Video only learning from end to end neural networks in Tesla FSD 12 is like the large language model Chatgpt. Adding more video data and compute will make Tesla FSD vastly better in months. Tesla FSD is also the basis for the real world AI for Teslabot.
Tesla FSD, robotaxi level FSD and factory and real world usable Teslabots should all arrive within 24 months.
Tesla increased their available compute by seven times this week by enabling 10,000 Nvidia H100 chips. Tesla now has over 100,000 Nvidia A100 chips worth of AI compute.
In 10 years we will look back at the first public demo of a robot that learned to move around the world by watching only videos.
This is a paradigm shift in how software is built.
What is also mindblowing is that the inference compute power needed for 8 cameras running at 36FPS is only about 100W on the Tesla-designed AI computer. This puny amount of power is enough to achieve superhuman driving!
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.
I’ve read some review of Musk’s test drive with this new NN system/ I also watched a bit on YT, reviewer said (correctly) that there was an intervention in 20 something minute, the same average intervention ratio, than in previous system(coded by hand by programmers). His conclusion was that as long as there are the same number on interventions per, let’s say 5 hours of driving, it’s not really an improvement.
And I agree, when we will see significantly lower number of interventions, then we can say, world has changed, so far, we are at the same/similar level as previous ‘FSD’ version.
There are literally infinite number of new, bizzare situations on the road that may happen to any Tesla each day, and we don’t know how NN will be reacting to them, as long as there will be confusion and cars will be stopping, we won’t have FSD. We can say we have it, if a person without drivers licence, sitting on back of the car can use robotaxi and have long trips, without being stopped each few hours.
I am a believer, but it’s to early to say NN will be successfull and will solve the hardest, rarest, weirdest situations, we need more data
I think we are 5-10 years away to true self driving and that is IF. Too many exceptions.
“Tesla increased their available compute by seven times this week by enabling 10,000 Nvidia H100 chips. Tesla now has over 100,000 Nvidia A100 chips worth of AI compute.”
H100 are more powerful than A100, but still almost 7x times more compute from 10k gpu compared to 100k older gpu?
Even if they are up till maximum 30x faster, they are maximum 9x faster for AI training.
I am not certain if they will perfect this in two years. I am certain that if they do it will be world changing. There are 3.5 million truck drivers in the USA. There are 20,000 car crashes per day. If your timeline is correct my further prediction is humans are not allowed to drive on open roads within 15 years.
Could you send your child to school in one? Could it travel to the school, wait in line and drop them off?
I honestly don’t know. But, that is how I would define a true robot automobile.
Category: Science
Source: Next Big Future
