Chuck Cook has drone video of Tesla FSD 12.5 supervised making many perfect unprotected left turns. Tesla FSD is the Full Self Driving product which is a driver assist system the human driver must supervise the driving system. As of Tesla FSD F12.3.6, human drivers will not need to touch the steering wheel.
It is unprotected left turns (no lights) with heavy traffic. Tesla basically had this fairly perfected by 12.3.X but it is an indication that there is no hesitation or other issues on it. It is the only video we have of the testing. The other big things to see will be going in reverse and perfectly dealing with human hand signals, flashing lights and all emergency vehicles. Other robotaxi like Waymo and Cruise had issues with emergency vehicles and construction.
Chuck Cook is famous in the Tesla community for his thorough testing of previous versions of Tesla FSD.
This videos shows several Tesla test cars using FSD 12.5 to make the difficult turns.
Here is my priority list of what Tesla FSD 12.5 should hopefully have improved and added versus older versions.
An protected left turn is a common critical safety capability and is the ninth item on the checklist of features below. The checklist was not updated for four months. There have been several improvements for 12.3.6 being able to handle many capabilities better. There has been improvement in 12.3.6 on avoiding pot holes and with understanding hand signals. The common critical safety capabilities should go from 2 out of 6 green to 6 out of 6. They were at 7 out of 10 and should be 9 or 10 out of 10 or even blue where it works all the time perfectly.
It would be a huge advance for full self driving automation for the system to work perfectly most of the time but good the rest of the time. Obviously, it would be best to have it work perfectly every time but progress is first to get to perfect most of the time and good the rest of the time.
Tesla FSD 12.5 has not been released to the public but it will likely go to other Tesla employees on the weekend and could go to the first wave of the public next week.
I discuss with Herbert Ong how Tesla FSD 12.5 could change everything.
The chinese robotaxi companies and Waymo in the US have Lidar, sensor and hypermap dependent systems. This is why they are only using a few hundred to a thousand vehicles in small geofenced areas. The Chinese robotaxi companies want to rapidly scale to profitable service with tens of thousands to millions of vehicles by 2030. IF Tesla FSD 12.5 and later versions of Tesla full self driving solve driving then scaling to 100,000 vehicles next year and a million vehicles the year after can be done. This would be with human safety drivers and remote drivers until regulators are satisfied with the safety data.
However, the robotaxi partners would need to buy many Tesla vehicles to get market share and awareness in each city and they would need to work out all of the operational, charging and other issues of a full Uber-Didi ridehailing service.
Here are some of my other videos that explain how vastly improved automated driving changes the game with each tripling of miles per intervention
I explain the China automaker, robotaxi and ridehailing market.
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.
So who was NOT in the way, when the car/truck turned left? Think about it…What happens when something does happen, that’s utterly unexpected? When something happens, no person did imagine? Think a computer can figure it out by just processing 1’s and 0’s real fast? I wish it was so simple. IMO, data, and intelligence are not the same. Awareness and intelligence are also, not the same. Intelligence is the beginning of perception of something beyond you. IMO, primitive animals may likely perceive, and do, things beyond who they do notice before they “know” who they are. Maybe not so primitive. I wonder, if self awareness, evolved from something we observed, and did not connect the dots.
But invented the dots, for us to connect? I digress (I do that FREQUENTLY) AI to me, represents many contradictory issues. Your dealing with a technology that is on a physical sense, fundamentally static. Integrated circuits can NOT change their physical morphology. I hit my laptop with a hammer, what’s the consequence? If I hit it hard enough, I will need a new laptop. Even if I hit it “a little hard”, and damage certain components, it will NEVER repair itself, let alone “improve” itself at a physical level.
It can’t do that. The IC’s will never morph into structures they are not now. Yes, programing can do amazing things. And I’m amazed at what programmers can do with such physically limiting technology. (They DO impress me…) I have said before on this site, and I freely admit, I am prejudiced in favor of biology. That’s my background. There is so much about biology we don’t understand, but we do know this: It’s unimaginably robust, and so adaptable. You can kill 99.9% of the bacteria that infect you, but that last fraction of a percent? It’s able to kill you. I’m sure it’s nothing personal, the bacteria just will do whatever it has to, (without actual “intent”) to survive.
OK, just got cut off. Don’t like that. Much of what I tried to write is gone. I’ll pick it up later. Don’t like what I say? Say so. Don’t try to put a figurative pillow over my face. I’m gentle, but not passive. I’m sure it was an accident on this sites part.
Some problems with hand signals.
They are NOT standardized; ‘stop’ is pretty straight forward, but ‘speed up’, ‘slow down’, ‘wait a sec’, etc are iffy.
What you do NOT want is some bum on the side of the road able to stop your car just by standing on curb and lifting his hand up.
The problem with FSD lies in the edge cases. Being able to do most road maneuvers normally is really the lowest bar for any technology that calls itself ‘Full Self Driving’. The harder part will be making sure that it drives safely some number negligibly close to 100% of the time.
Is the implication that Tesla’s self driving module couldn’t previously make left turns? That must have made for some circuitous routes.
You beat me to that thought by a minute!
It is unprotected left turns (no lights) with heavy traffic. Tesla basically had this fairly perfected by 12.3.X but it is an indication that there is no hesitation or other issues on it. It is the only video we have of the testing. The other big things to see will be going in reverse and perfectly dealing with human hand signals, flashing lights and all emergency vehicles. Other robotaxi like Waymo and Cruise had issues with emergency vehicles and construction
I have added in a crowdsourced checklist of features for FSD and how good some people thought the previous versions were.
FSD 12.5 will also merge highway and street driving. This means FSD smoother and safer than autopilot for highways. Autopilot was already about 12 times safer then average USA drivers.
Merged means no handoff problems going between highway and city streets or vice versa.
Ah. I figure there had to be something more than just turning left. Though you’d think left turns with lights would just be a restricted subset of ‘unprotected’ left turns; You can never really be sure when someone is going to run a red light.
Have they cracked backing up with trailers yet?