There are 5 Linux laptop at home. None crash. Except the one my wife uses, when she touches the screen. It just goes dark with a few bright pixels on the top line. Nothing in the logs. Last week I changed the inside screen cable... and the problem disappeared. Must have been some kind of short because the cable is twisted in weird ways in the hinges. So I don't think it was an OS problem !!!

LLMs have never been rules-based "agents," and they never will be. They cannot internalize arbitrary guidelines and abide by them unerringly, nor can they make qualitative decisions about which rule(s) to follow in the face of conflict. The nature of attention windows means that models are actively ignoring context, including "rules", which is why they can't follow them, and conflict resolution requires intelligence, which they do not possess, and which even intelligent beings frequently fail to do effectively. Social "error correction" tools for rule-breaking include learning from mistakes, which agents cannot do, and individualized ostracization/segregation (firing, jail, etc.), which is also not something we can do with LLMs.

So the only way to achieve rule-following behavior is to deterministically enforce limits on what LLMs can do, akin to a firewall. This is not exactly straightforward either, especially if you don't have fine-grained enough controls in the first place. For example, you could deterministically remove the capability of an agent to delete emails, but you couldn't easily scope that restriction to only "work emails," for example. They would need to be categorized appropriately, external to the agent, and the agent's control surface would need to thoroughly limit the ability to delete any email tagged as "work", or to change or remove the "work" tag, and ensure that the "work" tag deny rule takes priority over any other "allow" rules, AND prevent the agent from changing the rules by any means.

Essentially, this is an entirely new threat model, where neither agentic privilege nor agentic trust cleanly map to user privilege or user trust. At the same time, the more time spent fine-tuning rules and controls, the less useful agentic automation becomes. At some point you're doing at least as much work as the agent, if not more, and the whole point of "individualized" agentic behavior inherently means that any given set of fine-tuned rules are not broadly applicable. On top of that, the end result of agentic behavior might even be worse than the outcome of human performance to boot, which means more work for worse results.

More like the total opposite, I'd say.

I can't imagine what's the business value of having ChatGPT doing a BloodNinja impression. It's not good for PR, it'd risk exclusion in serious environments, school and the like, it'd risk legal trouble, the list goes on. The potential for trouble far outweighs any possibly benefit, which is what? There's only downsides because it'd go wrong in some way sooner rather than later.

No, doubling down on serious, well paying uses and removing controversial ideas of little worth is exactly what looks like a clear, decent strategy here.

Ever read The Last Ringbearer ? It's the LOTR, but see from the loser side ("History is written by the winners"). And it shows the orcs as some kind of industrial society (aren't they ?) and the elves as some kind of lordly rulling class (aren't they ?). It twists things around in a way that can give you cold sweats...

Simple, just ask the LLM to fix it, doh ! /s

Yup. Nowadays there are some scam sites that scrape the entire content of others, change the stylesheet, put everything on sale at 50%, let the orders roll in and run away with the cash. Some even manage to stay up for years. One the ripped me off 20$ for a hard to find bike part 2 years ago is still up and running for instance. So good luck to those using AI purchasing agents, they'll need it !

I browse with ad blockers and JS blockers and increasingly more sites just block me outright or won't work ("Content available after you disable your ad blocker" "NO").

Are you inferring that since they let you use a mobile device that they think they're secure?

This is the same group of organizations that will allow you to withdraw cash if you know a four digit code that hasn't been changed in the account holder's entire lifetime. The same group of organizations that will allow you to ACH transfer funds from an account because you know the account number and routing number, which are both on every check they hand out.

BTW, none of the things in the previous paragraph are enough to secure a transfer of funds. Banks always have additional assurances. They cash checks because they don't really cash them, they put them in holding status until they get confirmation from the other side. They allow ACH transfers because they know exactly who is performing them and have a lot of legal leverage to recoup any lost funds. They allow fishy (to a certain extent) credit card transactions because they charge enough interest to eat the losses. They allow teller withdrawals because they have cameras and the FBI will actually go after those who commit fraud.

90% of bank security is invisible to the public. It annoys me when people say "We can do this, my bank does it and they consider it secure." You almost certainly aren't doing what your bank does from a larger perspective.

Particularly since the name is already taken by a revolutionary telescope... that will now take bad images thanks to all those fucking satellites passing in front of it !

Or chloromidians? Was there any mention of chlorine on it?

In the absence of chlorine, would flouromidians form? Would flouromidian-based life have any unusual properties?

Or whatever it is that tricorders detect?

Or souls? Did it make a detour through our solar system to pick up a few souls for later use?

Never would have believed it was a fake.

Let me just add that if you provide a way to copy files via a physical cable, there should be a way to find those files, no ?!?

Maybe locking down the filesystem makes sense on a smarttv (not sure about that one as I don't have one) or a car, but a phone/tablet is now as general a computing device as can be (you still need to copy your pictures to your PC, and copy your mp3 from your PC). I don't know about the Mac, last time I had one was in 2000 and I did not like it back then.

I don't really want to go back to that issue, but your first point makes no sense to me. What is a "device with no direct file access" ? There's a file browser. Apps save files. If the files are already on the device, what sense does it make to transfer them to a server on the other side of the world just so they can make it back. That's total bullshit.

Any time I use anything Apple, I'm violently reminded of why I never want to use anything Apple. Kind of offtopic but last week my dad asked me to copy some directories of images from his Linux PC (which I admin remotely without issues) to his iPad. It took me no less than 2 fucking hours to figure out how. I could connect via USB and transfer easier than I would have expected... but then impossible to find the images on the iPad, no matter what app I used. I tried several other methods to.