electric cars.

Kitchen Knife Forums

Help Support Kitchen Knife Forums:

This site may earn a commission from merchant affiliate links, including eBay, Amazon, and others.
By releasing a product into the wild
Yeh, "autopilot" is a huge marketing stunt .. but it may have really paid off. We'll see. Access to real data is huge.

Tesla also doesnt use LiDAR.
Didn't Tesla drop lidar because they think they can do it with just camera? This was recent.

I didn't know Google is that far ahead "officially", I'm a bit hesitant to call them that far ahead due to exactly what you said about Tesla's access to training data .. and the part that Tesla's are already out on the road, and I haven't seen google one in practice yet.

Google certainly has much trust from past achievements though. I'm not sure why, but if they would release this tech like they did maps and gmail so everyone could use it .. that would really jolt the industry forward. One can dream.
 
while attention has uses outside of language models, it is not even debatable that it has had the largest effect in language by a mile IMO.

Sure. I agree :)

I am just noting that attention networks aren't exclusively used by NLP (although they were developed there).... and there are tasks within autonomous driving that could benefit from attention networks.

saying that the hype of deep learning (as it currently exists) is disproportionate to its actual results in autonomous driving is IMO, about 10 degrees too kind. I say this as a person with multiple DL models in production.

Yes and no.

Without being too emphatic, all I am trying to communicate is that deep-learning is not a silver bullet. Not necessarily to you... but to anyone who might read the post. Autonomous driving... deep learning... AI.... big data... a lot of these terms get thrown around and conflated in popular culture. I am sure you and I might agree on what is realistic and what are unfounded expectations

However, it is well worth recognising the contribution deep learning has made to computer vision at a practical level. You cant really do high-level computer vision research without stubbing your toe on deep learning. This very much includes autonomous driving. Any autonomous vehicles using machine vision will be running some sort of deep learning network to interpret the video stream. I agree... forget the hype. In some senses deep learning for perception tasks is routine and even mundane. But since it has become almost a prerequisite, I feel one has to be forced to acknowledge that its impact on autonomous driving has been enabling and significant (albeit, not the first coming of robotic jesus).



Im not 100% sure what your argument is.

I think where we differ is the role of technological 'step changes'.


autonomous driving is probably a solvable task in the long run but so are lots of problems that got worked on for decades. that's my suggestion here; without a major change, it doesnt matter how many things you try and bolt together like an erector set cars wont drive themselves

Like you say! Time is progress! I think this the stronger position.

Artificial neural networks are what? More than half a century old? Backpropagation decades old... 'Modern' stochastic gradient descent and running networks on dedicated hardware (GPU/TPUs) are about 15 years old? All these leaps can smash one problem and subsequently open many doors... but they hardly ever solve very general and difficult problems.

You dont need particularly general AI to solve autonomous driving (my previous silly remark about changing light bulbs). But the system can't be a narrow AI either. Writing a championship level chess algorithm, or finding cats in pictures... or translating literature from spanish to english are all difficult problems. But we can solve those problems with highly optimised algorithms for a dedicated task (narrow AI). Major leaps have a more profound effect on these well defined tasks. Driving is a mix of complex and interacting requirements... so the system has to reflect that. In other words, be slightly less narrow.

I think the secret sauce will be in learning how to bolt things together. Not one great leap. This is my point. Sure... very specific major changes will help... but I just dont imagine a major leap in one subsystem solving the global problem. Like I mentioned earlier you could crudely break down autonomous driving into perception, planning and control - indeed any robotics problem. Deep learning has excited people because it is having a profound impact on perception systems. But that doesnt necessarily solve the planning problem... you might use deep learning... you might not. You can have a really technical and complex discussion about localisation without even mentioning deep learning... and so it goes.

And this is where we are. We have really competent building blocks that make up something like 95% of an autonomous car... but that 5% is really important and it is going to be a hard slog completing. It is likely this will come through improvements in many fields. Some major. Many more incremental.
 
Didn't Tesla drop lidar because they think they can do it with just camera? This was recent.

:)

Who knows the real reason 😂

Everything Musk has said about LiDARs could just be backward justifications. Early 360 degree LiDARs cost ~$75K. Clearly that cost is not viable for consumers. Since then the cost has plummeted to about $10K... better... but still a significant cost. If you adopt an aggressive market strategy, you can't wait for the price to drop. To me this better explains why Tesla does not use LiDAR than any nonsense Musk has said. I am sure the price will keep dropping. I'd say it is at acceptable levels for narrow fields of view (e.g. Ibeo).

Can you build an autonomous car using only cameras? Probably... Should you?? Probably not... Not even Tesla do - they also use radar. Redundancy is a useful feature in systems with spectacular failure modes. Do you want your autonomous car to have human vision or superhuman vision? Seeing in the dark; seeing while driving into the sun and seeing through fog/snow/rain all sound like pretty useful capabilities to me. Our current technology does not provide those capabilities in one sensor, so a mix is required. LiDAR can play a role in that.

One of Tesla's first fatalities could have been prevented with LiDAR. The driver was watching Harry Potter whilst on autopilot. Apparently the computer vision could not differentiate between a white truck and the open sky (over exposed?). The car drove underneath the side of the truck at 100km/h - decapitating the occupant. A forward facing LiDAR system would have likely detected an obstacle in front of the vehicle and intervened.


I didn't know Google is that far ahead "officially", I'm a bit hesitant to call them that far ahead due to exactly what you said about Tesla's access to training data .. and the part that Tesla's are already out on the road, and I haven't seen google one in practice yet.

Clearly you don't live in San Fran or Phoenix ;) (ignoring your avatar location information 😁)

I think this goes back to a previous point though. I think other companies are being more cautious and even working with municipalities to roll out trials.

It is hard to say whether the Tesla data set can tip the balance. Good data is worthwhile... inconsequential data is an encumbrance and shyte data is a liability. I am sure Tesla has hired some top guns... but autonomous driving is a software and AI task... Google has a lot of form in that space. They also know how to handle large volumes of data and have the infrastructure for it. This means they can take some computational shortcuts in dense urban areas (like SF or PHO).
 
Last edited:
I am sure Tesla has hired some top guns...

Tesla hired Karpathy but one cannot build ML with reputation alone.

In fact the quality of your individual MLEs and DSs is almost never the determinant of how good your product is. Tesla and Uber manage to hire people with impressive resumes but see my prior comment. Anyway they aren't any better than NVidia or Google's teams and the fact is the latter two seem to make significantly more material progress every year, probably down to the same **** culture that has plagued Tesla. Uber has the same issue, fwiw, and despite hiring tons of qualified people and trying to hire even more (for example, by pestering certain grumpy KKF members) have had to dump their self driving division to stop hemorrhaging money at quite such an extreme rate.

Anyway I can only say so many times that Tesla's got issues and have people not listen before I throw a clot in my brain.
 
i just sms:ed my boss telling him i needed a 1k €$£ raise a month.

and that i would also need a tesla from now on. but none of these fagg0t ones with only 400hp, but the real ones with 700hp.

its only 4 in the morning here. we'll se how it goes. i think i have a good bargaining standpoint. he's gonna love this.

From a bicycle to 700hp🤪
 
Here plenty houses have solar panels electric car works well with that.

I'm in 5g stocks it's a game changer with huge rise in data capabilities. Don't buy that self driving cars will be universal in a decade as some predict. If battery technology gets better
electric will take off. China has lots of electric
car companies. Some will become world producers in the future because of cheap labor.
Even Japan & US car companies building more
electric cars. I wouldn't buy Tesla stock at these levels, plenty other players in the works.
 
Google knows things.
I suddenly get electric car compare videos on the tubes.
(not an endorsement of this channel, first time seeing the dude, but seems like just a dude talking about cars reasonably honestly and mostly about the feels, not the pure on-paper stuff)

His opinions in short:
- Tesla - boy racer gadget that does the one thing amazing well, go fast in a straight line.
- Lucid - refined luxury sedan for 50+ year olds, made by nice people.
- EQS - magical, serene, the pinnacle of German luxury, reasonably priced given what you get
- Taycan - lots of porsche tax but you get an amazing balanced sporty porsche feel.
Then a bunch of discussions about how Tesla and expectedly Lucid are often under advertized range, Tycan is way over and EQS is a bit over advertised .. so in practice, they are all not that far apart from each other.

How close do you think is he?

 
Back
Top