There will still be fraud, but when you remove the layers and costs private insurance add on, it is much easier to disincentivize and detect fraud, especially cases were people up code or change codes to get paid more.

Because you just have one source of data coming in and out.

By the way, actual doctors also play fast and loose with coding, but that's to get paid for actual care.

What stunningly bad marketing. Look, Microsoft did do a lot of stuff, but even the most strident fan would find it hard to compare it to a fundamental breakthrough in physics.

I use a lot of Microsoft stuff, but I know the success of a lot of players in the early PC years was on the back of really shady stuff.

They really are trying to bring in collectors that have no idea of the history of computing here, I think.

For endless reasons. I mean, if he wants to train a LLM on all his, err, musings, go ahead, but nobody is keeping that going for long.

Frankly, a bot that just spews out variations of "I'm so smart" and "the singularity is a thing because (random)" would be just as good.

People get old, they die. When they can die well, that's a good thing.

Not sure why anyone would expend the effort for that particular feature. And I use Windows daily and there's a lot I like about Windows 11, but I am accepting that there will be this wave about putting Cortana 2.0, err, CoPilot in everything until this LLM craze flattens out.

Not that I mind smarter tools. It would be nice if I could type "Hey, I need to get milk when I am at the store" at, say, Cortana Redux and a polite notification would pop up when I am in a grocery store.

I know, privacy and all that, but so much data is available on anybody that I'll take some wins from it when I can.

I don't see the point of Recall. It just seems like a tool to help those that whose attention is so spread thin they need snapshots of the past hour to remember what they were working on. Don't get me wrong, browser history is useful, but this is just too much.

And it will affect the counties that consistently vote for people that would never even consider the idea that health care is fundamental right that every single person should have access to without fear of cost.

Basically, they will ring as much as they can out of rural and community hospitals, but that is short lived. The doctors can't even come in and stop it, because regulations allow a vulture equity firm to buy up hospitals, but they can't be employee or doctor owned.

The doctors will burn out and move to urban areas where there is still demand and actual non-profit systems that are trying like hell to stave off for-profit acquisition.

It's either single payer or a health system that is completely collapsed under mismanagement for profit.

But, no, we need to spend trillions on weapons programs that don't work, prop up defense contractors that just leech off our taxes to maintain stock pricing.

You could cut 4 trillion of defense spending over 10 years and you get nothing but a more efficient and nimble defense force. Sure, we can't spend trillions trying to install democracies in random countries.

But, I'll take that over nobody ever going into bankruptcy over medical bills by a system where insurance and equity take up 33% of very cent spent to just deny care.

CUDA is a very thin abstraction over NVIDIAs hardware. So, getting CUDA ported over to any other architecture is possible, but it won't be as efficient. And, CUDA isn't open source, so there's no way any tweaks or fixed that AMD (or Intel) would need to work better with their GPUs would be accepted.

There's just a lot of inertia around CUDA. It's been used for over 10 years and it just worked. It is only until now that there is such a demand for this kind of hardware outside supercomputing that the issue of being tied to one vendor is now a very pressing one.

This seems to fall to the misconception that the hardest part about writing is the initial draft. It isn't. It's the conceiving what to write and the rewriting and edits it takes to get a good piece. This is the exact same thing with coding. AI generated text quickly falls to such a low level that it puts more of a burden on the editors than editing a decent first draft.

People are so impressed by transformer based LLMs being conversational that they just overlook the actual quality of the generated output objectively. They give it more credit because it is being done at all.

And it all does fall apart. I asked to ChatGPT to create a class that uses bipolar coordinates to represent a vector. It happily spit out the code. For a polar vector.

Intelligence is not a thing that is so precious we need to mimic it. It fact, it is one of the most wasted resources around. The whole point of pushing AGI as a productivity revolution is to scare workers out of fear of becoming obsolete.

And there is a simple solution to all this. Organized labor and collective bargaining. If companies try to make it too hellish for people to work solely to grow market value, you don't get access to the skilled labor that you still need. And by doing that, you actually make salaries and working conditions better for everybody. Could LLM could be a very useful tool for workers. Sure. Just note how everybody goes to how it will replace workers versus help them do work with less stress or time? Odd that.

Without access to the Google Play Store, it was doomed. You think that such a feature would be a boom to Android development, but the development chain is all about emulation that rules out running production apps.

You aren't going to get some magical leap that allows a LLM to run affordably in desktop computer, much less a laptop or phone. So, you have to have these in the cloud, but you still have to power and cool them all. And you aren't getting 2x the performance every 18 months. It was three years at best and terms of tensor accelerators, it will probably take five years to see if feature shrinks will allow for core and memory gains for that next 2x.

The push to run smaller models with AI specific chip features is weird, because, well, these models don't scale down well at all. You can't just carve off bits of it for specific stuff and have still work. But, hey, sell that new hardware I guess.

In the end, the improvement it has over regular search just isn't worth the massive extra costs. It still requires humans to review the output to see if makes sense, and that is always the case, because no matter how fancy the math, there is always a tradeoff between precision and recall in learning models.

Unionized labor and collective bargaining. I know people complain about how unions protect lazy people and are bloated and out of date, but most of that is based on considered anti-union campaigns by companies that want shareholder gains over everything else, including the workers that actually make those profits. Let's be honest, there's some lazy people in all the workplace that manage to worm their way into an organization, unions or not.

Also, unions are organized by its members and I think software engineers and IT people could actually make an effective, innovative union that works well.

Fabs are so incredibly expensive and require so much specific talent that you can't just apply the startup model to making them. You work with the established players who are (looks around) already going at a pretty good pace to make new fabs already.

This is one thing I find odd that nobody is talking about with LLM based AI. We know it takes some heavy duty silicon, but they are acting as if Moore's law was still in play and we aren't up against some really tough thermal and power limits. There's no "twice the computation" for a small enough bump in watts and heat.

Honestly, I hope this AI bubble pops quickly. I know it won't, there's still too much cash sloshing around. But, honestly, human intelligence isn't really hard to find. It's frankly wasted.

Honestly, I admire how well web developers are keeping themselves busy and self employed by having to constantly rewrite web sites because Javascript frameworks have the shelf-line of a tomato in the sun.

Honestly, Microsoft, just make the Store desktop application work better. Argh.

While I do think that software programmers are not as different as some of them would like to think, there are real advantages to not treat all workers the same.

There are plenty of people that are fine getting paid to watch pretty mindless content and take easy quizzes. However, there are those employees that uniformly hate such things.

Of course, those are often the employees you want to retain, because they often want to just get stuff done.

There is no specific reason for this user to run Linux. and the cost of a license will be nothing compared to the countless lost hours of trying to learn something new.

Also, I am constantly reminded how little power users
understand regular users. Use the man command? By the time you managed to explain how to get a terminal window open and the basics of using a command line, you could have Windows up and running. And man documentation is next to useless for non power users.