this post was submitted on 01 Aug 2025
1093 points (99.1% liked)
Technology
73567 readers
2922 users here now
This is a most excellent place for technology news and articles.
Our Rules
- Follow the lemmy.world rules.
- Only tech related news or articles.
- Be excellent to each other!
- Mod approved content bots can post up to 10 articles per day.
- Threads asking for personal tech support may be deleted.
- Politics threads may be removed.
- No memes allowed as posts, OK to post as comments.
- Only approved bots from the list below, this includes using AI responses and summaries. To ask if your bot can be added please contact a mod.
- Check for duplicates before posting, duplicates may be removed
- Accounts 7 days and younger will have their posts automatically removed.
Approved Bots
founded 2 years ago
MODERATORS
you are viewing a single comment's thread
view the rest of the comments
view the rest of the comments
I don't know, most experimental technologies aren't allowed to be tested in public till they are good and well ready. This whole move fast break often thing seems like a REALLY bad idea for something like cars on public roads.
Well, the Obama administration had published initial guidance on testing and safety for automated vehicles in September 2016, which was pre-regulatory but a prelude to potential regulation. Trump trashed it as one of the first things he did taking office for his first term. I was working in the AV industry at the time.
That turned everything into the wild west for a couple of years, up until an automated Uber killed a pedestrian in Arizona in 2018. After that, most AV companies scaled public testing way back, and deployed extremely conservative versions of their software. If you look at news articles from that time, there's a lot of criticism of how, e.g., Waymos would just grind to a halt in the middle of intersections, as companies would rather take flak for blocking traffic than running over people.
But not Tesla. While other companies dialed back their ambitions, Tesla was ripping Lidar sensors off its vehicles and sending them back out on public roads in droves. They also continued to market the technology - first as "Autopilot" and later as "Full Self Driving" - in ways that vastly overstated its capabilities. To be clear, Full Self Driving, or Level 5 Automation in the SAE framework, is science fiction at this point, the idea of a computer system functionally indistinguishable from a capable human driver. Other AV companies are still striving for Level 4 automation, which may include geographic restrictions or limitations to functioning on certain types of road infrastructure.
Part of the blame probably also lies with Biden, whose DOT had the opportunity to address this and didn't during his term. But it was Trump who initially trashed the safety framework, and Telsa that concealed and mismarketed the limitations of its technology.
You got me interested, so I searched around and found this:
So, if I understand this correctly, the only fundamental difference between level 4 and 5 is that 4 works on specific known road types with reliable quality (highways, city roads), while level 5 works literally everywhere, including rural dirt paths?
I'm trying to imagine what other type of geographic difference there might be between 4 and 5 and I'm drawing a blank.
Yes, that's it. A lot of AV systems are dependent on high resolution 3d maps of an area so they can precisely locate themselves in space. So they may perform relatively well in that defined space but would not be able to do so outside it.
Level 5 is functionally a human driver. You as a human could be driving off road, in an environment you've never been in before. Maybe it's raining and muddy. Maybe there are unknown hazards within this novel geography, flooding, fallen trees, etc.
A Level 5 AV system would be able to perform equivalently to a human in those conditions. Again, it's science fiction at this point, but essentially the end goal of vehicle automation is a system that can respond to novel and unpredictable circumstances in the same way (or better than) a human driver would in that scenario. It's really not defined much better than that end goal - because it's not possible with current technology, it doesn't correspond to a specific set of sensors or software system. It's a performance-based, long-term goal.
This is why it's so irresponsible for Tesla to continue to market their system as "Full self driving." It is nowhere near as adaptable or capable as a human driver. They pretend or insinuate that they have a system equivalent to SAE Level 5 when the entire industry is a decade minimum away from such a system.