AND. And build better sites. That's my opinion.

The pathways for these intrusions are going to be found and exploited. It'll be by greedy blackhats, nation-state blackhats, and maladjusted domestic blackhats but it's going to happen.

The key difference is that domestic actors - be they individuals or companies - can and should be prosecuted. It doesn't matter if it's a script kiddie using non-AI tools or some frontier model-training corporate entity... accessing computer systems you don't have the rights to is illegal.

As for the "safety" of these tools, it seems to me that they're being developed with a default-allow-plus-blacklist model of permission instead of default-deny-plus-whitelist model. Someone will find ways to "trick" them into working around blacklists of behavior and access. They should be scoped the opposite way so the only behaviors and targets they can use is limited. I totally get it that reduces their utility dramatically but... tough.

The problem is they're not currently good at not pattern-matching.

That's debatable.

Part of why I read is plot/story/character, which generally survives any edit process. The other part of why I read is creative word-choices and sentence structures. Authors like Douglas Adams made deliberately convoluted structures that amused. The style of an author is - to me - an important thing.

Can LLM editors today accept and accommodate the voice of an author? No.
Will they ever? Almost definitely.

Where I'm at is that these tools need to be better-than-human before they should be deployed on creative material. This isn't just spell-checking, it's editing, which does involve understanding of the material and what an author is doing and why, and accepting that the results don't always fit the norms and shouldn't. Creative material is about pushing boundaries into new frontiers, and while LLM will probably get there, it's not there today and I don't want it blending my favorite authors' into some beige hellscape.

Nobody's going to read it, expert or not. It's a wall of... whatever. It's not written or formatted in a way that is enticing to a reader.

Remember... LLMs regurgitate materials that already exists. People who are familiar with the topic likely already know anything accurate or factual in it. People who aren't familiar with the topic could go somewhere reliable to ingest the facts. This unsolicited pile of words... serves nobody.

The key word is "it." There is no singular "it". Flattening the category of AI-enabled software to a singular "it" implies that there is a flat truth that applies to that entire category.

The AI-enabled software "it" that is running a store into the ground doesn't present a meaningful threat to anyone.
the AI-enabled software "it" that is being used by black-hats to find security vulnerabilities does.

AI is a tool. Think of it not unlike nuclear. Nuclear power plants aren't an existential threat, despite the harm a plant accident can cause. Nuclear weapons are.

What's clear is that trying to simplify complex topics into bite-sized "common-sense" topics doesn't work in 2026.

No, I don't think the tech-bros are being honest. But I don't think it matters anymore. Someone is going to weaponize this technology and a lot of people are going to suffer because of it. Not Skynet scenarios but rather Cyberpunk / Blade Runner scenarios. There's a real danger to the health and well-being of a large portion of the human race, for sure. The tech-bros aren't going to let that stop them from getting richer, for sure.

What we - the 99% - can or should do about that, I don't know.

Fair enough, and the feedback is always appreciated. All I can offer in return is to bold what I wrote...
"On the gripping hand, there could be individual-level cases that do have society-wide impact."

Thank you.

Wow. Brutal.

I was very careful in my choice of words. This isn't about understanding. It's about feeling, which is why I used the word "feel" instead of "think" repeatedly.

I understand the role of the US Supreme Court and its equivalents in other countries. It's just that I'm uncomfortable with it being triggered in this case. I'm sure that doing so will be technically correct if allowed, but I don't feel that this case is a witchhunt... it's a witch. Yes, that means I've convicted Bankman-Fried in my heart. Granted. The courts to date have done the same, so it's not like the conclusion of guilt is absurd.

I have mixed feelings on this.

On the one hand, someone who believes they have been wrongfully convicted of any crime should be allowed to have that conviction reviewed by every level of justice that exists.

On the other hand, I feel something like a supreme court should spend its time on law matters that impact a large portion of their country. Issues and topics that refer to policy and fundamental law. Spending their time examining a single person's case feels like a use of their time that is unjust to the rest of the country.

On the gripping hand, there could be individual-level cases that do have society-wide impact.

In the end, this is only happening because there's a lot of money involved, which also feels unjust. A random citizen wrongfully convicted of theft or fraud could never get their case before the US Supreme Court, so why should this guy? I hate it all.

I have no mod points, so here, have an "I agree". On all points. It's one thing to say "I didn't like it" or - even better - "I didn't like it because...", but dropping random "it sucked" - or worse - "it really sucked" is... lame.

I found it entertaining, which is what my primary criterion is for entertainment.

There's (probably) a fundamental difference, even if you want to equate AI recombination and human inspiration.

Human artists - in theory - used legal means to hear what inspired them. Could be listening to it on broadcast radio, where the right to hear it was arranged through ads. Could be listening to it on a mix tape from a friend, which many countries have laws for. It's not assumed, suspected, or proven that the musician who cites Sting or Ozzy stole the albums they're inspired by.

The tech bros pretty openly did exactly that.

If they ever could've, they would've. That should tell you they couldn't, and the above is the reason, I think.

Almost without fail the people doing that are working within legal rights, doing parody. Comedians or sketch comics. There has been a legal carve-out for those types of activities. Even Elvis impersonators have to be very careful; their actual performance doesn't invoke legal liability but any merch or an Elvis-themed Vegas wedding facility does trigger royalties. It's very different to have a comedian performing "wouldn't it be funny if Famous Person was in this situation" than "we're making more Die Hard movies and Bruce Willis can go ahead and stop us from using his likeness if he can... oh... right... he can't... so eat a bag of dicks."

It's August 24th right now. "February or March" is around 5 months away. Which is not a year.

Due to weather, yes, plausibly. This is literally rocket science and weather can absolutely merit a scrub.

Again, this is rocket science. The rocket scientists seem to think that visible light is somehow helpful to the landing's odds of success. It passes the sniff test to me; voluntarily sacrificing such a useful spectrum seems unwise to me.

Indeed, yes. Do you have some grounds for your doubt?

That's interesting but drum machines and synths are still instruments in a human's hands. Think about some of Pink Floyd's early works where they used tape loops and whatnot. The automation became an instrument to incredibly ground-breaking effect. Think about the entirety of Depeche Mode's discography. Huge amounts of artificial instruments but the arrangement can be very good, and they have made new arrangements throughout the decades when touring, just like purely accoustic bands. Then there's Michael Cretu's band, Enigma, which is literally a portable sound studio, yet his stuff is quite influential and complex.

The difference in my mind is that AI isn't any of those things. It's recombinant, not creative. It will never invent... until true AI in the sense we used to use the term exists.

Task Manager is for application performance. It tells you what resources a given task/PID/application is using. It tells you what memory usage is, CPU usage is, GPU usage is, per application. Extending this to NPU hardware is very much the purview of Task Manager.

While that is true, it is also not what this is about.

They're expanding Task Manager to give you insight and monitoring capabilities for when AI is running. They're not injecting AI into Task Manager itself.

I never demanded perfection.

Worse, your qualification is meaningless. A 100% failure rate would be good enough provided additional measures that compensate for it.

This thing screws up one in every 5,000 visitors. Look up Sainsbury's. Three million customers per day. That's 600 failures per day. With around 1,500 locations, that's an average failure at every location every two to three days. Maybe that's happening when your best manager is on-hand to take a look and conclude "huh, well, this looks a lot like the known shoplifter but... well... maybe not". Or maybe that's happening in the middle of the night when your minimum-wage student decides to not serve a lawyer, doctor, or garbageman who deserves it.

A screwup once a year per location would be high. Once or twice a year for the entire chain... maybe. Maybe acceptable. But you go ahead and set your personal bar wherever you want. Until you get the cops called on you.