But AI cuts out the social aspect, like a comment "you do know that you need to pass an exam, right?" from the person you copy from. And AI makes this far too easy to do.

Of course. When I was in university (pre dot-com era) the Internet was around to communicate with (everyone exchanging ICQ numbers was a thing), but if you wanted to copy off someone you either had to bump into them on campus, or phone them at home, and if you did that, you had to do it at reasonable times because parents and other things often hate the phone ringing while they're trying to sleep. Or if it's the weekend, they may not be home.

ChatGPT is available 24/7. It's right there waiting for you to use it. It doesn't matter if it's 5 minutes before you have to hand in an assignment.

It's still not AI that's the problem - it's students not putting in the work. Whether it's in the past and copying off other people, or it's currently using AI, it's the same result - students who cheat will do worse.

Well, given that most of the high level people in government were basically plucked from the on-screen talent of Fox News, I think the phishing campaign might be easier to detect as the emails would appear to be written by someone competent.

After all, Trump wants to invade Portland (OR) because Fox decided to show some 2020 footage of a protest, and Trump thinks that's still happening, half a decade later.

It could be as simple as students using AI where they used to use someone else.

After all, in the "pre-AI" age, students would routinely copy code and other things from other students, and you can tell they did it because their grades generally were worse.

These days, I'm sure instead of asking other students, they are asking AI, and all we're seeing is the same thing - the students of the past simply copied one another, the students today, rather than copying, ask AI. Just like students of the past used paper mill sites to write their homework, now use ChatGPT to do same.

Of course, I can't say I am completely innocent of the practice - we were in a group doing a group project. I and someone else smarter than me were working on something extremely complex for the rest of the group while they worked on something much simpler for another class. In the end we figured out the complex work, and presented it to the rest of the group to learn from our knowledge, and they submitted the group project with all our names on it. I saw what they submitted and studied from it so I did learn, and the rest of the group learned from us where we all got good grades in the end. It's just if it was divided equally among everyone the two projects would've been a mess and it was simpler if we split the work the other way

8bpp support and HDR are not the same thing. You can have 12bpp JPEGs that are NOT HDR. They are often called "deep color". Many newer digital cameras support 12bpp captures and you can display them on a 8bpp display by dropping the lower 4 bits.

JPEG may not support HDR, even though it supports 12bpp images. While you need more bits to support HDR, often they can be supplied as a separate map because you only need extra bits for the extra bright pixels (HDR means you have increased headroom to handle really bright areas, but those are generally only a portion of the image).

You cannot display an HDR image on an SDR display without a technique called tone mapping that maps HDR values to SDR in a way that avoids crushing the dim shadows to black or oversaturating the bright areas to white. That includes a 10-bit HDR image on a 10 bit SDR display - you must apply a tonemap.

So Google using a JPEG base layer plus HDR map may be the proper way to support it - SDR devices can decode the JPEG data as-is, while HDR aware apps will decode the JPEG base layer and then apply the HDR highlights.

Which processor?

Probably a middle of the road STM32 or something similar .

A modern SoC like what's in your smartphone doesn't have a 2000 page datasheet. It doesn't have a datasheet period - Qualcomm and such work with people who develop products and they have a whole collection of documents from register lists, to subsystem documents, to various technical addenda like boot processes, ROM elements, etc.

A low end microcontroller probably has a datasheet on the order of a few hundred pages that documents everything it does.

A more complex chip like a low end ARM SoC would factor in around 2000 pages or so, like a STM32 class processor.

Also it would generally be like that because anything more sophisticated and you use the code provided by the vendor - like for the aforementioned Qualcomm and such SoCs, Qualcomm provides the reference code you start from so you can compile and run and start developing, referencing the supplied documents only when you need more details.

Let's compare, shall we?

Little Bobby Tables:

• No framework required: conventional database entry + payload only
• Wreaks havoc in an instant
• Total size: 32 bytes

This:

• Downloads ollama (672 MB, on Windows)
• Downloads a 14 GB data file for the model itself
• Requires a bare minimum of 16 GB of VRAM—and still runs like absolute molasses, eating up all resources
• Total size: 15 GB

Personally, I'm on Team Tables here. Maybe in a decade or three this will be practical.

I plan to achieve this by the late 2030s, maybe 2040.

In other words, it's just handing the technology over to CATL in China.

CATL got tons of Chinese government money to do big R&D and are currently the leaders in sodium ion technology. And the batteries are to be cheaper (lots of sodium ions around - we have oceans full of it, and the anode and cathode materials are cheap and plentiful).

Government investing in new technology helps promote use of that technology. Trump is basically handing China the next generation battery technology and likely to be able to sell it stupidly cheap suppressing research in the technology by any other nation. Even if the next administration restores funding, the time lost could basically mean there's no way you could catch up to China's advancements and cost.

There's also another dirty little secret - you can't design a chip for one fab and then transplant it over to another fab. Every fab is unique in its own special way and there are unique design rules for each fab.

Even if TSMC runs a new fab in AZ, you can't just take a chip designed for the fab in Taiwan and run it on AZ unaltered. You have to redesign it and get all new masks made (at $100K each mask - so a chip may need 20 or more masks making the cost of a new mask run over $2 million). So you have to redesign your chips to the new fab rules which can take months to a year depending on how much debugging you need. And because the rules change, the performance envelope changes, so your chips under the new fab may not run as fast so you need to redesign it to be more performant, or rework the masks so variances in exposures give you faster transistors.

Each fab is special and even if you have the same equipment, each has unique quirks that make it so you can't just run chips on it.

There's also a huge supply chain issue - Japan is a big supplier of EUV supplies, namely the chemicals like the photomask agents and other things.

For 20 years, plus or minus, personal computers reversed that idea.

It's likely just because spending long times sitting on the toilet causes hemorrhoids. It's just that doom scrolling is what people are doing now on the toilet, rather than reading.

There were previous studies saying spending more than 15 minutes or so per movement increased your chances.

Hope it has a charging socket too. I think people don't realize that pulse-width modulated LED lights don't really provide enough useful light to charge anything, but maybe this is for road warriors who want to show off their Logi gear on the beach.

Must be those fancy dimmable lights, because plain old regular LED bulbs don't use PWM - they are using a linear regulator that just controls the current through the bulb. This compensates for varying line voltages (so they don't dim when you turn something big on) and such.

But the solar keyboard they used to make didn't need much power at all - I've had it for a long time now and replaced the battery twice. The battery prefers to be very shallow discharge and charge so the first time I had it replaced under warranty (but picked up a cheap replacement battery off eBay) where I learned if you discharge it a tiny bit before charging it, its life quadruples than if you discharged it deeply then charged it.

(it's like you can run 120mAh through it at shallow discharge but only 30mAh at deep discharge. That's how low power it was because it was like a 5mAh battery and took a week to get down 10%).

Exactly. It's hard to categorize foods.

UPFs are often stuff you can make in your own home today - take say, french fries. It's hard to see what makes it "ultra processed" when for most home cooks, it's usually just potatoes, with a little batter coating of flour, water, salt, deep fried and then sprinkled with salt or spices. It's hard to say that's bad for you, even though the fries you get at the fast food joint likely is.

Even potato chips can also be made at home with very little processing yet is still a UPF, especially if you like most people buy it in a bag. Pringles, well, you can understand why it's a UPF since it's potatoes, converted into a mash, mixed with other stuff, rolled into a sheet, then sent through a cutter and mold and then fried.

Other stuff is easier to categorize like cakes, cookies and other baked goods, since they don't represent the source foods they came from - you can't imagine a cake from a stalk of wheat. Likewise, we have bread, probably the first ultraprocessed food.

Same goes with dairy products - cheese and yogourt straddle the line between processed and ultraprocessed (especially American cheese, usually referred to as "processed cheese product").

Where you draw the line is hard, and it's basically impossible to say all UPFs are "bad" because that would cut things out like dairy products, or bread products. You might as well toss out trail mix because part of the mix is often stuff like pretzel sticks and such.

It's such a thorny category because we want to categorize UPFs as bad, yet that often categorizes other things as bad. It's probably better to just categorize things like foods with added sugar are bad, or deep fried foods are bad, rather than the source of preparation. Cheese is good (even though it's a processed food), but not too much, because it has a lot of salt in it, as is yogourt (but be careful not to eat too much since many have added sugar).

Teenage me would have been confused. "What is Microsoft thinking, releasing their source code for all to see?" Back then, open source was still kind of a fringe concept.

No it wasn't. Open source was a big reason for the microcomputer revolution in the 70s - everyone was sharing around source code freely. The Apple II even came with full source code for the ROM.

The key points that changed it would be first, BIll Gates' famous "piracy" letter where he called the people freely copying Microsoft Basic as pirates, and second, well, all those Apple II clone lawsuits where they took the Apple II ROM code and used it to create their own Apple II clones.

But the spirit lived on, because as we moved from microcomputers to home computers, people were wanting to learn to program them, and books filled with (often lengthy) BASIC programs were common - because printing out the source code was often the cheapest way to distribute software. In a way, open source lived on because those source code listings in books and magazines that you laboriously typed into your computer were the source code of the program you were going to run.

Distribution on tape and floppies were for the "closed source" programs of the era.

It was really the 90s did the whole era of closed source programs taking over with the popularity of the IBM PC and MS-DOS and everything being shipped as binaries, but the spirit never went away, since Linux was a budding upstart back then.

So much ÃoeTop X thingsà ItÃ(TM)s genuinely infuriating how bad the internet is getting. Used to be able to find answers to questions, tutorials on almost any subject in seconds. Now you have to wade through all the shite to find one that just about answers your question. In a Ãoeget off my lawnà moment I miss the old internet. Could almost trust things you found online or at the very least verify them easily.

Luckily, 95% of the AI slop videos are easily detected - the thumbnail almost always has a woman in some sensual pose (likely AI generated), with the question on a rather bland background. The first second is a text box with the question being asked appearing. That you can tell is an AI generated video and to mark it as "do not show this channel again" because they produce about 1000 videos a day on crap like that. (it's just AI generated video using AI generated text and an AI generated voice with AI generated images, and it goes 24/7).

That eliminates most of the AI slop and often the preview is good enough to identify it.