We Need FSD Full Self-Driving to Be Imperfect

[Darthamerica]

In the pursuit of fully autonomous vehicles, we must accept that no complex technology can be flawless from the outset. Tesla’s Full Self-Driving (FSD) system is a prime example. If we demanded perfection before release, such technology might never see the light of day. By embracing the imperfections and risks that come with scaling FSD, Tesla takes a bold first-mover approach.

Real-world incidents—including crashes and, unfortunately, fatalities—generate invaluable data, which Tesla uses to iterate and improve. This pragmatic approach accelerates the technology’s maturity, unlike the cautious stances taken by more risk-averse competitors. Over time, other automakers will either develop their own systems or license Tesla’s, riding the wave of innovation.

The real impact is exponential growth in self-driving capability. By risking imperfection today, we pave the way for mainstream adoption tomorrow. As the technology improves, the world will gradually come to accept self-driving cars as a norm, transforming transportation.

Be bold, be cautious, enjoy the ride.

Sponsored

 

[Mark Shaw]

Life is imperfect- and until all vehicles on the road can communicate with each other FSD will be imperfect. I hope to be long gone before that happens.

 

[cybercoffee]

While I largely agree, I think FSD on the CT is completely on par with FSD on my other teslas. I've used it for a couple hundred miles yet and no close calls. I'm very pleased with it and think that they really took the time to release a fully baked product.

 

[Jack27]

So here’s my maybe uneducated on the subject question.
when I went to Austin I got into and saw many human-less taxis driving around now in Los Angeles I’m seeing allot of the test cars driving with no one.
yes I do know they have someone monitoring them that can shut them down but would be way to late if a problem happen. how is that technically that good when using my FSD I feel I need to take over atleast 1 time a use ?

 

[Darthamerica]

While I largely agree, I think FSD on the CT is completely on par with FSD on my other teslas. I've used it for a couple hundred miles yet and no close calls. I'm very pleased with it and think that they really took the time to release a fully baked product.

I think it’s good as a driver assistance feature. I do see a path to actual FSD, but as is, it’s still got a way to go. From what I’ve experienced with CT FSD(S) over the last week I would say that it has the potential to be at least as good as a human driver over time.

The biggest challenge are all of the corner cases and Teslas as is are going to be limited to whatever the current SW version has been trained to handle. Since there are near infinite numbers of corner cases, there will always be situations these Teslas can’t handle with the current SW implementation(offline learning).

 

[thech0zen1]

Real-world incidents—including crashes and, unfortunately, fatalities—generate invaluable data, which Tesla uses to iterate and improve.

You are an absolute psychopath.

There should be federal regulations in place for FSD until it’s safe for drivers and pedestrians. People’s lives are not a commodity used to improve technology.

 

[Chris9702L]

So wrong! I’m in the aviation field and most major improvements and improvements are from accidents. An example is the Delta crash in Texas from a microburst. Little was known about them but the accident brought awareness and systems to notify of one is occurring.

 

[Ray in montana]

In the pursuit of fully autonomous vehicles, we must accept that no complex technology can be flawless from the outset. Tesla’s Full Self-Driving (FSD) system is a prime example. If we demanded perfection before release, such technology might never see the light of day. By embracing the imperfections and risks that come with scaling FSD, Tesla takes a bold first-mover approach.

Real-world incidents—including crashes and, unfortunately, fatalities—generate invaluable data, which Tesla uses to iterate and improve. This pragmatic approach accelerates the technology’s maturity, unlike the cautious stances taken by more risk-averse competitors. Over time, other automakers will either develop their own systems or license Tesla’s, riding the wave of innovation.

The real impact is exponential growth in self-driving capability. By risking imperfection today, we pave the way for mainstream adoption tomorrow. As the technology improves, the world will gradually come to accept self-driving cars as a norm, transforming transportation.

Be bold, be cautious, enjoy the ride.

Very well put. Thank you.

 

[HaulingAss]

In the pursuit of fully autonomous vehicles, we must accept that no complex technology can be flawless from the outset. Tesla’s Full Self-Driving (FSD) system is a prime example. If we demanded perfection before release, such technology might never see the light of day. By embracing the imperfections and risks that come with scaling FSD, Tesla takes a bold first-mover approach.

Real-world incidents—including crashes and, unfortunately, fatalities—generate invaluable data, which Tesla uses to iterate and improve. This pragmatic approach accelerates the technology’s maturity, unlike the cautious stances taken by more risk-averse competitors. Over time, other automakers will either develop their own systems or license Tesla’s, riding the wave of innovation.

The real impact is exponential growth in self-driving capability. By risking imperfection today, we pave the way for mainstream adoption tomorrow. As the technology improves, the world will gradually come to accept self-driving cars as a norm, transforming transportation.

Be bold, be cautious, enjoy the ride.

I think you mean well, but we don't need fatalities to improve FSD, the improvements happen from disengagements. Because disengagements are used to show where FSD needs more video data to train on. The idea of having supervision 100% of the time is that the driver can disengage if there is a safety issue. If the driver doesn't disengage before a serious accident, it's the driver's fault for not paying attention.

Statistics plainly show that a human driver with FSD is safer than a human driver without FSD. That means there is no human cost to the current development path. It's actually already saving lives, thanks to the people using and supervising it.

 

[Darthamerica]

I think you mean well, but we don't need fatalities to improve FSD, the improvements happen from disengagements. Because disengagements are used to show where FSD needs more video data to train on. The idea of having supervision 100% of the time is that the driver can disengage if there is a safety issue. If the driver doesn't disengage before a serious accident, it's the driver's fault for not paying attention.

Statistics plainly show that a human driver with FSD is safer than a human driver without FSD. That means there is no human cost to the current development path. It's actually already saving lives, thanks to the people using and supervising it.

Fatalities are unfortunately a common occurrence in the development of cutting-edge technologies. For instance, with the advent of FSD, there have been and will continue to be fatalities. I am not implying that FSD is inherently flawed or that it will not eventually become fully autonomous. My intention is simply to highlight the inherent risk associated with such technologies.

 

[Darthamerica]

You are an absolute psychopath.

There should be federal regulations in place for FSD until it’s safe for drivers and pedestrians. People’s lives are not a commodity used to improve technology.

You’d be surprised.

 

[BornToFly]

You are an absolute psychopath.

There should be federal regulations in place for FSD until it’s safe for drivers and pedestrians. People’s lives are not a commodity used to improve technology.

If you remove suicide from the statistics, human drivers cause more death every year than guns (and injury rates are astronomical). FSD has already decreased the accident rate and death rate compared to human drivers. The goal is to get rid of human drivers, which are surprisingly bad due to inattention, and very poor judgment (speeders, DUI - which is really bad now with more pot users, etc). To say it can't be released until it is perfect means a lot more death. Personally, I would like to see much stronger penalties for injuring or killing someone with a car. If you hit and kill my wife while she is on her bike, you need to spend the rest of your life in jail.

 

[HaulingAss]

Fatalities are unfortunately a common occurrence in the development of cutting-edge technologies. For instance, with the advent of FSD, there have been and will continue to be fatalities. I am not implying that FSD is inherently flawed or that it will not eventually become fully autonomous. My intention is simply to highlight the inherent risk associated with such technologies.

I think I'm talking right past you.

Why do you think fatalities are resulting from the FSD development program? The statistics clearly show supervised FSD results in less fatalities than a human driver alone!

 

[eswimm]

You are an absolute psychopath.

There should be federal regulations in place for FSD until it’s safe for drivers and pedestrians. People’s lives are not a commodity used to improve technology.

Ultimately, any vehicle safety improvement brings trade-offs. Seat belts and airbags have been responsible for vehicle deaths but are ultimately responsible for saving a lot more lives.

Are FSD and similar autonomous technologies successes or failures if they halve the vehicular fatality rate? I think data has pretty clearly shown that FSD in tandem with an attentive driver is safer than the driver alone and FSD with an inattentive driver is still safer than the inattentive driver alone.

Tesla has numerous checks in place to enforce driver attentiveness.

Sponsored