![]() |
I installed Tesla Level Autopilot on my 7th Gen ES.
Thanks to Comma.ai - I now have Tesla level highway autopilot installed on my Lexus ES300h 7th Gen. Please take a look at the video I made below. The system blows the stock lane keep out of the water. The car can literally go a hundred miles on the highway now without me even touching the wheel. It’s truly special.
|
I'm going to be landing at LAX on September 2nd. I use the 405 South to CA-73 South, exit at Newport Beach. Just figured I'd let you know. ;)
|
starting price of $2,199 and $2,799 for the crosscountry one. wow. This company was founded by a hotshot Playstation hacker and made a huge flash a few years ago then silence. I thought it might had went bankrupt or sold off long ago.
|
I plan to keep my hands on the steering wheel for now... :driving:
|
yeah me too hehe. the self driving feature is cool and all but it is training the user to be so dependent on it that it is a recipe for a disaster in the making. study after study has shown when it faces driving situations the AI doesn't know how to deal with it will pass the control over the human and by that time it is usually too late. That seems to be the default error handling action that basically passes the liability back to the human driver.
For Tesla it seems when the outcome is good they take the credit otherwise it's the human driver's failings. I have no doubt that with enough R&D, testings, and sacrificed lives in long term the AI driving will be a lot safer than humans. I am not willing to be their test subject while paying for it. |
Thats the latest one. The one i have is 1099.
|
Originally Posted by ionian
(Post 11112088)
yeah too me hehe. the self driving feature is cool and all but it is training the user to be so dependent on it that it is a recipe for a disaster in the making. study after study has shown when it faces driving situations the AI doesn't know how to deal with it will pass the control over the human and by that time it is usually too late. That seems to be the default error handling action that basically passes the liability back to the human driver.
For Tesla it seems when the outcome is good they take the credit otherwise it's the human driver's failings. I have no doubt that with enough R&D, testings, and sacrificed lives in long term the AI driving will be a lot safer than humans. I am not willing to be their test subject while paying for it. This is level 2 as well. There is actually a small IR blaster and camera which tracks the user’s eyes. If you’re distracted the system disengages. It’s a total game changer. |
Interesting product. Keep us update with your periodic posts about on on-going experience. We want to know the good, the bad, and the ugly.
|
Yep been following that guy for years and his videos and progress. Seems alright. He started with the original iphone jail break if i'm not mistaken, in 2007. I was waiting to see someone uses this on a Lexus product. i remember in 2016ish or so when he first installed it on a corolla. whenever the corollas started coming out with the hardware/cameras.
and probably a bit misleading to call it tesla self driving. maybe tesla-ish-like self driving. |
Yeah, this product is a cool project for those with tinkering mindset. It was kinda affordable at $1k for last gen model but now the current gen just doubled the price overnight and that just moves it out of the range for more potential buyers. To improve its accuracy in more driving conditions and situations it will need a lot of data to train the AI/ML. Tesla is king in this space and they have massive data points compare to their competitors. I don't see how comma.ai can compete. They are a niche player in the end and most developers won't waste their time with it. Best George can do is hope it will get good enough so he can cash out to a company looking to enter the market and don't want to build one from ground up. With comma.ai raising the price their data points went from small to smaller.
Data is king for self-driving vehicles. |
Originally Posted by ionian
(Post 11112333)
Yeah, this product is a cool project for those with tinkering mindset. It was kinda affordable at $1k for last gen model but now the current gen just doubled the price overnight and that just moves it out of the range for more potential buyers. To improve its accuracy in more driving conditions and situations it will need a lot of data to train the AI/ML. Tesla is king in this space and they have massive data points compare to their competitors. I don't see how comma.ai can compete. They are a niche player in the end and most developers won't waste their time with it. Best George can do is hope it will get good enough so he can cash out to a company looking to enter the market and don't want to build one from ground up. With comma.ai raising the price their data points went from small to smaller.
Data is king for self-driving vehicles. |
I can't speak for others, but I actually enjoy driving my cars - and being in control - especially my Corvette. The "wind in your hair" experience. No automatic driver system for me... A bunch of BS IMO...
:driving: |
I'm considering the version 3 after some time. The version 2 has had so many returns due to the hardware dying that it wouldn't be worth dropping $1k on it.
|
I really like the concept of this product of decoupling the driving assist functionalities from car manufacturing. Many car manufacturers are too conservative, whether it's for liability reasons, for safety, or for the lack of algorithmic development. Comma is not the only company offering dedicated driving assist technologies though, but it's certainly the riskiest one and an immature product in my personal opinion. I'd recommend anyone to carefully think about this product before committing to it and always leave your hands on the steering wheel. Here is why:
The main reason is the serious lack of hardware. Tesla, for instance, has 8 cameras, 12 ultrasonic sensors, and optionally radar. The latest Comma has 3 cameras and leverages the car's original radar. Tesla has a proprietary computational chip offering 144 trillion operations per second (TOPS). The latest Comma, in comparison, uses a Qualcomm 845 smartphone chip with 3 TOPS. The most critical drawback comes from its lack of redundancy and error tolerance. Tesla's car-specific chip offers full control redundancy whereas Comma's off-the-shelf chip can possibly crash under a tiny event upset just like the occasional freeze of your cellphone. I don't mean it's impossible to achieve great driving assist with meager hardware; in fact, several L2 driving assists systems have similar TOPS or rely on pure camera perception. However, no other system is as vulnerable as Comma which combines the bare minimum hardware in every aspect as it's intended so. The second reason is the serious lack of training data. A good training is the foundation of any good AI/ML system. Tesla has multi-BILLION miles of training data collected by all Teslas on the road. Major car manufacturers like Ford, Toyota, etc. have a data volume that is about one order of magnitude lower. What about Comma? I don't know, but I'm sure it's not even remotely close because a start-up doesn't have the same resources and money. Also, I would suspect that the system is much less road tested or rigorously verified through regression after any version update. Admittedly, when you look into Comma's documentation (the newer version of Openpilot is largely based on the same principle), the control logic is not at the same level as Tesla. More advanced L2 ADAS system like Tesla builds an environment map to make global decisions (which is more aware of the surrounding condition other than merely the frontal car), whereas Comma implements ACC and lane centering separately in a modular fashion. This strategy is almost identical to the car's original ACC+LTA, but with a smoother and more capable LTA. At this point, I truly hate the advertisement of Comma which positions it as a rival of Tesla because they are fundamentally different. It's probably okay to drive Tesla Autopilot/Ford CoPilot/Cadillac SuperCruise hands-free (although I wouldn't) as these offer a coherent steering+acceleration control and are more environment-aware, but it would be much riskier to drive a simple, decoupled L2 system like Comma hands-free. There are some other concerns regarding its reverse-engineering-enabled integration but I think I've said enough. I'm probably always on the defensive side of aggressive ADAS but particularly so towards these less verified products. That's just my 2 cents from the perspective of a hardware security engineer and hearing my wife' who actually works in an autonomous driving company as a verification engineer always seeing the worst case :) |
Thanks for sharing your insight and expertise on the matter. As much as some people like to trash Tesla I appreciate what they are trying to do. I think part of the problems with autonomous driving is two folds from mindset perspective:
1) the overselling of its current benefits 2) consumer optimistically over reliant on their vehicle's capability and not fully understanding the system limitations (e.g. people napping in their Teslas while barring down the high way) I am sure the people in this forum are very safe and courteous drivers and they would never be driving distracted, aggressively, or under influence. We all have our anecdotal stories as to which states have the worst drivers out there on our the neck of the woods. I live in Delaware and for me the most aggressive drivers I see on my local roads are the ones from NJ and NY. According to insurify (https://insurify.com/insights/states...-drivers-2021/) Ohio drivers are the worst drivers in the nation for 2021. The sooner AI makes driving safer the better it is. |
| All times are GMT -7. The time now is 09:28 PM. |
© 2026 MH Sub I, LLC dba Internet Brands