Great for bombing hospitals and other infrastructure at least...

> The sky is not falling, and WebUSB is not going to cause it to fall.

You could always write a native app. It's always been possible that way.

Sorry, no, I am not supporting Windows, MacOS, iOS, Android or multiple other platforms when I could just target one single platform - the web browser.

>WebUSB isn't a driver, it relies on underlying usb drivers.

I never said WebUSB "was a driver". But it does not rely on specific usb device drivers being installed for specific devices.

>What is the need here for a webpage that also needs to access specific hardware not exposed generically?

There are already plenty. You don't need to ask me because you'll likely just shit on what I write, so here's a list compiled by someone else:

https://github.com/webusb/awesome

That just scratches the surface.

I'm more interested in WebBluetooth, which Apple is also blocking from standardization. I manufacture a bluetooth enabled device that I'd like to have a simple web application to interface with, rather than needing to pay Apple for the privledge to develop an app for their app store, where they can then extort me for $$$ for any sales made through the app.

It's no different for WebUSB, it has many, many uses, but Apple is choosing profit over progress.

Alcohol costs the UK 4-5x more than smoking. Coincidentally, it's the upper classes drug of choice. Must be a coincidence though

I’d say cocaine is the upper class drug of choice. Regardless, alcohol is every classes drug of choice. The debate over whether the government is hypocritical or not kind of ignores the reality that British voters don’t want alcohol banned. So the government isn’t going to ban it. Which is broadly what you’d want a government to do!

> Upper classes drug of choice

You're joking me. It costs more in Australia for a pack of cigarettes than it does for multiple beers or even a bottle of decent wine.

Alcohol is not the upper classes drug of choice, its all classes drug of choice.

As the US found out, alcohol is very very hard to ban because it is very very easy to make.

Also popular and part of the culture. Jesus turned water into wine but didn't pop out for 20 Marlboro.

Weed and tobacco are also very easy to make. They literally grow on trees[1].

[1] Technically, herbaceous plants.

Really in the case of tobacco, (almost) no one is going to grow it. It's a massive pain in the ass when most people are addicted to the nicotine. Synthetic nicotine in vapes are what would be black marketed these days.

It's way easier to ship as well discreetly, borderline impossible to seize in reality, which is probably one of the reason in SEA they are about to ban vaping, it's really a huge gateway to transport anything, very rarely LE is opening open and testing what the vape contains, so transporting large amount of any substances has never been easier.

Weed and tobacco smoking are also easy to detect by people who don't want secondhand smoke. And if it were illegal, they could report it.

At least alcohol produces side effects that people enjoy. Smoking pretty much only has negative side effects once you get hooked.

Have you ever smoked? I feel like I only hear this sentiment from people that have never tried nicotine.

Nicotine absolutely produces effects that people enjoy. Smokers don't just do it because they want to smell bad and look cool.

Yes, I smoked for a decade. The only noticeable effect it produces after a while is providing relief from nicotine withdrawal symptoms. It does feel similar to regaining focus or calming your nerves, so smokers trick themselves into thinking that's what it actually does. Nicotine is also way, way more addictive than alcohol. I've gone months without alcohol with almost no mental effort but day 3 of quitting smoking was probably one of the most miserable and challenging of my life.

Is there proof that the positive effects are still there after you're hooked? Or are the "positive effects" at that point just a cessation of the negative effects of withdrawal?

Yes, absolutely. It's a stimulant, similar to caffeine. Just like how nearly everyone adjusts their caffeine consumption based on the situation (got to buckle down, drink an extra cup of coffee), people do the same with nicotine. It also still works as an appetite suppressant.

Now, the euphoric effects that you get at first, those very rapidly go away with tolerance. With habitual use, you probably only experience a tiny shadow of that with the first hit of the day, or a respectable replay if for whatever reason you go a couple days without (which is heightened by the cessation of withdrawal). The nausea and disorientation also go away, which is nice since otherwise it would be a problem.

Alcohol is the deadliest and has the biggest social costs of any drug. Nicotine is second, heroin is a distant third.

Drugs that are largely harmless, like MDMA, are illegal with heavy penalties.

Drug policy is largely nonsense and rampantly hypocritical.

MDMA is a lot more acutely dangerous than nicotine, and somewhat moreso than alcohol. If you drink too much, you'll vomit, and for the most part be fine. Obviously that not always true (I'm sure everyone knows at least one person who had to have their stomach pumped in college), but for the vast majority of users, their body's natural defense against being poisoned works fine.

An MDMA overdose, however, needs active, external cooling to ride out. We don't really have a natural safety valve for overconsumption.

That's not to say it should remain banned (I'm quite pro-legalization myself), but it's not entirely arbitrary to have MDMA banned versus other, less acutely dangerous drugs. Better examples of unjustifiably banned drugs are psychedelics such as LSD.

Just ban smoking indoors then

Peter thiel the rational investor that dropped millions on a personal vendetta? Peter thiel the gay man who roleplays as Christian for some reason?

> I feel like my brain is starting to atrophy

On the upside, there wasnt much to atrophy in the first place

There's only so far engineers can optimise the underlying transformer technique, which is and always has been doing all the heavy lifting in the recent ai boom. It's going to take another genius to move this forward. We might see improvements here and there but the magnitudes of the data and vram requirements I don't think will change significantly

State space models are already being combined with transformers to form new hybrid models. The state-space part of the architecture is weaker in retrieving information from context (can't find a needle in the haystack as context gets longer, the details effectively get compressed away as everything has to fit in a fixed size) but computationally it's quite strong, O(N) not O(N^2).

I’ve read and heard from Semi Analysis and other best-in-class analysts that the amount of software optimizations possible up and down the stack is staggering…

How do you explain that capabilities being equal, the cost per token is going down dramatically?

Optimizations, like I said. They'll never hack away the massive memory requirements however, or the pre training... Imagine the memory requirements without the pre training step....this is just part and parcel of the transformer architecture.

And a lot of these improvements are really just classic automation or chaining together yet more transformer architectures, to fix issues the transformer architecture creates in the first place (hallucinations, limited context)

Exactly this. To actually visualize the sheer scale of the VRAM wall we are hitting, I recently built an LLM VRAM estimator (bytecalculators.com/llm-vram-calculator).

If you play around with the math, you quickly realize that even if we heavily quantize models down to INT4 to save memory, simply scaling the context window (which everyone wants now) immediately eats back whatever VRAM we just saved. The underlying math is extremely unforgiving without fundamentally changing the architecture.

Neither is this the first time nor are they really confronting it