- cross-posted to:
- technology@beehaw.org
- cross-posted to:
- technology@beehaw.org
Yesterday evening, Tesla reported first-quarter earnings for 2025, and they were abysmal: Profits dropped 71% from the same time last year.
Musk sounded bitter on the call with investors that followed, blaming the company’s misfortune on protesters who have raged at Tesla dealerships around the world over his role running DOGE and his ardent support of far-right politicians.
“The protests that you’ll see out there, they’re very organized. They’re paid for,” he said, without evidence.
Psst… you know that Waymo is also looking into AI models that don’t use lidar? And it’s not just Tesla and Waymo now either, there’s some others as well.
https://waymo.com/research/emma
Thats not to say that it will be a success, but ya know, maybe it’s actually possible, and even Waymo seems to think it’s possible and worth researching. Thats a lot of people looking into something thats clearly “fundamentally flawed”
And found it wanting compared to systems that incorporate LiDAR
I was just looking through the linked PDF and I’ll actually retract what I said earlier about not finding it wanting, and them just listing limitations.
In the research paper linked it is more explicit
“Despite these promising results, EMMA is not without its limitations. In particular, it faces challenges for real-world deployment due to: (1) limitations in 3D spatial reasoning due to its inability to fuse camera inputs with LiDAR or radar, (2) the need for realistic and computationally expensive sensor simulation to power its closed-loop evaluation, and (3) the increased computational requirements relative to conventional models. We plan to better understand and address such challenges in future work.”
This is more clear that it’s not just listing limitations, but it also finds it not doing as well because of it.
That I’d call wanting.
deleted by creator
As I understand it, it is fundamentally possible, but at this point one of the worst bottle necks is the physical processing power required to analyze the video feed. Millisecond paced analysis of multiple aspects of multiple HD video feeds in real time by “AI” still takes some pretty heavy duty computational hardware at this point, which is impractical in a real world vehicle. It is prudent to do the research now though to be ready for the day when the hardware catches up. That said, I reckon that Waymo is in a much better position to do this research than than Tesla. Being under Google/Alphabet, they have access to the nearly 2 decades worth of Google Maps and Google’s machine vision training work, among other things, like not being a meme stock company headed by an emperor whose clothes are beginning to dissapear.
Some people are jumping on LLM hype to try to shoehorn it into a hot technology? What a surprise!
Research is one thing, the production line is another.
It is fundamentally flawed because everyone that tries it ends up killing motorcyclists.
If they can kill motorcyclists faster than motorcyclists can kill themselves that is impressive.
I want you to seriously think for a moment if you would say that to someone who has lost a friend or a father, who was doing nothing wrong, to a preventable issue resulting from Elon Musk being a penny pinching cunt.
I’d not intentionally be a dick to someone who obviously has learned that lesson the hardest way.
I’d absolutely say that to someone random in the hopes that it saves someone’s friend or brother. If you’re riding a motorcycle without always having how dangerous they are near the front of your mind, you’re not long for this world.
Assisted driving is nowhere near the primary cause of motorcycle deaths. Phones are worse, by at least a couple orders of magnitude.
(And yeah, removing lidar was fucking stupid)