Dream drive pro and autonomy subscription

[LSteve]

Lucid Air runs on a nvidia Orin SoC. This chip is rated for 256 INT8 TOPS (trillion operations per second). TOPS is a measure of how powerful the chip is.

The Gravity uses 2x Orin SoCs or about 500 TOPS.

Midsize uses 2x nvidia Thor SoCs. These are the latest generation of chips from nvidia and can do ~1000 TOPS each so about 2k total. The self-driving Gravitys for Nuro/Uber also use this setup.

Funny thing is, Tesla's current gen HW4/AI4 is estimated at 150-175 TOPS but their software is completely custom. It's also probably not capable of full unsupervised driving but Cybercabs are launching on AI4 so who knows. Maybe that's why the rollout/testing has been so slow. Their next gen AI5 hardware is rumored to have 2000-2500 TOPS.

Lucid uses Nvidia's alphamayo as the base for their ADAS systems. Unlike Tesla, Nvidia has no real reason to make their model efficient. Maybe the Air could do it if Lucid spent a lot of money on it but they don't have the money to spare. Gravity sounds like it will be able to do supervised point-to-point eventually (and they had hands-free city driving as coming in 2027 in the investor presentation) but who knows when.

Another comparison point is Rivian Gen2 (R1T, R1S, and launch edition R2s). They use the same setup as Gravity (2x Orin). Unlike Lucid, they are going the Tesla route and building their own custom ADAS software. They have stated that supervised point-to-point ADAS is coming to Gen2 vehicles. Unsupervised P2P is supposed to come to Gen3 models. This generation will use custom designed chips capable of 1600 TOPS. Gen1 Rivians are essentially done. They use some old Mobileye hardware that is nowhere near capable enough.

The guts of the Air are outdated, and the Gravity will soon be outdated. This isn't even taking into account the number of ECUs used in these models. Midsize is going to run on 3 chips. Gravity has 13 and Air has 30ish?. Lucid can't just slap the latest gen chips into these two models. It would take complete redesigns to bring them up to date. This is probably why Tesla discontinued the S and X. They did not think spending the money redesigning the internals for AI5 would be worth it with how few they sell every year.

If I owned an Air, I would not hold my breath for new ADAS capabilities beyond a self-parking feature that is slated for 2027.

So it sounds like *if* they wanted to, they could probably bring something close to Tesla's autopilot to the Premium version of DD Pro in the Air? Tesla pulls that off with HW3, which I have to imagine the Air hardware probably beats. But why would they even if they could? That would make DD Pro even less worth it, unless they could get highway point to point on the Air as well, like Tesla's older version of FSD, before they rewrote the stack for city streets. All of this probably isn't in the budget I'm guessing, which is a bit of a bummer.

 

[Buffalo Bob]

So it sounds like *if* they wanted to, they could probably bring something close to Tesla's autopilot to the Premium version of DD Pro in the Air? Tesla pulls that off with HW3, which I have to imagine the Air hardware probably beats. But why would they even if they could? That would make DD Pro even less worth it, unless they could get highway point to point on the Air as well, like Tesla's older version of FSD, before they rewrote the stack for city streets. All of this probably isn't in the budget I'm guessing, which is a bit of a bummer.

...but thinking about it, wouldn't proper accounting require at least some of the $9k I paid for DDPro to be sitting somewhere in Lucid's coffers until they make further progress toward completion of the feature set I paid for?

 

[martinjlm]

...but thinking about it, wouldn't proper accounting require at least some of the $9k I paid for DDPro to be sitting somewhere in Lucid's coffers until they make further progress toward completion of the feature set I paid for?

Based on my time in Product Portfolio Planning at a major OEM...nope. That's like numbering grains of rice.

 

[Buffalo Bob]

Based on my time in Product Portfolio Planning at a I'm hidden major OEM...nope. That's like numbering grains of rice.

I really didn't mean that some of my $9k specifically should be set aside, or even some of DDP charges in the aggregate. I just meant that if people pay extra for a specific upcoming fearure, a responsible company would not just abandon the effort and spend the money needed for completion on something else.

If they want to abandon the effort, at least a partial refund would be in order. Even with that, it's bad enough that I would never have bought the car in the first place if I thought they would just do that.

 

[Ktw13]

I use my Air DDPro on every work commute, this morning it was really foggy and it performed well.
To me its well worth the one time spending.

 

[martinjlm]

Based on my time in Product Portfolio Planning at a major OEM...nope. That's like numbering grains of rice.

I really didn't mean that some of my $9k specifically should be set aside, or even some of DDP charges in the aggregate. I just meant that if people pay extra for a specific upcoming fearure, a responsible company would not just abandon the effort and spend the money needed for completion on something else.

If they want to abandon the effort, at least a partial refund would be in order. Even with that, it's bad enough that I would never have bought the car in the first place if I thought they would just do that.

It might not seem like it in my initial response but that's not what I meant. The "grains of rice" would be attributing revenue from vehicles sold to groupings of content and using that as a basis to reinvest in that form of content. Department budgets are set based on the approved technology plan and are forward looking, not drawing from results of product in commerce.

 

[Buffalo Bob]

It might not seem like it in my initial response but that's not what I meant. The "grains of rice" would be attributing revenue from vehicles sold to groupings of content and using that as a basis to reinvest in that form of content. Department budgets are set based on the approved technology plan and are forward looking, not drawing from results of product in commerce.

Shouldn't the revenue from the yet-to-be-developed features never be booked in the first place? My overarching point is that rules must somehow be broken if the bottom line includes $$$ for something that is never delivered. Worse yet if it is no longer intended to be delivered.

 

[Aka-none]

I could tolerate the current level of hands-on / eyes-on Highway Assist if the hands-on part was based on galvanic skin response. With it being torque based, I cannot help but torque the wheel in the direction I want to go, so it ends up just about indistinguishable from driving the car myself in Adaptive Mode.

Hands-off attention monitoring is even more baffling. I can drive the car in hands-on / eyes-on mode for hours and only get a 'pay attention' warning if I do something stupid like fiddling with music selections too long. Just the way it should be! If I switch to hands-off / eyes-on mode, I can't get five minutes at peace. Why would the 'pay attention' bar be set lower for hands-on mode than it is for hands-off mode? To me, that makes no sense at all.

As far as wondering how Lucid is going to handle the massive deliverable shortfalls from hands-off / eyes-off.... I hate to say it, but it looks they are going to stick with a case of selective amnesia... Perhaps from staring too long at the Lidar/Radar/Sensor arrays while wondering how the hell Tesla beat them so badly with nothing more than a few crappy cameras.

So Tesla isn’t hands off eyes off as that doesn’t match the definition of supervised, dispite what people say or think. Tesla themselves call it supervised.

https://www.tesla.com/fsd/safety

Tesla Autopilot and FSD crashes hit record 207 in a single month

Tesla reported a record 207 Autopilot and FSD crashes to NHTSA in a single month — the most ever — even as it redacts nearly every crash narrative.

electrek.co

Tesla FSD Safety Statistics: What the Data Can—and Cannot—Show

Read Tesla's collision-report methodology alongside NHTSA reporting limits and independent partial-automation research.

www.informedforlife.com

Tesla FSD Safety Record — Crash Data & NHTSA Investigation Analysis

Deep dive into Tesla FSD safety: crash statistics, NHTSA probes, fatality rates, and how Full Self-Driving compares to standard Autopilot.

www.autopilotwatch.com

Here's 13 crashes tied to Tesla's expanded Full Self Drive probe

U.S. regulators expanded a probe into the effectiveness of Tesla's Full Self Drive software at navigating in reduced visibility conditions.

www.usatoday.com

So all those folks not supervising in a Tesla fad aren’t following musk oxes directions. And remember that’s the company that treats fed as a parlor trick, I mean really mad max mode? That’s a safety oriented design criteria?

 

[Northpointe]

Shouldn't the revenue from the yet-to-be-developed features never be booked in the first place? My overarching point is that rules must somehow be broken if the bottom line includes $$$ for something that is never delivered. Worse yet if it is no longer intended to be delivered.

The terms of sale would govern this. I don’t recall seeing anything suggesting that the purchase of DD Pro was a guarantee of a certain level of future vehicle autonomy. Most references refer to semi-autonomous driving which is a bit nebulous. It may have been over hyped by salespeople but that in and of itself is likely not enough to pursue a remedy. It wasn’t being offered on a deposit basis or with any assurance that a refund would be given if it fell short of full autonomy. As such a balance sheet reserve wouldn’t be warranted.

 

[martinjlm]

Shouldn't the revenue from the yet-to-be-developed features never be booked in the first place? My overarching point is that rules must somehow be broken if the bottom line includes $$$ for something that is never delivered. Worse yet if it is no longer intended to be delivered.

Wish I could answer that. Never had to deal with that when I was in Product Portfolio Planning. But I guess my best recollection would be that revenue was always tracked on a vehicle program level, not technologies or features. There was always a lot of forward looking projection at a technology level for the purposes of business plan development. Occasionally there would be a “did you meet your promised targets” review at a technology level, but never a full tracked accounting.

 

[DS9]

So it sounds like *if* they wanted to, they could probably bring something close to Tesla's autopilot to the Premium version of DD Pro in the Air? Tesla pulls that off with HW3, which I have to imagine the Air hardware probably beats. But why would they even if they could? That would make DD Pro even less worth it, unless they could get highway point to point on the Air as well, like Tesla's older version of FSD, before they rewrote the stack for city streets. All of this probably isn't in the budget I'm guessing, which is a bit of a bummer.

I enquired about hands on lane center in DD Premium during an owners event in July. Lucid followed up by email saying that DD Premium in the Air didn't have the required equipment for hands on lane centering.

 

[bcjenkins]

I enquired about hands on lane center in DD Premium during an owners event in July. Lucid followed up by email saying that DD Premium in the Air didn't have the required equipment for hands on lane centering.

That response is a head-scratcher given our 2020 LEAF has excellent hands-on active lane centering supported by a single, forward-facing monocular camera.

Does the DD Premium equipped Air *not* have any forward-facing cameras?