18H·

Car manufacturers and self-driving cars

Somehow, posts about $TSLA (+2,64 %)
$MBG (+1,52 %)
$VOW (+2,24 %) and the assistance systems always seem strange to me, so I asked ChatGPT.

(Prompt: "self-driving cars Mercedes vs waymo vs tesla"

AI says:

Investor perspective:

Whether self-driving cars based on cameras will get approval in Europe for fully autonomous driving remains to be seen, because there is definitely a lot of criticism of the pure AI+ camera strategy.

It may work faster in America and Asia, where the media can be bribed even better and accidents are less relevant than sales figures, but in the EU a very large disinformation campaign would have to work against the will of the European lobby, which uses the expensive lidar systems.

Conclusion again from ChatGPT:

Tesla, as usual, has not unlocked another level on the launch date now, so they are still level 2, right? https://www.handelsblatt.com/technik/it-internet/autonomes-fahren-tesla-startet-robotaxi-dienst-mit-einigen-einschraenkungen/100135657.html

Luca

@Lvca

4D·

WAYMO: "TESLA WAS RIGHT, WE WASTED YEARS"

$TSLA (+2,64 %) Waymo has released a new research paper that basically says this: if you want smarter self-driving cars, you simply need to throw more data and computing power at the problem.

That's exactly what Tesla has been doing for years - ditching traditional methods, building gigantic AI supercomputers and collecting data non-stop with their fleet of vehicles.

Corry Wang, who heads up the computing infrastructure at Anthropic, put it in a nutshell:

"Waymo was wrong. Tesla was right."

Badass.

If this proves to be true, Tesla could be on the verge of a GPT-3-like level-up in autonomous driving - while Waymo is just watching in the rear-view mirror.

15 Comentarios

Nom@user28461

18H

Let me help you with ChatGPT:

Tesla vehicles rely exclusively on so-called **vision-based systems**, i.e. cameras supported by **AI models**, to recognize the environment in their autonomous driving approach. This also works at night, for several reasons:

---

### 🌙 Why does Tesla's camera system also work at night?

#### 1. **High-resolution cameras with good low-light performance**

* Tesla's cameras are specially calibrated and have a high sensitivity to low light.
* Similar to modern smartphone cameras, they can deliver usable images even in low light conditions.

#### 2. **Processing with AI (Neural Nets)**

* The software also processes noisy or dark images using neural nets that have been trained to recognize objects in many different lighting conditions - including night, twilight, headlights, streetlights, etc.

#### 3. **Utilization of vehicle lights**

* The vehicle's headlights illuminate the road in front of the car and enable the cameras to detect important elements such as road markings, pedestrians or obstacles.
* Rear lights of other vehicles help the software to identify them correctly.

#### 4. **Infrared elements (passive)**

* Some cameras have some infrared capabilities, which improves night vision performance - although Tesla does not use active infrared or night vision cameras like some other manufacturers.

---

### 🚫 Why no radar or LiDAR?

Tesla deliberately decided against **Radar** and **LiDAR** because:

* Vision-only AI should extract all necessary information from image data in the long term.
* Radar in earlier versions sometimes led to contradictory results with the camera data.
* LiDAR is very expensive and, according to Tesla, not necessary if the AI is sufficiently trained visually.

---

### Conclusion

Tesla can also drive safely at night** because:

* The cameras can "see" enough with headlights and modern image sensors.
* The AI is trained to reliably process relevant information even in the dark.

•

Moritz@leveragegrinding

18H

@user28461 prompt? otherwise you can simply leave out the "nevertheless Lidar is significantly better" at the end.

•

Nom@user28461

18H

@leveragegrinding Prompt: "How can Tesla drive at night if they only rely on cameras?"

••

Nom@user28461

18H

@leveragegrinding Of course LIDAR is good, but it is expensive. So get rid of it if you can do it well enough otherwise.

••

Moritz@leveragegrinding

18H

@user28461 you simply ask

"why is tesla's camera-based self-driving controversial and what do the accident rates say?
maximum 100 words"
"Tesla's camera-based system ("Tesla Vision") is controversial because it dispenses with lidar and radar and relies purely on neural networks - which can be prone to errors in poor visibility or unexpected situations. Critics complain about a lack of redundancy and safety risks. Tesla's accident statistics show lower rates with Autopilot activated, but are difficult to compare as they are mostly based on highways (fewer accidents) and are published by Tesla itself - independently verifiable data is often missing."

•

Moritz@leveragegrinding

17H

@user28461 Doing without a technology just because it's still expensive doesn't make that much sense. I'd rather my cheaper car didn't have the function than be lulled into a false sense of security and get it cheaper.

I'm really looking forward to a registration in Germany, because then there will be neutral accident figures.

•

Nom@user28461

17H

@leveragegrinding And now? Shall we show each other how to get ChatGPT to say what we want to hear?

You said it seems Spanish to you - if you had wanted to, it would have seemed less Spanish to you now, but your defense/resistance is just too strong. You WANT to think that.

Investors are simply betting that the vision-only system is more competitive (because it is cheaper) and is good enough or even better than humans due to the AI / database - so they buy. Nothing Spanish.

••

Nom@user28461

17H

@leveragegrinding Then you wait on the sidelines until never-never day, while investors (!) think long-term and anticipate this development for many years / assume that it will happen. It does not make sense to enter the market until the company has grown / become valuable.

•

Robin@robiin

17H

@user28461 Oh Nom what can you do. Some people here in the forum don't realize why Tesla specifically reports the system as level 2, while it is far more capable than comparable level 2 systems. We can't get against the wall :D

•

Nobody@Nobody_123

16H

@robiin But now I'm curious, why does Tesla only report the system as level 2?

••

Robin@robiin

15H

@Nobody_123 If you were to report it higher, you would have to guarantee more in case something happens. So it's completely unnecessary if you can continue development even if the system is only officially registered as a level 2 system.

•

Ver todas las 3 respuestas adicionales

PaulPK@PaulPK

12H

Europe is lost and irrelevant - they will allow it after it becomes clear that it is 10 times safer than humans - unless the wrong parties block it again.

•