Again, you're not understanding the branches of AI unique and required for androids (NLP and embodied cognitive models) - which Tesla has none of in FSD. So they don't have this tech but would have to develop it. For FSD, the branches of AI is sensor fusion, vision, machine learning, deep learning, reinforced learning and control systems which are all needed in some form or manner for androids.

That is totally incorrect. embodied robots need to react in real-time and learn in real-time based on multi-sensory feedback. Going to the cloud (even with edge computing) is too slow and simply too much data to transmit. You know that FSD does not require an internet connection right to make decisions? It's all done locally (often referred to as Hardware 4.0), and data is only sent back to Tesla for other reasons.

You're also mistaken that these models can't be run locally. We have lightweight neural networks, convolutional network networks, quantized neural nets, reinforcement learning with proximal policy optimizations, tinyML, etc etc. These are widely used in many applications today.

Never said that, I said it adds additional complexity using a set of AI technologies outside and on top of what was developed for FSD. You seem to think Tesla already has the full 'AI stack' with FSD and just need an additional sensor input and volia! That is far from the reality. All I am staying is that none of Tesla's showpiece have actually demonstrated embodied cognitive models nor understanding of natural language. Robots which can be remotely controlled, with precision and balanced walking is nothing new and not interesting.

 

^^ great post 703.

humanoid robots will mostly be used initially for specific repetitive boring tasks, taking drudgery out of humans' lives. they will have to react to circumstances, for sure, but in a huge number of use cases, what they're reacting to will be pretty limited (like operating on a factory floor or a warehouse).

they may also be used, gasp, via remote human control. there are many use cases for this too, like going into dangerous situations or getting someone out of a burning building.

this idea that they'll be partying and serving drinks any time soon (because it's just like fsd lol) is a fantasy.

 

Cybercab will definetly be confied geographically, this is probably why it will come to market in next few years.

I mean how do you think it will not - it requires a home base to run? Home base will have to be built by Tesla.

 

literally almost every car demo on a car show is rigged. They are showing preproduction hand built vehicle that is 2-3-4-5 years away from final version.

Rimac had a autonomous taxi demo 4 months ago, and local media tore through him due to preproduction version being remote controlled (as if production version would drive in a warehouse floor among audience), which was in some ways political (they dont like the fact he is famous now and got rich quick) - where European and US media were mesmerized with the same demo.

Which again, makes it all silly and political. Unfortunately this is the world we live in - and while Musk maybe brought pieces of it to himself due to twitter, Rimac certainly did not - yet he got the same treatment.

What was interesting is the concept - 2 seater, no controls, home base that will auto clean and power. Definitely locked to a city where they have detailed mapping done themselves.

 

What if Tesla added a steering wheel and pedals to the Cybercab and that is the production vehicle of the $30k car..? I thought the look of it was really cool. If Musk announced that, I think the stock would go crazy.

 

That was my dark horse bet, but the train-van came out instead

 

Define sensor fusion, and tell me what the sensors are.
FSD does not require an internet connection to operate services like Waymo 100% do.

 

I just wonder if it might be part of a different set of announcements in the future. That event, focused so much on robots, AI, etc. may not have been the time to unveil it. But if you're already making the car and just need to add a few components...... makes a lot of sense to me.

 

Cybercab will require internet connection.

It also means nothing.

 

It will not require a connection to drive down the road in other words if connection drops it won't have to pull over.

 

Nor does Waymo.

“Driverless vehicle company Waymo made waves with its recent launch of a paid, autonomous ride-hailing service in San Francisco. However, the company has deliberately excluded 5G technology from its core offering, opting instead for on-board computing.In response to inquiries from Light Reading, a Waymo spokesperson revealed that the company's self-driving vehicles are not reliant on a continuous wireless connection for safe operation, nor do they use vehicle-to-vehicle (V2V) communications.

"We believe the most optimal way to operate autonomous driving technology is for the compute to be on board and for it to make decisions, without needing to rely on cell signals and remote operators," the spokesperson wrote. "We don't want to have a situation where, say, if the car lost cell connection, it couldn't make a left turn. Therefore, everything from custom made maps to robust neural nets that inform our perception and routing runs on board."

Waymo's vehicles are equipped with a cellular connection, which is used to supplement the autonomous driving system with real-time road conditions, said the spokesperson. The company declined to provide any more details on its usage of cellular connections, including whether the company's cars use 4G or 5G and which service providers they might connect with.”

https://www.lightreading.com/iot/way...the-background

 

That's good to see thanks for that, last I checked Waymo couldn't drive without being tethered I thought that was very dangerous.

 

It’s almost impossible to keep track. Everything is evolving so quickly.

 

https://www.opentrain.ai/glossary/sensor-fusion

rest you can work it out.

 

It is not at all.

Little chance of anything shown here will be in production in next 2 years.