Understanding What ‘Good Enough’ Means for Auto Wipers and the Limitations of AI/ML Solutions
ML progress in auto wipers has been slow, despite the huge trove of feedback data available. This suggests that the promise of AI/ML is weak and may not be able to do all that much. Auto Highbeams have seen improvement recently, but overall, wipers still don't work as well as they should. The problem is compounded by the fact that Tesla has prioritized other projects over perfecting wipers, and different definitions of 'perfect' mean that it's unlikely to ever be perfect for everyone. This highlights the importance of understanding what 'good enough' means for a particular product or service before relying on AI/ML solutions.
The data available for auto wipers is vast, but it's not enough to make them perfect. The data needs to be properly organized and labeled in order to be useful. This requires a lot of manual effort, which can be costly and time consuming. Additionally, the data needs to be constantly updated as conditions change. This means that AI/ML solutions are only as good as the data they're given, and if the data isn't up to date, then the results won't be either.
The other issue with relying on AI/ML for auto wipers is that it's difficult to define what 'perfect' looks like. Different drivers have different preferences when it comes to how their wipers should work, so it's impossible to create a single solution that works for everyone. This means that AI/ML solutions will always be limited in their ability to provide perfect results.
Finally, Tesla has prioritized other projects over perfecting auto wipers. This means that while they may have the best data and algorithms, they don't have the resources or time to dedicate to making sure their wipers are perfect. This further highlights the importance of understanding what 'good enough' means before relying on AI/ML solutions.
Overall, progress in auto wipers suggests that AI/ML is still far from being able to do all that much. The data needs to be properly organized and labeled in order for it to be useful, and even then, it's difficult to define what 'perfect' looks like. Additionally, Tesla has prioritized other projects over perfecting auto wipers, which limits the potential of AI/ML solutions. It's important to understand what 'good enough' means before relying on AI/ML solutions, as they may not be able to deliver perfect results every time.
How long has it been since the M3 was released?
It was released in 2018, so it has been nearly 5 years.
What data does Tesla have access to regarding auto wipers?
Tesla has access to a trove of feedback, including when a user overrides auto and picks a setting (or turns on / off). This provides them with rich data for machine learning.
Has there been any progress with auto wipers?
There has been some improvement, but not as much as expected given the amount of time that has passed and the amount of data available.
Are there any other features that have seen improvement recently?
Auto Highbeams has seen some improvement recently.
What are the issues with auto wipers?
When using FSD/AP, they are locked on, and none of the manual modes have fine adjustments. In heavy rain, they either go into unnecessary high gear or sit dormant until woken up. In light rain, they go crazy when there's exterior light sources but, when it's dark and no lights in its eyes, the windshield gets coated before it ever considers a swipe. In light snow with little moisture, they pretty much run on a dry windshield which just makes noise and annoys until turned off.
Why is AI/ML progress so slow?
AI/ML progress is slow because development efforts are often put on the back burner in favor of more fun projects. Additionally, different definitions of “good enough” can lead to disagreements between CFOs and buyers. Finally, AI/ML progress is slow because of the complexity of the data and the difficulty in training models to accurately interpret it.