If trash will end on the oceans anyway, it's better to dissolve into microplastics than not dissolve and trap marine life.

It makes more sense than you think. Most of Reddit content is stupidity, but human stupidity. LLMs need human intelligence (even lack of) to be trained, because training on LLM content quickly degrades the training. And Reddit have one of the largest human-generated repositories out there.

They saved $2 million without increasing staff. And the issue is not the current price they pay, but that the price keeps increasing too much. So in 10 years they will still be paying around the same for the engineers, but would pay 2 or 3 times more than that to Amazon.

It's a paradox, but this helps bringing in more clients. If they don't charge for egress, anyone can try Amazon just to see how it works, and leave if things aren't good. Amazon then gives them a lot of "free for an year" resources, they end up entangling themselves to the neck on Amazon infrastructure, and cannot leave.

If Amazon charges outrageously high fees for egress, clients won't even try because "if we don't like Amazon we have to pay too much to leave."

I saw elsewhere that a large cylinder produced a few microvolts.

The jump from the 486 to the Pentium is greater than any jump from the last decade. My daily driver is a Xeon E5-2667 from 2014, and paired with a RTX 3060 I can run Cyberpunk in 1080p on Ultra at 70-90fps, and it's a processor from a decade ago. It's like running Windows NT in a 386 with great performance, not "to show it runs."

The company already folded, the owner got 3 months on prison IIRC. There's no point on keeping the company under supervision, the market closed all doors for them, nobody trusted them anymore.

I understand that the company should have worked better to protect their data, but they are more a victim here than a culprit. Calling both the company and the scumbag that hacked them the same is an injustice, it's like calling a ransomware victim a culprit because they didn't secured their systems. They were negligent, their database got hacked, their customers got harmed, they got put out of business. It was not on purpose. The scumbag piece of trash did all this on purpose.

And six years for him is too little. But at least he got some prison term...

I think this is as far from OK. In Brazil a guy got charged and thrown in jail for stealing cell phones based on dark, pixelated surveillance camera images, and even proving that he was more than 50 miles away minutes after the alleged crime, he still got jailed because AI Face Recognition said it was him.

If AI cannot be used as evidence, it must be thrown away from the proceedings. Anything done by AI can and will influence anyone involved, and if the judge already have a bias against the person in front of him, those biases will be reinforced when AI commits a mistake against the defendant. And if the AI commits a mistake in favor of him, it's easy to dismiss as "AI error."

I've seem and read about enough misjudge cases that I doubt this ends well.

Ever heard of GPS Spoofing? The camera uses GPS for location data, and probably for timestamping (GPS is basically a very precise date-time broadcast anyway). Spoof the GPS and the camera will put the intended location AND timestamp as legit.

My company forbids me from running Brave because "Brave have internal VPN that can be activated and bypass enterprise policies."

I doubt they will allow Chrome if this is activated and they cannot deactivate it using group policies or registry keys.

It's not a server application (I hope), so the telemetry will just access Intel servers and post a bunch of data. Depending on how the data is encrypted and signed, someone could abuse this service by posting a flood of fake data.

It could be abused, as any piece of software, but I don't expect remote code execution on it. Maybe a local privilege escalation, an unsafe driver leading to code running as SYSTEM, things like that.

The cleaning guy went to great lengths to stop the alarm. Isn't a case of "unplug the freezer to plug the vacuum..." here, it was a deliberate action that took some effort. "Don't disconnect anything." is a simple instruction that a 3 year old can understand, no need for specialized cleaning services...

The strike isn't about moderation, is about SE (StackExchange) telling one thing to the mods and applying another thing site-wide without much notice. Mods have been fighting AI-generated content and suddenly SE says "no more bans on users posting AI content."

I work with z Series, but on Linux, not zOS. And a System Z running Oracle DB on Linux is a beast. It's even faster than an quivalent Oracle Exadata costing double or more. I migrated one Oracle database for a client once, and in 2-3 years the savings on licensing paid for the mainframe. We migrated from a 16 processor cluster to a 2 processor mainframe, with more performance, and paying 1/8 in license. So only with the savings, he could get a shiny new mainframe every 3 years, forever.

A mainframe IS expensive to buy, no way around that. But if you are running a product that charges per processor (Oracle, OpenShift, and others), it's a no-brainer. Linux running databases can achieve 20 or even 30 times more performance per processor than any x86 system. So if space, electricity or licensing fees is an issue, slap a mainframe running Linux on your solution and be happy.

I've doing something like this myself for years. I run Linux, and have a script running every minute taking screenshots and processing them. I use ImageMagick tools to treat the files: "import" to take a screenshot, "compare" to say how much the picture changed since the last run (and stop the script if nothing changed), "convert" to make two versions: one bigger with contrast up to feed "tesseract" for OCR, and one smaller with less colors for storage.

A few times a day I feed the text to ElasticSearch (when the load on the system is low) and I have a PHP script to search for things. So when I need something I read or saw, I can search and have both the text and the image. Tesseract results are mixed, sometimes flawless that I don't need to look at the images, sometimes is a garbled mess. I've tried AI-powered OCR, they work better, but turns my computer into an area heater.

It uses less than 700MB per year, so I believe it's worthy. And uses only tools on my system, so even if that startup says "it never leaves your computer" you cannot say for sure that someday in the future that changes and nobody notices. It's portable too: the pictures are png, the text are plain .txt, ElasticSearch indexes can be exported (or you can import them on any other fuzzy text search). If that startup dies (or forces itself out of your system), you would lose everything you saved.