Computers get old and EV tech is constantly improving. ICE has been stable for decades. Tesla believes they can increase the market value of older EV's eventually by pushing autonomous driving.
I like the direction of https://slate.auto. Module, bring-your-own-computer. We'll see if they allow affordable trade-in's for upgrading battery/motors.
Is it really true that ICE has been stable? Cars seem to have been getting many innovations, especially with power, torque, and reliability. We probably don't hear much about it because it is low profile stuff and a mature product.
The newest vehicle reliability advances are _less_ reliability via:
- cylinder deactivation
- ubiquitous turbos
- gasoline direction injection
- more computers
- generally higher cost of repairs (eg: if a car from 2023 needed a headlight it would cost much more than a car from 1998 needing a headlight, and even if they both had the same failure rate the reliability of the new car would be worse from cost alone)
This is well informed and correct. Understand this if you buy a new ICE vehicle: the drive train is uneconomic to repair out of warranty: do not imagine you'll keep it long term or hand it off to a kid or whatever.
Costs drop like a rock once vehicles get old enough that they can't be financed, which means dealers won't sell them, which means they won't be bundled with 3rd party warranties and service plans, which means that the owner will be the one paying and there's actual pressure to control costs (because the owner has no party down the line to pass costs onto).
For example, for the longest time Nissan CVTs were "nonrebuildable, send it back and buy a reman" now any idiot can rebuild one for under a grand in parts.
4L60 and E4OD rebuilds were also $$$ for a long time now they're dirt cheap too.
I wouldn't cite the 4L60E in this argument. It's an ancient design, as Wikipedia puts it, "The 4L60E is the electronically commanded evolution of the Turbo-Hydramatic 700R4, originally produced in 1982." The 700R4 is a THM350 from 1969 with an additional overdrive gear.
I'm not writing about 50+ year old platforms, around which a huge market of suppliers and technicians has evolved. The ICE transmissions GM sells today are vastly more complex, with zero design commonality with the classic stuff, and enjoy none of the benefit of long adoption that make the classic stuff cost effective. Further, and this is the important part, because the lifecycle of everything ICE is much shorter now, measured in years as opposed to decades, they never will.
So ten years from now, when your circa 2020 10L60 dies, there won't be a transmission shop in every town that's equipped and stocked to deal with it cheaply. The cost will be greater than the value of the vehicle. And that's my point: these vehicles are not going to be economic to operate out of warranty.
You can trace all these designs back forever. It's more "inspired by" than actual incremental revision in most cases. There is just about nothing but some vague shapes that look similar and maybe some bolt lengths that are common between a 4L60 and the TH350 era stuff.
>Further, and this is the important part, because the lifecycle of everything ICE is much shorter now, measured in years as opposed to decades, they never will.
The average car on the road is lasting longer as the years go on. People said the same things when fuel injection came out.
>So ten years from now, when your circa 2020 10L60 dies, there won't be a transmission shop in every town that's equipped and stocked to deal with it cheaply. The cost will be greater than the value of the vehicle.
I'll take that bet. Modern transmissions are stupidly easy to rebuild from a skills point of view because they replace all sorts of mechanical adjustment with "hurr hurr we just PWM the solenoid to make it go BRRT and if the BRRT is too rough the computer algorithm will turn it down". Yeah there's more components, but those are easi.
> do not imagine you'll keep it long term or hand it off to a kid or whatever.
i don't agree, my friend has been driving the same Toyota LandCruiser for 20 years. I will have no problems handing my 2016 4runner down to my kid who turns 16 this December. The 4runner will last another 10 years easy.
Respectfully I specifically wrote "new." 20 year old SUVs predate the issues of new vehicles.
> my 2016 4runner
Even a 2016 vehicle predates what I'm pointing out. A 2016 4runner likely has a 270hp naturally aspirated V6 with modest power and a relatively basic 5 speed auto transmission. A 2025 4runner is a turbo charged 4 cylinder making nearly 2hp/cu in. and an 8 speed transmission. The former is much simpler and thus economic to maintain and repair compared to the latter.
ICE vehicle drivetrains have changed radically in only the last few years. They're almost universally using forced induction, integrated into unserviceable castings, to attain far higher volumetric efficiency, equivalent to high performance engines of not long ago. Gone are the 4-5 speed transmissions and transaxles: 8-10+ speed is the norm, and the complexity follows here as well. They are absolutely intolerant of neglect and abuse. Repairs are complex and likely to fail: manufacturers have taken to replacing major assemblies in leu of attempting repairs as they would have in the past. Part of that is due to the unserviceable nature of these components. Another part is the lack of dealer talent to deal with their own products. Another part is that the manufacturing lifecycle of major assemblies is now extremely short: only a couple years, whereas 10+ years was normal for common platforms as recently as the the 2010s.
What this means is: when these new products are no longer supported by manufacturers, who will drop their supply obligations rapidly as they legally can due to short lifecycles, parts will be fabulously expensive and difficult to obtain and the skills and tools necessary to repair them will be rarified and also expensive. Post-warranty ICE vehicles will be uneconomic, full stop.
Pretty much all of these reliability reducers are manufacturers trying to eek a little more MPGs by throwing lots of complicated technology at the problem, which introduces a lot more failure points.
Headlights and taillights on my current vehicle are supposedly around $1500 each, mostly due to a bunch of sophisticated sensors being built in.
Back in the 80s headlights were standardized (in the US at least) - you either had rectangular or circular. They were available at every auto parts store. Now they're a special order item from the dealer.
New synthetic oils are very durable. They actually do last a long time.
There are oil tests that confirm this. Also, 10,000+ mile oil changes are not new, and there are tons of used vehicles on the market, running around with long oil change intervals, and high mileage.
But you can't change a headlight bulb if they're a sealed unit, because those don't have bulbs, but with basic tools, you can certainly change a headlight.
But also, if you have a vehicle with a sealed headlight, you're probably not having to change it every other winter.
Actually ICE has improved significantly in the last 20 years, even though progress was held back by emission and safety regulations.
In the year 2000 a sports car would have around 200/250hp while these days any sports car that wants to boast power has around 500hp or more.
Perhaps you're not into cars much but if you compare top cars on track days etc. you will know there have definitely been huge changes.
Though during the last 20 years repairability and reliability also took a hit.
The must fun I've ever had while driving was with the family '99 Honda Civic. Zero power. Zero torque. But, man, it was like driving a go kart in the streets. I'd zip through mountain passes with the engine revving to hell but in reality I'd only be barely going the speed limit.
I did the same mountain pass in my Subaru recently and didn't even notice I was going 20mph over the speed limit. The engine was silent and still responsive.
Those are improvements overall for the environment and safety of course. However as usually improvement in one area is not necessary an improvement in another area. I was referring to "sport" performance of a car or ICE engine then the emission regulation held back the power output and performance of the drive train and the safety regulations added a lot of weight and size to the cars therefore hurting handling.
Said no pedestrian hidden behind my A-pillar (just kidding I still drive a 1980s minivan, one of the poor people ones, not the hipster one, don't get your hopes up).
I'm not sure they've improved that much. Comparing my 1990 miata/mx5 with modern cars I prefer not having electronic screens and bleeping things all over the place. The only thing I'd chose to modernise is engine efficiency which is maybe a bit better now.
A couple of suggestions:
1. Instead of multiple tiny inputs, you could have a single form with a date picker which defaults to todays date. Clicking a column header could set the date input value.
2. An end user wont know about the formatting requirements for entering task duration. A more user friendly option would be to parse the duration from the task name. Using some regex to match, heres an example.
// Look for time patterns like "30min", "1h", "2 hours", etc.
const timeRegex = /(\d+)\s*(min|mins|minutes|h|hr|hrs|hours)/;
const match = text.match(timeRegex);
Apple's chips are based on the ARM architecture, which is inherently more power-efficient than the x86 architecture used by AMD and Intel for their laptops. Apple also designs these chips in-house.
Apple is using a combination of high-performance and high-efficiency cores (Big.LITTLE architecture), allowing the chip to balance performance and power usage based on the task at hand. They also include dedicated hardware for specific tasks (like video encoding/decoding), which can be more efficient than performing these tasks on general-purpose cores.
AMD and Intel design chips for a wide range of devices and systems, requiring a more generic approach. This limits their ability to optimise for specific hardware and software combinations. Apple designs chips solely for its own products, allowing for tighter integration between chip, hardware, and software.
I understand that Apple controls much more its ecosystem than either AMD or Intel, but I cannot imagine a M1/2/3 being less "generic" than an AMD/Intel, care to elaborate?
Also, the question is what makes the architecture more power hungry. Originally there was this disctinction of CISC (intel/amd) vs RISC (arm/apple) but my understanding is that at least intel/amd have some cisc runnning on risc architecture, meaning, externally it looks like a cisc/x64, but internally is a risc or riscish.
Pure Vision currently. The last we heard about LiDAR was the issue of combining Lidar Data with Vision in the NN. Which technology should take priority in the NN? For example LiDAR might recognise the shape of a sign but vision will know if its a stop sign or not.
I'd be super interested to see if someone has successfully combined RGB data with LIDAR data.
This is basic university robotics. Sensor fusion ! There are a whole host of techniques for dynamically updating confidence between two or more sensors estimating the same values. Kalman filter being the standard approach (which for reference was used in the apollo missions 50 years ago - developed a lot since then)
And yes, it is commonly applied with vision models. There are a host of combined rgb/lidar, structured light, depth camera and more setups in the labs students are working on at my local university, and have been for at least 6 years
For reference, I, a computer science undergrad, learned this kind of sensor fusion theory in an elective class that was just an excuse for a soon to retire professor to play with lego robots
You can know more about sensor fusion than Elon does by reading a literal blog post.
The NN shouldn't have to "choose" one or the other. It's a classic "Not even wrong" question!
mailerlite.com has a free tier for 12,000 emails per month but limited to 1,000 subscribers.
If you don't mind writing a bit of javascript, you could use Cloudflare workers which allow sending emails via MailChannels API. Be sure to setup certificates so people cant spoof your domain.
Interesting post, I've not used YAML outputs as of yet. When using GPT3.5 for JSON, I found that requesting minified JSON reduces the token count by a significant amount. In the example you mention, the month object minified is 28 tokens vs 96 tokens formatted. It actually beats the 50 Tokens returned from YAML.
It seems like the main issue is whitespace and indentation which YAML requires unlike JSON.
Yes, minified JSON would be even less tokens than YAML.
But:
1- LLMs tend to have very hard time to produce minified (compacted) JSON in the output, consistently.
2- As for compacted JSON input- Empirically it seems that LLMs can process it quite well for basic cognitive tasks (Information Retrieval, basic Q&A, etc), but when it comes to bit more sophisticated tasks it fails compared to exactly the same input, uncompressed. I've mentioned and provided examples in the comments of this article.
What do you enjoy most about software dev? Open source is great for basically any area of software development, you could create some PRs to existing projects you find interesting. I'm a frontend dev but recently enjoyed learning about Deno and built a tiny blog using their Deno KV service, I'm now trying to built a super basic sandbox like codepen using Cloudflare Workers and KV.
If you're not a fan of serverless, fly.io or render.com are great for hobbyist projects.
Also ChatGPT can be great to help work out a project idea.
I like the direction of https://slate.auto. Module, bring-your-own-computer. We'll see if they allow affordable trade-in's for upgrading battery/motors.
They could also work with CommaAI for autonomy.