The other thing that most people don’t focus on is how we train LLMs.

We’re basically building something like a spider tailed viper. A spider tailed viper is a kind of snake that has a growth on its tail that looks a lot like a spider. It wiggles it around so it looks like a spider, convincing birds they’ve found a snack, and when the bird gets close enough the snake strikes and eats the bird.

Now, I’m not saying we’re building something that is designed to kill us. But, I am saying that we’re putting enormous effort into building something that can fool us into thinking it’s intelligent. We’re not trying to build something that can do something intelligent. We’re instead trying to build something that mimics intelligence.

What we’re effectively doing is looking at this thing that mimics a spider, and trying harder and harder to tweak its design so that it looks more and more realistic. What’s crazy about that is that we’re not building this to fool a predator so that we’re not in danger. We’re not doing it to fool prey, so we can catch and eat them more easily. We’re doing it so we can fool ourselves.

It’s like if, instead of a spider-tailed snake, a snake evolved a bird-like tail, and evolution kept tweaking the design so that the tail was more and more likely to fool the snake so it would bite its own tail. Except, evolution doesn’t work like that because a snake that ignored actual prey and instead insisted on attacking its own tail would be an evolutionary dead end. Only a truly stupid species like humans would intentionally design something that wasn’t intelligent but mimicked intelligence well enough that other humans preferred it to actual information and knowledge.

Yeah, even an established creator is going to have a hard time moving their audience.

If YouTube weren’t a near monopoly it would be different. Then other companies would be competing for creators.

Making it worse is section 1201 of the DMCA. It makes it a crime to circumvent access controls. In the past, Facebook was able to grow by providing tools to interface with MySpace. People didn’t have to abandon their MySpace friends, they could communicate with them through Facebook, and Facebook could ensure that messages sent on its platform arrived to people still on MySpace. But, if you tried that today Facebook has access controls in place that make that a crime. The same applies to YouTube. Nobody can build a seamless “migrate away from YouTube” experience because YouTube will use the DMCA to block them.

The governments of the world need to bring back antitrust with teeth and force interoperability.

I’m not even convinced that an “AI tutor” is better than nothing.

Probably better than sitting in a room staring at a blank wall. But, is it even as good as playing with lego? Playing tag with friends? Drawing with crayons? Taking a pet for a walk?

All other things being equal, it would save a lot of lives to replace every human driver with a Waymo car right now. They’re already significantly better than the average driver.

But, there are a few caveats. One is that so far they’ve only ever driven under relatively easy conditions. They don’t do any highway driving, and they’ve never driven in snow. Another one is that because they all share one “mind”, we don’t know if there are failure modes that would affect every car. Every human driver is different, but every human is more or less the same. If a human sees a 100 km/h or 60 mph speed limit on a narrow, twisty, suburban street with poor visibility, most of them are probably going to assume it was a mistake and won’t actually try to drive 100 km/h. We don’t know if a robo-vehicle will do that. AFAIK they haven’t found any way to emulate “common sense”. They might also freak out during an eclipse because they’ve never been trained for that kind of lighting. Or they might try to drive at normal speeds when visibility is obscured by forest fire smoke.

There’s also the side effects of replacing millions of drivers with robo-cars. What will it do to people who drive for a living? Should Google/Waymo be paying most of the cost of retraining them? Paying their bills until they can find a new job? What will it do to cities? Will it mean that we no longer need parking lots because cars come and drop people off and then head off to take care of someone else? Or will it mean empty cars roaming the city causing gridlock and making it hell for pedestrians and bikers? Will people now want to live in the city because they don’t need to pay for parking and can get a car easily whenever they need one? Or will people now want to live even farther out into the suburbs / rural areas because they don’t need to drive and can work in the car on the way into the city?

Personally, I’m hopeful. I think they could make cities better. But, who knows. We should move slowly until we figure things out.

a silicon valley AI project to put transit workers out of work

Silicon valley doesn’t have objectives like “putting transit workers out of work”. They only care about growth and profit.

In this case, the potential for growth is replacing every driver, not merely targeting transit workers. If they can do that, it would mean millions fewer cars on the road, and millions fewer cars being produced. Great for the environment, but yeah, some people might lose their jobs. But, other new jobs might be created.

The original car boom also destroyed all kinds of jobs. Farriers, stable hands, grooms, riding instructors, equine veterinarians, horse trainers, etc. But, should we have held back technology so those jobs were all around today? We’d still have streets absolutely covered in horse poop, and horses regularly dying in the street, along with all the resulting disease. Would that be a better world? I don’t think so.

It’s another project to get AI money and destroy labor rights.

Waymo obviously uses a form of AI, but they’ve been around a lot longer than the current AI / LLM boom. It’s 16 years old as a Google project, 21 years old if you consider the original Stanford team. As for destroying labour rights, sure, every capitalist company wants weaker labour rights. But, that includes the car companies making normal human-driven cars, it includes the companies manufacturing city buses and trains. There’s nothing special about Waymo / Google in that regard.

Sure, strengthening labour rights would be a good idea, but I don’t think it really has anything to do with Waymo. But, sure, we should organize and unionize Google if that’s at all possible.

Transit is incredibly underfunded and misregulated in California/the USA

Sure. That has nothing to do with Waymo though.

robotaxis are a criminal misinvestment in resources.

Misinvestment by whom? Google? What should Google be investing in instead?

AFAIK they’re as safe as SawStop table saws. There has only ever been one collision involving a Waymo car that resulted in a serious injury. It was when a driver in another car, who was fleeing from police, sideswiped two cars, went onto the sidewalk and hit 2 pedestrians. One of the cars that was hit was a Waymo car, and the passenger was injured. Obviously, this wasn’t the fault of Waymo, but it was included in their list of 25 crashes with injuries, and was the only one involving a serious injury.

Of the rest, 17 involved the Waymo car being rear-ended. 3 involved another car running a red light and hitting the Waymo car. 2 were sideswipes caused by the other driver. 2 were vehicles turning left across the path of the Waymo car, one a bike, one a car. One was a Waymo car turning left and being hit on the passenger side. It’s possible that a few of these cases involving a collision between a vehicle turning and a vehicle going straight could be at least partially blamed on the Waymo car. But, based on the descriptions of the crashes it certainly wasn’t making an obvious error.

IMO it would be hard to argue that the cars aren’t already significantly safer than the average driver. There are still plenty of bugs to be ironed out, but for the most part they don’t seem to be safety-related bugs.

If the math were simple and every Waymo car on the road meant one human driver off the road with no other consequences or costs, it would be a no-brainer to start replacing human drivers with Waymo’s tech. But, of course, nothing is ever that simple.

Source: https://www.understandingai.org/p/human-drivers-are-to-blame-for-most

Waymo times than Teslas?

Imagine you’re the guy who invented SawStop, the table saw that can detect fingers touching the saw blade and immediately bury the blade in an aluminum block to avoid cutting off someone’s finger. Your system took a lot of R&D, it’s expensive, requires a custom table saw with specialized internal parts so it’s much more expensive than a normal table saw, but it works, and it works well. You’ve now got it down that someone can go full-speed into the blade and most likely not even get the smallest cut. Every time the device activates, it’s a finger saved. Yeah, it’s a bit expensive to own. And, because of the safety mechanism, every time it activates you need to buy a few new parts which aren’t cheap. But, an activation means you avoided having a finger cut off, so good deal! You start selling these devices and while it’s not replacing every table saw sold, it’s slowly being something that people consider when buying.

Meanwhile, some dude out of Silicon Valley hears about this, and hacks up a system that just uses a $30 webcam, an AI model that detects fingers (trained exclusively on pudgy white fingers of Silicon Valley executives) and a pinball flipper attached to a rubber brake that slows the blade to a stop within a second when the AI model sees a finger in danger.

This new device, the, “Finger Saver” doesn’t work very well at all. In demos with a hotdog, sometimes the hotdog is sawed in half. Sometimes the saw blade goes flying out of the machine into the audience. After a while, the company has the demo down so that when they do it in extremely controlled conditions, it does stop the hotdog from being sawed in half, but it does take a good few chunks out of it before the blade fully stops. It doesn’t work at all with black fingers, but the Finger Saver company will sell you some cream-coloured paint that you can paint your finger with before using it if your finger isn’t the right shade.

Now, imagine if the media just referred to these two devices interchangeably as “finger saving devices”. Imagine if the Finger Saver company heavily promoted their things and got them installed in workshops in high schools, telling the shop teachers that students are now 100% safe from injuries while using the table saw, so they can just throw out all safety equipment. When, inevitably, someone gets a serious wound while using a “Finger Saver” the media goes on a rant about whether you can really trust “finger saving devices” at all.

Anyhow, this is a rant about Waymo vs. Tesla.

Now you just need to slay the Apostrophe Monster.

Except Gates is a piece of shit. You don’t need to shout at Gates, but nobody should ever meet him and treat him like a human.

No she wasn’t. She was never part of IBM at all.

She simply knew the chairman of IBM because they both served on the United Way board of directors. She was also a lawyer, as was Gates’ dad, which is a likely reason that the contract that Bill signed with IBM was so incredibly friendly to Microsoft.

https://en.wikipedia.org/wiki/Mary_Maxwell_Gates

Who’s Gate?

If Microsoft hadn’t been around Apple would have probably defined the early PC era. The Apple II was released in 1977, 4 years before IBM decided to enter the home market with the PC.

Or Commodore might have been the one to dominate. They sold about 5 million Amigas.

Or it could have been NeXT after Jobs was forced out of Apple and started a new computer business.

The winner turned out to be Microsoft, but desktop computers were well on their way to being a standard thing long before Microsoft / IBM got into the market.

it’s the reason so few people use FOSS products.

It’s a reason. Another reason is all the stuff that Microsoft was found guilty of doing during their conviction for abusing their monopoly.

The last major American privacy law, the 1988 Video Privacy Protection Act was passed in 1988 by Reagan. The only reason it happened is that politicians realized that their privacy was affected. Robert Bork was going through his Supreme Court confirmation hearings and someone got a hold of the tapes he had rented and published them.

Politicians were worried about their own personal privacy, so they passed a new law to protect the privacy of people’s video tape rentals.

Maybe the fact that the targets here were politicians will mean that something will happen with data privacy, for once.

Back when I first started using Linux, it was rare to have more than one PC in a house. Now I personally have 3 computers, a desktop and a couple of laptops, and a tablet, and a phone, and some old barely-working tablets and laptops in a drawer.

It is definitely the case that I’ve had to use one of the other machines when the Linux desktop had issues. OTOH, I’ve also had to use other computers to help me out with a Windows issue (though it wasn’t an OS error, it was a drive that went bad).

It’s funny though. Back in the day when I only had the one computer, I was able to troubleshoot issues with it while still using it. That was probably only possible because tech was less advanced. For example, it was possible to browse the web effectively using a text-only client. Back then websites were simpler and Javascript was pretty much non-existent, so if you were troubleshooting a graphical issue you weren’t so crippled. Similarly, you weren’t so crippled if you couldn’t use GUI programs, because in those days almost every GUI program had a console equivalent that worked as well if not better.

These days, it’s pretty likely that the info you need will be on YouTube – obviously not very useful from a console, or a Discord chat – same problem.

My NAS is one place where I wouldn’t risk anything that isn’t rock solid. Even if you don’t lose data, the NAS is infrastructure that should always be available.

I’m used to (on Windows) occasionally having the nVidia driver break things so the computer blue screens. At that point, your computer is shutting down and there’s nothing you can do about it.

It was weird under Linux when I had an nVidia bug and the display stopped working, but the computer was still alive. I was able to SSH in and do a graceful shutdown. It was weird to watch because my display was completely frozen. The mouse pointer didn’t move, the clock wasn’t updating, but the windows were still all there. But, behind the scenes everything was working normally (bar high CPU usage because something else in the system was bothered by the display being screwed).

As nice as it is that Linux responds a bit better to bad nVidia drivers, it’s also annoying how poor the quality of those closed-source drivers is. There are certain kinds of bugs that apparently have been issues for years and nVidia just isn’t addressing them.

Back in the days when you needed to write your own modelines, that definitely wasn’t true. You screw up your modelines and X emits signals that your monitor can’t handle and you’re out of luck. It was very normal to spend a lot of time editing your Xorg.conf file until it worked with your monitor.

You must have come along at a time between fiddling with modelines being a thing, and Wayland taking off.

Everyone around him must hate him so much, to allow him to go out in public looking like that.