Telsa finally gets competition - Longbow Roadster

[neptronix]

Ex-Tesla Alums Debut New Electric Roadster Named To Taunt Elon Musk | Carscoops



Celeritas Levitatis means the Speed of Lightness and it’s the hallmark of Longbow. This new automaker out of the UK believes that it has the solution for the overweight automobiles that crowd our streets. “Weight invites complexity, blunts agility, and dulls the senses,” it says. Fixing those issues required going back to the drawing board. What it’s come up with are two new electric sports cars that weigh less than a Mazda MX-5.

 

[neptronix]

No idea if this company will deliver at all but happy to see competition in this space, personally been waiting for a little light electric car since the 2000's.. more competition = better prices for the consumer.

 

[Comrade]

more competition = better prices for the consumer.

But they are making a high priced electric sports car.

RAV4 is the best selling non truck vehicle. Similar format but slightly smaller compact SUVs is what is selling well all across the world. If they made compact electric SUVs, that would at least have buyers interested in the format, even if at elevated prices compared to ICE/hybrid, and would add competition.

Electric sports cars are a niche in a niche.

 

[neptronix]

It's about half the price of a Tesla Roadster and it will likely encourage other manufacturers to build something similar if successful. That's the good of it to me.

The primary roadblock to making something like an electric Acura Integra or Corvette has been energy density and with today's cells. It's only now possible to build such a vehicle without greatly sacrificing range and requiring something like a carbon fiber chassis to reign in the weight.

Yes, any company could make a lot of money selling boring obese SUVs, but there are still a subset of car drivers like me who enjoy excellent driving characteristics, and for the time being, Teslas are the only cars that are relatively affordable that can scratch that itch.

For me, Tesla is out as a brand of car to consider for these reasons:
1) The company is vehemently anti-repair; as bad as Apple computer.
2) The company has bad customer service/support
3) Elon's involvement in politics has a large % of people boycotting the brand; this casts a dark shadow over the company's future. Not many companies can survive losing half their buyers.

..so i am happy to see that Tesla has competition.

 

[Hillhater]

Elon's involvement in politics has a large % of people boycotting the brand;

I suspect this is just a media invented factoid !
Tesla’s sales situation is mainly the result of increased competition with more value choice for EV buyers compared to previous years . Pointing at Musk is just a easy justification for buying from competitors.
At least Musks political activities are open, obvious , and presumable ok with the majority of US voters, as compared to the dubious activities of heavily politically influenced Chinese EV producers !

 

[neptronix]

It's not, there's a ton of people who have been selling theirs out of dislike for Elon for about 5 months now.

There is a big drop registered in German Teslas:
https://www.news.com.au/technology/...a/news-story/87488c5db4e04ff9e120a8ef2ff8530f

At least half of Americans don't like Elon's involvement in our politics. But they don't dislike it as much as the Germans, for interfering with their politics as well.
Tesla is down ~23% in marketshare around 2024, and they don't have Chinese competitors to blame here.
Tesla Sales Figures – U.S Market

Getting in bed with the government even more did good things for Tesla's stock until the backlash started



I don't think it's just a media thing hillhater, polarizing his former customers wasn't a good business move and has negative effects on the business that can be measured.

 

[Comrade]

there's a ton of people who have been selling theirs out of dislike for Elon for about 5 months now.

It just looks like a lot because they are all on Twitter. This is just a fad like any other. The emotionally oriented neurotics that make up the cohort of Tesla sellers don't remember anything that's not on MSNBC that day, and Musk isn't going to be on TV forever.

 

[neptronix]

They're also on reddit and everywhere else.
All the liberal people i know IRL who have the money and interest to buy a Tesla, won't touch the brand at all either.

I don't know if the boycott will last or not. It would probably take a few successive awesome product launches for some people to rethink it, lol.

 

[Epithemeus]

Elon lost me when he said you shouldn't use lidar for self driving features and just cameras are enough. Ridiculous. Also the lack of physical controls for a lot of things. And what Nep said.

I appreciate what Tesla did for the industry. Pushed some boundaries and convinced a lot of people of the viability of electric. It's time for their spotlight to end though. So this is great news.

 

[Comrade]

Elon lost me when he said you shouldn't use lidar for self driving features and just cameras are enough. Ridiculous.

I've been driving for decades using only 2 eyes and never had any kind of an accident. Many other people also have excellent driving records using eyes only. If I can do it, why can't a Tesla that might have 10 eyes, is never distracted, and has way better sensitivity in low light conditions?

 

[neptronix]

Elon lost me when he said you shouldn't use lidar for self driving features and just cameras are enough. Ridiculous. Also the lack of physical controls for a lot of things. And what Nep said.

Oh yes i forgot, lack of physical controls and overcomplicatedness is also another reason i'd never buy one.

I appreciate what Tesla did for the industry. Pushed some boundaries and convinced a lot of people of the viability of electric. It's time for their spotlight to end though. So this is great news.

I agree with that so much. What sucks is that they never proceeded to lower the cost so that the everyman could afford one. They jus loaded the Model 3 up with doodads and made it a semi luxury car. Lame.

I've been driving for decades using only 2 eyes and never had any kind of an accident. Many other people also have excellent driving records using eyes only. If I can do it, why can't a Tesla that might have 10 eyes, is never distracted, and has way better sensitivity in low light conditions?

Because AI still sucks even if we throw another 10x compute power at it. Lidar helps solve the problem that computer vision at speed still sucks, at the expense of having some expensive, ugly equipment attached to the car.

 

[classicalgas]

I've been driving for decades using only 2 eyes and never had any kind of an accident. Many other people also have excellent driving records using eyes only. If I can do it, why can't a Tesla that might have 10 eyes, is never distracted, and has way better sensitivity in low light conditions?

...because computers are stupid compared to humans when it comes to sorting and identifying objects in their visual fields.

 

[JackFlorey]

If I can do it, why can't a Tesla that might have 10 eyes, is never distracted, and has way better sensitivity in low light conditions?

Because

1) computers don't have the spatial awareness that people have
2) for the near field, Teslas don't have parallax sensing like we do (2 eyes)
3) Teslas don't have ears like we do, and cannot detect things like the noise of a car coming around a corner
4) on average, people are bad drivers with their current sensory equipment

It was a mistake to remove the additional sensors. The relatively new field of sensor fusion gives you more information from the combination of two sensors than you get by just adding the two inputs. Even a single decent forward facing radar would avoid a lot of problems.

 

[Epithemeus]

I've been driving for decades using only 2 eyes and never had any kind of an accident. Many other people also have excellent driving records using eyes only. If I can do it, why can't a Tesla that might have 10 eyes, is never distracted, and has way better sensitivity in low light conditions?

Because AI has the chance to not only match us but to be so much better than us. Cameras and lidar in combination. The more senses, the more information to make the right decision.10 eyes, no distractions, no difference at all between light and dark, combined with a full 3d map of the immediate surroundings unhindered by optical barriers.
It's very entertainment oriented but this video makes some excellent arguments.

 

[neptronix]

Well here's the real problem.
All current AI has to be trained on solutions ahead of time because it is unable to think sufficiently fast or deep in real time.
This is a substitute for lack of reasoning capability.

You can have all the sensors in the world and we still regularly see Waymos doing weird shit like this because they're in some situation they've not been trained on before and go bezerk in response.



Early AI models that CAN do reasoning require bucketloads of extra computational power. They can reason their way out of things not in the training set a little better, but are still imperfect, and horrendously slow.

Computer chips are now getting ~10% faster per year, down from the repeated 2x speed increases we saw under 'moore's law' in the 1970's-2000's, so we don't have a bevy of extra computational power per watt, nor electricity to fuel it.. with today's technology, to add reasoning to these cars would require putting a baby datacenter in them, but even still, it wouldn't operate as well as a human in these weird edge cases, so the benefit isn't really there.

 

[classicalgas]

"All current AI has to be trained on solutions ahead of time because it is unable to think sufficiently fast or deep in real time"

Current research on humans indicates we do that too ( work from predictions based on experience) since the world moves too fast for us to just respond to complex situations.

One of the advantages we have over AI is being able to quickly adjust when current reality doesn't match past experience.

Well, some of us, anyway...

 

[Comrade]

Because

1) computers don't have the spatial awareness that people have
2) for the near field, Teslas don't have parallax sensing like we do (2 eyes)
3) Teslas don't have ears like we do, and cannot detect things like the noise of a car coming around a corner
4) on average, people are bad drivers with their current sensory equipment

It was a mistake to remove the additional sensors. The relatively new field of sensor fusion gives you more information from the combination of two sensors than you get by just adding the two inputs. Even a single decent forward facing radar would avoid a lot of problems.

1) they do
2) maybe not now, but it doesn't mean it can't be implemented
3) maybe not now, but it doesn't mean it can't be implemented
4) people don't get into accidents because they lack sensors, it's because they do stupid things (driving in heavy snow or rain where they can't see falls under stupid)

The complaints with vision only navigation are valid today, given the state of technology today. Yes, computers are not as good humans vision. But only today. There is no doubt it will get better in the future and exceed the best of humans drivers.

 

[Comrade]

with today's technology, to add reasoning to these cars would require putting a baby datacenter in them

As was exemplified with DeepSeek, computing power in AI is not a predictor. It's all about efficiency of the code.

For example Mobileye chips used by say Toyota for their Safety System suite of features, and many other manufacturers, uses less than 3 watts for the video processing unit. With those 3 watts it can detect lines on the road, read road signs, detect humans and vehicle shapes, vehicle headlight or taillight, etc.

Highly optimized code and highly optimized hardware is where Tesla is going to go. That technology is still in its infancy, and it's already pretty good.

 

[JackFlorey]

1) they do

No, they really don't. People have a few tens of millions of evolution designed to let them live in a complex and dynamic environment. They can (unconsciously) track thing and then keep track of them even when they go out of sight (sensor) range. You need miles of code to do that for computers.

2) and 3) - sure, perhaps someday that will happen. Which goes back to the original comment - it's dumb to remove those extra "senses."

4) People get into accidents for a great many reasons - largely stupidity, but often a contributing factor is a lack of input. Example here would be kid with his music playing at 90dB who doesn't hear the truck horn.

There is no doubt it will get better in the future and exceed the best of humans drivers.

Yep. And one of the reasons it will get better is that sensor fusion - ability to use more than one sensing modality. Of the really big failures of the Tesla autodrive system, most would have been avoided or ameliorated either with radar or lidar.

 

[Comrade]

And one of the reasons it will get better is that sensor fusion - ability to use more than one sensing modality.

It seems like you are trying say that vision only self-driving vehicles fundamentally can't be safe enough for widespread adoption because they don't have LIDAR? Is that right?

 

[neptronix]

As was exemplified with DeepSeek, computing power in AI is not a predictor. It's all about efficiency of the code.

A good part of their optimization magic is simply writing assembler code for a specific generation of GPU.
The problem is that GPUs don't have an agreed on architecture.. they evolve too fast, so never established that. This makes for painful hardware upgrades because the underlying architecture changes often enough.

One big problem is that a lot of LLMs run on python and uses python-C/C++ interfacing to spot optimize.
Python is one of the slower high level languages and i'm thinking that the 'glue' code may provide a substantial power hit..
But i don't know enough about this area.

For example Mobileye chips used by say Toyota for their Safety System suite of features, and many other manufacturers, uses less than 3 watts for the video processing unit. With those 3 watts it can detect lines on the road, read road signs, detect humans and vehicle shapes, vehicle headlight or taillight, etc.

Highly optimized code and highly optimized hardware is where Tesla is going to go. That technology is still in its infancy, and it's already pretty good.

That's quite impressive, i didn't know that.
This hardware and software is probably custom cut for a limited scope of operations.. unlike Tesla's current system that is trying to handle 100% of human-road interactions.. which is tough to do.

 

[JackFlorey]

It seems like you are trying say that vision only self-driving vehicles fundamentally can't be safe enough for widespread adoption because they don't have LIDAR? Is that right?

I wouldn't say that. If the criterion is "better than an average driver" then you can probably get there with vision alone. But a car with ranging sensors/systems will always be safER than an equally competent car without them.

 

[classicalgas]

For example Mobileye chips used by say Toyota for their Safety System suite of features, and many other manufacturers, uses less than 3 watts for the video processing unit. With those 3 watts it can detect lines on the road, read road signs, detect humans and vehicle shapes, vehicle headlight or taillight, etc.

...but can that system identify a child that's fallen on a basketball into traffic, with a dog in close pursuit? Next to a turn signal on the car the kid just fell past, in the rain?

That's a visually complex occurrence, and the particular visual signature of that event isn't likely to be in the AI's training.

Brake, swerve, or both? Braking effort will reduce swerve traction, and vice versa. Does the car know that there's a place to swerve into without checking and processing first? Does the car even have the swerve option?

 

[Comrade]

If the criterion is "better than an average driver" then you can probably get there with vision alone.

But what if the criterion is "better than the best human driver"? Due to simple factors like no distractions, more and better placed visual sensors, better optical qualities than the human eye, etc.

 

[Comrade]

...but can that system

It has no "AI" in those chips whatsoever. But it does fairly advanced image recognition using very little power because it's specialized hardware and specialized software.

The point is simply that it's an apples to orange comparison, to compare general purpose GPU power requirements for self-driving "AI" image processing today, to what it would look like in the future when it is mature and will be produced by tens of millions of units per year on specialized hardware.