Search results

Intelligence is not a binary phenomenon, and the AI researchers are showing that they do have some approximate capabilities to reason and generalize. These capabilities are by no means perfect or universal. https://arxiv.org/abs/2310.13571 https://arxiv.org/pdf/2306.09841.pdf...

As it turns out the safety score is based on forward collision warnings per 1000 non-autopilot miles. In practice it means if you get a FCW you need to dilute it by driving off-AP/off-FSD in circumstances unlikely to make a FCW, which means easy highway driving where AP/FSD works well. So in...

That's all true, but even more important I think is the testing by the provider of the routing and mapping within the geofence---there sure seem to be many mapping errors or uncertainties (given Tesla experience) that likely need to be manually annotated after testing. At least all major...

It can be safer than a human in some cases (awareness of moving objects all around is better than humans today, so AP/FSDb on divided highway is good), and yet totally fail as a hypothetical L4 robotaxi. The safety we are seeing today is because humans override it and turn it off where the...

This also means that the insurers will be the ones to decide the actual risk with their dollars and won't be fooled by bullshit. Like climate/hurricane risk in Florida, they will look at objective science as much as possible for pricing. A non-affiliated insurance company agreeing to take on...

Did the car come from the US plant or the China plant? If China, no car there so far has used a NCA battery. It appears only Panasonic makes those. If US, in those years, it was all Panasonic NCA.

Can you repost the links without the Twitter? I cna't log into it at work.

Right, and given the hours that Tesla demands, they have to live physically in the most preposterously expensive part of the world outside Monaco (Palo Alto) and without the benefits of such a location. It may be intentional---they will only get very young people, and youth often has bad...

In what way do you prefer the drivetrain? To me that's a strength of Telsas.

Chill mode improves efficiency for me, and I'm sure 99% of it is changing human behavior.

what was your typical efficiency on original tires (MXM4?)

Back on topic: 11.4.7.3 It had 3 hard fails for me on a drive last night. One had it almost running a red left-turn light---it didn't think far forward enough to see it already turning yellow, it thought it could gun it and make it. Another had a fail on a well-trafficked intersection it...

With current management, pressure on margins and sales will likely lead to even more decontenting. That's the point of complaining about it before it happens.

The systems are better at detecting fast-moving objects, particularly out of front view, than many humans, particularly at night. I like using it for freeway driving at night for this reason. The routing and decisioning and planning is obviously worse, and it appears to be too much like the...

link works for me still. Maybe a temporary glitch.

Because the manufacturing and mechanical engineering talent is really good, and there aren't wealthier competitors with tons of VC capital or gobs of cash flow trying to hire them away. And dealerships of course. And because most of the time Elon's instincts in mechanical engineering are better...

Their interest is aviation of course, not automotive, and will accept a much higher price and lower volume.

Potentially but I think better cameras (and enough compute to take in the higher bitstream) would allow the statistical detectors to work better. They will also tune to favor low false negatives for any Vulnerable Road Users (pedestrians & bicycles) which will have some false positives...

and this is a case where a radar would help, as it did once before. Where it's probable there is no obstruction but a negative radar return makes it much more certain and safe to continue at full speed.

Some of that could be because of inadequate resolution of the cameras in HW3. 1280x960 is not enough at all for full view. Human foveas are better and humans have fast gimballing and scene assembly from quick views. And some is because artificial convnets do not see the way humans do, they...

EPA range is highway & city combined. Highway only at 70 is less. MT tests to 5% presumably indicated on display. EPA tests to physical battery empty, which is a few percent below 0 on display. (When new cars have a 4.5% buffer below 0). Unfortunately EPA website doesn't make highway range...

Its more likely that bug detritus on the windshield in front of the cameras induces a general haze and less clarity, and that might occasionally trigger PB from perception mistakes. It's also possible it's irrelevant---more undulating up and down roads, common in country areas and not...

The question is 'end-to-end *training*'. That is the real distinction: error signals backpropagate all from driving examples even through perception and intermediate states. not end to end forward inference.

that is exactly what WLTP is. It is calibrated upwards vs real world, but is more reliably precise.

Yes, that's the existing Karpathy solution on the perception side. It is trained against a large database of labelled video with human derived interpretation of 'objects', and then a rules & optimizer based policy planner is run with those features to drive the car. That gives what we see...

If it's really end to end training where human driving is the primary training signal along with generative video prediction (that is what is being demoed) then no they are not the same ones at all. The key element in any neural network training is the supervision signal. If that changes, then...

The Tesla energy screen with navigation destination accounts for everything and adjusts for everything and is very accurate and continuously updates. The top of the display meter should be permanently set to percentage, and some item in service tab gives current estimate of max full pack...

That's science fiction. The translation from linguistic generalities to policy in a complex perception world decades away. With end to end training you don't even have a defined internal representation for 'stop sign' assigned by humans, it will somehow generate something that you can post-hoc...

Actually that's really really difficult. With end to end training there isn't this ability to easily identify objects and make rules. And there is not a computational budget to run two entire full featured ML models simultaneously. They already do heroic tuning to reduce latency and time...

It is not, at least the models out of Shanghai are nearly the same size.

FWD for a new GM product at $49K vs a much quicker AWD Tesla Y at $49K? Which will be imminently in a refresh which with the 3 experience will improve its ride and comfort. Not going to be attractive at all. The new Bolt, with its non-Ultium CATL LFP pack might find a good niche at $37K...

No, EPA efficiency does matter. That is the EPA rule: test the default/common mode. I think it also may be a reason why the Porsche Taycan EPA is so bad compared to what they actually experience given the size of the battery and great aerodynamics: the Taycan default mode is weighted towards...

The new-technology V11 was supposed to be that version that they focused on for deployment and abandoned V10. I remember this same discussion happening previously. And non-FSD autopilot (the one most people are using) was supposed to be merged a while ago and that isn't happening. V11 FSDb is...

Karpathy leaves and a few months later it goes to *sugar*. Karpathy is a great scientist. Ashok Elluswamy seems to me more of a Elon-style booster. And Elon's public decompensating and mandating in-office 5 days a week, and declining stock price, are not at all helpful to attract and retain...

I wrote to the Discount Tire rep that occasionally posts here and pointed out that it was missing from TireRack a week ago. :)

There is a 4.5% buffer below 0% SOC. Which might be significantly more than 10 miles. EPA runs it to true empty. It might be that Tesla is more conservative in the in-car display of battery %age than others, and then uses EPA test aggressively without any deductions.

Just a quick comparison, a month ago, my experience in Edinburgh & Belfast were notably less expensive for dining & beer than San Diego, once you include the tax & tip additions. example, a US pint is $8.50 + 8% tax + 18% tip, $11 UK pub, a larger pint is 5.50 GBP, 6.67 USD. Ireland was...

The BMS calibration is irregular so while the battery degrades smoothly the estimate does not, and jumps around. And I think there is a period at early ownership where the estimate is nearly fixed to the rated miles, and then it releases to its running estimate. I saw that on mine. You...

There was a German test where they equalled or exceeded the Michelin Pilot Sport 5 perf, and greatly exceeded in rolling resistance. This is a fantastic result. https://www.tyrenews.co.uk/posts/hankook-ion-evo-crowned-winner-of-auto-bilds-reifencheck-tyre-test-for-evs...

He was better and more effective until about 5-7 years ago. He has declined in ability, and increased in arrogance. Henry VIII was a decent monarch in the beginning of his reign as well, but turned into one of the worst. Particularly after a brain injury---which I think Elon has from drugs...

That may be yet another Hankook EV tire. Or a renaming of the IK01 summer tire, to distinguish a summer performance line vs the all season better. Or an update of the IK01 summer tire. The summer Hankook iON appears to be sparsely available in USA now, look for code IK01 (summer) vs IH01 (all...

High mileage rating (durability) tire has two things: (1) less sticky compound---less grip, more durability, (2) higher tread blocks. (2) means more rolling resistance as higher tread blocks deform more when the tire is new. The weight isn't the central issue. Generally why performance tires...

Sure. Because finding rain drops on the window through the camera which is focused at infinity means trying to distinguish slightly hazy/fuzzy images from clearer ones. That's what rain would do to the image. That effect is simulated by other things, like glare and dirt films. It is why there...

they seem to be unavailable at TireRack, but are available at Discount Tire. Which bought Tire Rack company recently but clearly they are still operating independently, and will for a long time probably.

I'm an empirical scientist and let the data speak. Musk did not used to be like this previously. 20 to 10 years ago, when Tesla and SpaceX were starting, he added great value in finding and motivating great employees, and convincing tech bros to invest. The vision and drive and willing to...

I think with LFP the state of charge calibration and the estimation of the capacity is more erroneous because of the long flat mesa in the charge vs voltage curve. So the conventional algorithm was giving more noisy results in the field and causing customer complaints & spurious service...

Prediction: 2 years in (do we have any 2 year + LFP cars?) the displayed degradation rate slows. One day after warranty expires the algorithm changes. Nevertheless, the battery chemistry is as the science says and keeping state of charge low will lower degradation no matter what tricks are...

He did not. He funded early development and expansion and brought in many excellent engineers. Because the Board represents all stockholders. Elon has 15% or less now, and has been selling. His attention and skill are waning. This is the job of the BOD. This is no small grocery bodega, but...

What do you mean? There's an Apple credit card! But seriously the Apple Watch has been released. And shockingly the revenue of Air Pods alone is bigger than many well known companies. I agree that innovation is stagnant though execution is excellent. Tesla is going to face more serious...

Follow the science. Your battery very probably is better off than the other people's, you just don't see it yet. I thought the same was true for me, with 22LR with NCA. Original range 358. I was seeing extrapolated 340 miles (170 @ 50% displayed) after 15 months and was a little disappointed...