I'm sure the EVs the Chinese are currently selling in the EU would pass NHSTA certification. Are those the $10K Chinese EVs? Nope, those are absolutely death traps that have no chance of getting certified. However, the ~$30K Chinese EVs, those have a shot of passing NHSTA certification. Honestly, I hope Ford and GM stop being distracted with hybrids and commit, but I think it will take a systemic shock like the Chinese EVs arriving to break the grip of their ICE agenda and parts revenue stream. Same thing happened with fuel efficient Japanese imports in the 80s, the old guard poo pooed them and lost that market to the imports.

Ford and GM have been crowing that consumers don't want EVs. So what harm can possibly come from letting China flood the market with EVs nobody wants? If Ford and GM are right, the Chinese makers will be shooting their own foot. However, I have a feeling Ford and GM are being a little disengenous and the issue might be that people don't want Ford and GM base EVs at $65K a pop.

Funny how the narrative has changed. Weren't the legacy automakers were shelving EV plans because consumers didn't want EVs? So the Chinese EVs should find no demand, pack up their wares, and go home. So what harm could affordable EVs possibly cause?

Unless, and this is a stretch, what if consumers DO want EVs, just not $70K EV trims of $38K cars?. In that case the problem is automakers can't, or more likely don't want to, make EVs in a price range appealing to consumers. I'm all for anti-dumping legislation, but this doesn't feel like dumping, this is a rebuke to automakers who thought they'd be able to control the narrative and avoid selling EVs to keep the parts and service money tap flowing. The automakers should head the wake up, and for example, stop offering $70K* Ford F150 EVs and give consumers a $38K Ford F150 EV. If the Chinese can build them cheap, we're going to have to find a way to build them cheap while still turning a profit.

*The F150 standard battery with 230 miles of range is sub-ICE range, 300 miles is the minimum comparable for an ICS replacement.

Does that cause an issue with insurance though? I wonder if the VIN is "the same" for insurers, so it would only be an issue if the vehicle is a model NOT sold in the US.

Dang, they can't even round properly, also there's no way they're teaching C/C++, maybe JavaScript, how do you write macros in JS?

This is wrong on many levels but I'll single out the single most important reason why Rust will never be a good choice for an OS kernel, it contains hidden control flows.

What Rust are you referring to? Rust is anything but terse, it's what I image the offspring would look like if C++ and Perl had a one night stand. Is this terse and grokable?

use std::ops::Mul;

struct Factorial<T: Mul<Output=T> + Copy + From<u8>>(T);

impl<T: Mul<Output=T> + Copy + From<u8>> Factorial<T> {
        fn new() -> Self {
                Factorial(T::from(1))
        }

        fn compute(&self, n: u8) -> T {
                (1..=n).fold(self.0, |acc, x| acc * T::from(x))
        }
}

fn main() {
        let factorial = Factorial::new();
        println!("{}", factorial.compute(5));
}

This happens often if you prompt OpenAI with non-trivial coding questions. You get the "Certainly! ..." and stuff in a code block that looks like valid code gets output. To be fair, for relatively simple questions, it's correct most of the time, but for advanced topics, the stuff it hallucinates makes it borderline useless. For example, I recently asked it to write up a call using the somewhat obscure pragma API in Sqlite and it happily spat out a block of code dereferencing structs that don't exist in anywhere Sqlite.

The root of the problem is that "coding" is basically transcribing requirements into constructs that can be consumed by a state machine. No LLM is going to be able to infer all implied requirements for a piece of software from a few sentences. So we wind up with "prompt engineering", which is transcribing requirements into prompts. When you get a long enough list of requirements, you wind up with a large list of finely tuned prompts, which takes us full circle to "I can program in ChatGPT v3 and v4".

To me a risk is that people start buying the KoolAid and low level programming starts to die. Why? AI models have to run on something. Who's going to write and maintain the bootloader? Who's going to write and maintain the OS? Who's going to write and maintain the inference engine? Who's going to write the compiler that transforms and compiles the ML model? It's possible that AI will become self-hosted, and start building its own OS and runtime components. However, that has its own set of risks. Who's going to make certain the AI didn't hallucinate and create a buggy routine that resets a bank balance to zero if someone's middle name has more than 40 characters if Friday the 13 is in December?

Part of the problem is that the vast majority of KoolAiders don't know how GANs work. GANs inject a bit of randomness to generate a "novel" rearrangement of closely weighted tokens from a training set based on an encoded request. It's not doing hard research or gathering new insights, it's generating a derivative work from the research and insights it has been trained with. Granted, sometimes it leads to a permutation that hadn't been considered before, but being trained on 80 billion ways to skin a cat isn't likely to lead to a novel breakthrough in gene therapy. Someone has to do the hard research, postulating hypotheses, figuring out what the constraints are, what kills, what's placebo effect, what are the side effects, etc. That's all the stuff that makes AI look smart, but fundamentally it's just standing on the shoulders of work and research already done by us lowly carbon based lifeforms.

The flip side that's never mentioned. What about illiquid assets that have no market comparables such as collectibles and private companies, who determines the value? What about corporate entities, are they going to be taxed on "net worth"? If not, why not setup a shell holding company to transfer your wealth into it while you hold private shares with a par value of $0.01. Your $100B in company stock would now be $0.01 of wealth for you, while your shell company holds $100B on its books.

I'll add another wrinkle, once net worth is taxed, if you subsequently realize an asset sale, you've already been taxed on that asset, so do you get to deduct the tax that's already paid? Is the sale tax free? Since you never sold the asset, there's no transaction record, so what determines the value of that asset? For stocks is it the par value at issuance of $0.01? If it's the market price, what date / time is used? What if it's for a Picasso, who determines the value?

What about extraordinary events like Musk's $50B package being rescinded, would he be retroactively be able to claw back taxes paid on that net worth?

The other bit that's never discussed is that rich people don't like wealth taxes. France even scrapped theirs after 60,000 millionaires left the country as a result, leading to an overall decline in tax revenue. Make no mistake, anyone with wealth and accountants will exploit all possible avenues to avoid a wealth tax. For all the talk about billionaires pleading "tax me more", that's nothing but a front, anyone is free to donate as much money as they want to the Federal government. If Warren Buffet would like to donate $10B to the US government, good news, that's been an option since 1842, the link is right here, no congressional action required:

https://fiscal.treasury.gov/pu...

Oddly, no "tax me more" billionaire has elected to donate billions to the US government, maybe they don't know the treasury is ready and able to accept gifts?

There already is, TPUs (Tensor Processing Unit), GPU siblings that have massive amounts of memory, custom float types, and caches optimized for training / inferring ML models. The Nvidia H200 is a prime example. I'm not sure what Sam is gunning for here, maybe it's cheaper TPUs, because H200s, if you could find one, run $40K a piece.

If his idea is to come up with a processor that doesn't rely on tensors, that would require redefining the underlying basis for virtually everything we call machine learning today, which would be a multi-trillion dollar effort.

Just what I was thinking, so HP admits the security on their printers is garbage, and the solution is to buy HP cartridges rather than fix the garbage security on the printer. I like this logic, Microsoft should use it: You're using Windows with non-Microsoft software and you have a virus, well, that's the problem, only use software written by Microsoft and you won't have viruses, also please ignore the fact that the vulnerability being exploited lies in the OS.

Your numbers are incorrect, MacOS is just over 5.5% of Steam users, so over 3x the number of Linux users.

Source: https://www.statista.com/stati...

I dunno them being a "rarity", I was around the northern edges of the North East a few weeks ago dealing with barely above 0F temps, I'm not going to say there was an overwhelming amount of EVs, but in no way are they a rarity, I'd guesstimate it was a bit over 5% of the cars I saw. The majority were Model Ys and 3s, but there were a few Ioniqs, Bolts and Mustang EVs. Now, maybe Montana and North Dakota have different adoption due to greater distances and less charging infrastructure, but there is visible adoption of EVs in colder parts of the US, at least in the east.

If eating or "attacking" plants causes them to react to "pain", can we surmise that veganism is a form of plant cruelty? (Grabs bucket of sustainably raised popcorn)

Outside the office, I would estimate >50% of the laptops I see are Macs. So Macs dropping market share doesn't reconcile with what I'm seeing in the wild. At my office, it's close to >90% Macs. In my circle of friends and family, it's probably 75% Macs. I haven't personally seen any evidence of Windows is making a coming back. Granted, my experience is US centric. I'm sure less wealthy countries have a higher proportion of cheap Windows machines given there's no such thing as a cheap Mac. The limiting factor here is "desktop", because if we remove that filter, Windows is a small minority of all computing devices with Linux / UNIX holding the dominant share. To put that into context, there are almost 2x as many iOS devices as Windows devices in the world. If we include all Linux devices (equipment / servers / appliances), then Windows has 10% OS market share. In the grand scheme of things, Windows is becoming less and less relevant, try to name one startup that hung it's hat on Windows in the past decade. There are none.