 months. This will be a shocking series of transformations. Each new level of capability will have huge changes in financial and societal impact.
If every 1-3 months there is change as radical as going from a first time teenage driver to an average middle aged human driver then Tesla FSD becomes a completely different product every 1-3 months. This is before the robotaxi government approvals.
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.
We won’t have FSD before AGI/ASI. These “30-60 min drive without intervention” vids don’t tell us much. Only few guys on YT are uploading them. To see real progress, we need to see data from all Tesla owners having and using this system, not just few random (mostly fans) uploading few of their (I guess cherry picked) rides.
I see in other comments people are becoming more and more sceptical as time goes by and I have similar thoughts. I was believer few years ago, but I saw too many times those claims “next update/ version X.2.3 will be it” years are passing by and they still are making stupid mistakes.
PS. I believe we may have AGI between 2025-2030
The comparison in levels between teen driver and chauffeur is completely misleading. FSD is more like someone that somehow knows how to drive but suffers from sporadic unpredictable epileptic seizures. If thus epileptic chaffeur suffers on average one attack every 10 miles, or 30 miles or 1000 miles it makes little difference. You will not hure him. You might start trusting the system if a seizure occurs, statistically speaking, every million miles, but even in that case, for every million drivers driving a mile you will have an instance of failure in the first mile.
Lol… I have always referred to my Telsa Model 3’s driving ability as a “nervous 14 year old”.
Hm.. According to teslafsdtracker, there was no improvement in miles between interventions in the 12.3.x releases. 12.x releases were about as good as 11.4.
Hate to say it, but there is a huge discrepancy between Elons rhetoric and real data when it comes to fsd. Why does nobody with access to him ask him about this difference? Elon is talking about thousands of miles between interventions and the data shows 34 miles…
Users report no significant improvement in FSD in last two years. It still needs interventions at about 1000x the rate required to be as safe as a human, so still needs about 999 out of every 1000 current interventions fixed. Perhaps some of that performance gulf will be bridged by subsequent software updates, but for sure it won’t go all the way to being safe enough for full level 5 autonomy.
They are currently working with HW3 but have HW4 in some cars and HW5 to start rolling out next year is 50x higher processing power than HW3. They wouldn’t be doing that unless they thought there was a likely need for it – suggesting that they’re already pretty sure internally that current hardware with processing of ~150Tflops, equivalent to a mouse brain is woefully insufficient for the job. HW5 is closer to a dog brain, and more likely to work in predicting behaviours of other road users.
The legal issues with confronting that reality (ie all Teslas shipped to date not being able to do FSD) will be major.
Several states already allow robotaxi operations.
https://torc.ai/which-states-allow-autonomous-driving/
There’s likely ways to dodge regulations in other states with minimal repercussions. Some Uber operations were technically or totally illegal.
Tesla will probably find a “good enough” compliance path, rather than 100% compliance. Maybe they’ll declare that FSD (unsupervised) is a person and get it a driver’s license in one state, let courts figure out how to fight it.
It’s going to be in teenager driver format for a very long time, as long it continues to make mundane mistakes which only could be attached to a drunk blind person.
For Brian Wang, we will get LV5 FSD and AGI in 2025, LEV in 2029, and LK99 is the real thing. Pure delusion. Wonder how you get your 95% rate on metalucus from.