Fixed itDamn it. Don't point that out in text form for media to pick up on then claim another broken promise if it doesn't happen.
You can install our site as a web app on your iOS device by utilizing the Add to Home Screen feature in Safari. Please see this thread for more details on this.
Note: This feature may not be available in some browsers.
Fixed itDamn it. Don't point that out in text form for media to pick up on then claim another broken promise if it doesn't happen.
Almost certainly not.I wonder if Lex buy the dip before the podcast went public, does that count as insider trading?
Or, when someone who understands ML/AI and watched Tesla AI event then went all in, does that count as insider trading?
Almost certainly not.
Particularly not since Musk has said Level 4 will happen "Next Year" or something similar multiple times in the past.
Elon has made more wrong predictions about FSD than anyone else.Particularly not since Musk has said Level 4 will happen "Next Year" or something similar multiple times in the past.
I have been struggling with the concept of "insider trading" for a while, and my conclusion is that all of us participants in this forum are guilty.Almost certainly not.
Particularly not since Musk has said Level 4 will happen "Next Year" or something similar multiple times in the past.
Africa is already a dumping ground for used ICE, that mechanism is used to help support resale values in the developed world.Fun to speculate. My wife has a film camera that we haven't used in nearly 20 years. When we got our first digital camera, we stopped buying film for the old one and forgot about it. Now we have great cameras on our smartphones, so we have forgotten about the digital cameras too.
Maybe this transition is too simple as a model for the EV transition. After all, the cameras we have forgotten over the years are small and still sit somewhere in our closets. They were never such a large investment that we needed to sell them or make special arrangements to haul them away. But one thing seems clear, from the perspective of early adopters, you can easily trade in, sell or give away your ICE vehicle when you replace it with an EV. You can easily forget about you old ICE. It's someone else's problem.
Internationally, we know that affluent countries tend to buy new cars and a large fraction of used cars get exported to less affluent countries. So affluent countries that transition quickly to EVs could simply cast off most of the aging ICE fleet. Forgotten.
But this raises questions out the countries that import used ICE. Will they become flooded with cars that nobody wants. For example, parts of Africa have become dumping ground for cast off used clothing donated from affluent countries. The poor can pick off what they want from mountains of clothing. Perhaps we will see giant tracts of land in Africa covered with used vehicle, a defacto pick-and-pull for spare parts. A giant fleet of old ICE could cannibalize through to the next century. This could have adverse environmental consequences for poorer countries with cast off fleets. Proper recycling of used vehicles might not be available at scale in these places. When ICE values decline so much that affluent countries are "donating" ICE to poor countries, watch out. Donating easily crosses the line in to dumping. Dumping ICE into poor countries that lack automotive recycling infrastructure is negligence. It would be far better to recycle cars in affluent countries and donate useful reclaimed metals to developing countries if one has genuine charity in mind.
Of course, in affluent countries there very well may persist a nostalgia market for ICE. But my hunch is that the nostalgia market will fixate around certain iconic models and types of vehicles. Watch what ICE models remain in production the longest. There's a lot of forgettable ICE that will go out of production by the end of the decade. My own modeling suggests that from 2024 to 2030 non-BEV falls from 80% of new car market to 20%. This is a rapid collapse of demand for the most forgettable models of ICE. What's left in production after 2030 is largely niche, nostalgia or going-out-of-business models. Only the nostalgia models will have any sort of durable value. However, nostalgic petrol heads will have a huge supply of used vehicles to pick over for many decades. Nostalgia buyers may even find that they prefer pre-EV vintage model years to post-EV model years. As companies like Hyundai shut down ICE development units the quality of new ICE could decline. ICE makers just try to make the post-EV ICE models as cheap as possible with little inspiration or innovation. So the nostalgic car buyer could come to appreciate the older model years back when car makers still had some passion for antiquated technology.
Another twist here is that if you can use your smartphone to call a self-driving EV at lower cost than owning an EV or ICE, then quite a lot of us could forget even about owning a car. This would be like the digital camera we never use because smartphone.
Tony Seba predicts the clean energy and transport transition is largely complete by 2030. My hunch is that Tony is fairly right and EVs will be the fastest part of the transition.
TL;DR - Tesla's demand is the first sign, then the ever increasing higher profit margin, then scale (which is happening in 2022 to 2024), then marketing/FUD machines start losing traction, then legacy products filling dealerships (no demand), then legacy factories closing/retooling, then layoffs, then legacy brand mergers, then maybe bankruptcy.How did vinyl records, film camera, VHS/Beta/cassette/CD, tube TV, Atari, push button phones etc… businesses go. Like that, just louder.
Doesnt take much for Level 4. Remember it can be in specific areas in specific situations. Honda and Mercedes are doing Level 4 in bumper to bumper traffic under 40 mph on specific highways. Tesla could do it if they want to play this marketing game.Almost certainly not.
Particularly not since Musk has said Level 4 will happen "Next Year" or something similar multiple times in the past.
The disruption is when the mindset changes, when only a minority of buyers will buy a new ICE.Sebas numbers haven't panned out for years running now. Right idea, but way ahead of actual production capacity.
There's simply not enough batteries in the pipeline to replace 100% of just 2030 ICE sales.
Let alone replace the 20 years of such sales it takes to replace all the existing ICE on the road.
This is one place where, unusually, Elons predictions are far more realistic than that of others.... ~30 million EVs a year by 2030 is doable with battery scaling as we see it happening now (with the majority being Tesla).
But there's approaching 2 billion ICE vehicles on the road to replace.
Even if we pretended robotaxis exist we're not replacing even 1/4 of those by 2030.
Doesnt take much for Level 4. Remember it can be in specific areas in specific situations. Honda and Mercedes are doing Level 4 in bumper to bumper traffic under 40 mph on specific highways. Tesla could do it if they want to play this marketing game.
Mercedes is doing Level 3 I have undestood?
.
Mercedes Is First To Sell A Level 3 Autonomous Vehicle In 2022
The first vehicle you’ll be able to try this real, no-hands self-driving feature is the 2022 S-Class, but only some German autobahns for now.insideevs.com
Lots of good info on FSD v11 for those who don't have the time to listen.Elon Musk: SpaceX, Mars, Tesla Autopilot, Self-Driving, Robotics, and AI | Lex Fridman Podcast
I generally agree. NN performance generally scales with size. But I do not know and neither do you.find it unlikely that the final code for FSD will have smaller neural networks (as would be necessary to run it redundantly on HW3), as they had been planning to be able to run it on pre-3.0 versions. So it appears as if the NNs keeping getting bigger. That fits with Karpathy saying that one big feature of the NNs they use is that their performance scales with size.
Thanks for the summary.Lots of good info on FSD v11 for those who don't have the time to listen.
1. Was hoping to release it this year but is now delayed
2. Single stack
3. Re-write by replacing C++ with NN as the vector space compiler.
4. Reaching local maximum at how well C++ can generate vector space using NN's calculations (aka bag of points).
5. Taking away post processing of images which reduce 13ms of latency
6. Will require a complete retain of data due to #5
7. Surround video and train using surround video via auto labeling
8. It's in alpha
9. Add additional capabilities but didn't mention what they are
Overall design language
1. Simplify code
2. Decrease latency
3. Decrease or eliminate "jitter" which randomly general latency between 0-100ms. Said eliminate jitter will solve lots of issues with current software.
4. Working on a bare metal level compiler/software to optimize as much as possible for the limited amount of compute on FSD HW 3.0 which results in an increase of frames per second.
Agreed, but if you're short on time, you can skip through to Tesla relevant sections. They're well marked.Thanks for the summary.
However, I would think this has to be considered 'must-view/listen' video for anyone with even a passing interest in $TSLA.
Absolutely, however some people may be region locked or are at work and cannot watch the video.Thanks for the summary.
However, I would think this has to be considered 'must-view/listen' video for anyone with even a passing interest in $TSLA.
I've long been concerned by the lack of battery so I'd say your statement is spot on but bless Seba for calling the trend out before most others really had and for putting the story together in such a way that it was easy to understand. I really appreciate his work.Sebas numbers haven't panned out for years running now. Right idea, but way ahead of actual production capacity.
There's simply not enough batteries in the pipeline to replace 100% of just 2030 ICE sales.
Let alone replace the 20 years of such sales it takes to replace all the existing ICE on the road.
This is one place where, unusually, Elons predictions are far more realistic than that of others.... ~30 million EVs a year by 2030 is doable with battery scaling as we see it happening now (with the majority being Tesla). Current vehicle sales, annually, are more than double 30 million though.
And there's approaching 2 billion ICE vehicles on the road to replace.
Even if we could have 30 million EVs per year available today you couldn't replace even 1/4 of those by 2030. And we won't have remotely near that many EVs per year selling this year. Or next. The battery factories to do so aren't built yet.
Elon remarked at the shareholder meeting this year it'll be decades before EVs has entirely replaced ICE in the worldwide fleet.... and he's probably got a more realistic handle on how possible that is than Seba.