While you're correct, and I'm not disputing that, I often don't like an 'unadorned s' for pluralizing proper nouns and non-word things (like acronyms.) I wish there was a way to pluralize while still keeping the base word distinct.

As an example, multiple people named Judy definitely won't be "Judies", but I don't like them being "Judys" either, and definitely not "Judyes". "Judy's" seems more respecting of their actual name. I haven't figured out how to solve the collision with the possessive, and that's the fatal flaw of my argument. [Side note: The editor marked "Judys" and "Judyes" as being spelled incorrectly. Take THAT Merriam-Webster.]

So, while it's purely accidental and muscle memory in my stupid fingers, not lack of knowledge, when I always use "it's" as a possessive, it's a deliberate choice, having full knowledge, when I add extra characters to my pluralized proper nouns. Call it a rebellion.

Mostly because they are just statistical models. They don't understand anything at all. They just know that 'when b is near c, that often means we'll have an x followed by a y' That could be pixels or letters or wave-forms.

With that basis, it's almost miraculous that they do as good of a job as they are at 'pretending' to give coherent answers. That's why I always say, "AI is great, as long as it doesn't have to be correct."

I'd love to have AI that 'understood.' It didn't 'make up' answers, it 'looked up' answers. For example, if you want a URL, instead of 'constructing a likely URL' it actually went out and found that URL.

And it definitely needs to be 'deterministic.' Asking the same question again and getting a different answer is a big 'no no.'

Is this new 'physical model' based on rules and "understanding", or is it just a different pile of statistics?

The 'artificial scarcity' game just got moved up a level.
It used to be the products that were scarce, but when everything went digital, the product scarcity had to be artificial.
For a while, it was the creators who were scarce, (programmers, artists) but now that's not scarce either.

At this point, the only thing that's scarce is energy, water, and RAM.

I can't remember things well enough to know which is which....BUT!

It seems that the 'color' magic of "Warbreaker" would make for some excellent visuals. The first half of the story would make an excellent show as well...(I kind of remember the ending not quite having the same feel though.)

Is the difference between this new idea and a Chromebook just a middle-ware layer where the processing/rendering takes place? So instead of interacting with a browser, I interact with a picture of a browser and the 'computer' that generated that picture interacts with the actual webpage?

Is there a way to 'rearchitect' everything to not need a dedicated 'middle-ware' layer, but instead your 'display terminal' receives feeds of pre-computed and pre-rendered webpages and applications from multiple sources?

Aren't we already in a 'data-center' crisis? We're concentrating the 'compute' too much and it's overloading the grid. Taking the part that's still 'distributed' and centralizing that as well seems like compounding the problem.

This was my thought. What do the 'terminals' to these 'rented computers' look like?

How much 'local processing' do they need to do the communication and display? Wouldn't they still need RAM and hardware of some sort?

I have to assume that Google isn't the authoritative record for these domains. I guess they might be, and in that case the block seems 'somewhat' reasonable. However, if they aren't, and they are just an echo of the authoritative server, then it seems incredibly short-sighted to target Google here.
Is OpenDNS targeted as well, or xFinity, or any of a million other DNS 'repeaters.'

Canal+ is European. I wonder what percentage of European 'Internet users' have Google as their primary DNS server. I'm guessing those [famous IPs] are hosted in the US, and it would server Europe better to point at something more local? (Does Google have other DNS server IPs in other regions? Can you load-balance a single IP such that it doesn't require bouncing off an specific machine? Or is that what DNS would provide?)

Is there an actual JVM here? Looking at what TeaVM does, it seems it's more of a 'cross-compiler' (if that's a term, post says 'transpiles'), and it allows you to write JAVA but compile it to JavaScript (WASM). (Which are famously 'not the same thing.')

So, it's for JAVA programmers who don't want to learn JS? The deliverable seems like a monolithic, minified, partially obfuscated text file?

"Enhance 57 to 19. Track 45 left. Stop!"

Replying to your 'image generation' statement.

(This goes back to my previous post...which the punchline is: "AI is great, when it doesn't have to be correct.")

The mistake is asking it to draw a real person. AI shouldn't be drawing real people anymore than it should be recreating the Mona Lisa, (or writing an existing song, like a story from last week.) It should be generating NEW stuff. Ask it to draw a person that doesn't exist. That way it doesn't need to be correct, and it can do a great job.

The key is 'new' 'non-factual' stuff. For anything else.....use something else.

What I tell all my bosses when they ask me to 'use more AI' is this:

AI is great, when it doesn't have to be correct.

I love Suno. I think it's a miracle. AI making clipart, album covers, poetry? Fine. If it's something you enjoy, go ahead. (I have to say that AI 'copying people' and 'displacing artists' etc are separate problems from what I'm talking about, which is 'accuracy.')

However, if you want a 'fact' then you sure as crap don't want AI. It doesn't know if it's answer is correct or not. It doesn't know what 'correct' means. It just know that it seems similar other things it's seen.

The same goes for AI writing code. Code and algorithms require too much accuracy for me to trust AI. Maybe it's good enough for a 'pre-review' block-out, and saves you some typing. But I'd definitely expect a re-factor afterwards.

Hearing a song is not copyright infringement.

Memorizing lyrics is not copyright infringement.

Reproducing a song as a professional service definitely IS copyright infringement. An interesting question is whether the AI even knew it matched the lyrics. It might have had to do an 'after the fact' search with a 'plagiarism detector' to say, "oops, I guess that matches someone else.)

Blaming the prompts doesn't float, because that's basically saying "He asked me nicely."

---
I think I'd find out if the algorithm was deterministic. If you could find a way to always get it to copy something, that's a problem. If you can 'trick' it into sometimes copying something. Then the tool isn't actually accountable.

It's always been my comment that:

AI is great, as long as it doesn't have to be correct.

If you just want a picture, it can do that. If you want a picture of a specific person, it'll likely be wrong.
If you want it to write a song, it can do that. If you want it to write the sheet music for Beethoven, I wouldn't trust it.

That's the problem I have with AI at work. They keep asking us to integrate it into our Business Intelligence work, but I wouldn't trust it to analyze my business at all. Those answers are supposed to be correct. I'm not going to trust the AI to do it.

I'm a life-long computer guy....but I prefer to keep the computers out of the other appliances.

Finding a 'dumb' TV, or 'dumb' appliances is something you have to commit to. It's not easy, it's a lot of work.

I can't tell if it's because of the potential for additional income from upgrades/options, or whether it's setting the stage for 'recurring revenue' or whether it's just about data-collection and marketing, but they don't like to build 'dumb' appliances anymore.

I figured it might just be that all those features test well with the focus group and that's what they think the entire world wants. ( kinda like how all corporations thinks everyone wants AI in their toothbrush) But I can't imagine why they don't have 'base' models as well that save manufacturing/licensing costs. There's always gotta be the value option....right?

Why would you want to only edit cloud-storage files with your on-disk software?

If you're going to do cloud editing, wouldn't you use the cloud software?

--
(This question might have a better answer)

Why does MS want my files on One-Drive so bad? It costs them money. It's cheaper for them to let me edit local files than it is for them to force me to put them in a free One Drive account?
Is it really just in the hope that I 'might' decide to upgrade...even though I've rejected it over and over and over for the last 10 years?

Is there a clause in the EULA that lets them train their AI on all the files I keep in my free One-Drive account?

Is it just lock-in, and Libre Office can't access those files...and it's an extra step if I want to copy them over to GoogleDrive?

I just don't get it.