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.

Atlantic staff writer Jake Lundberg, who shops at the Granger, Indiana location, describes the stores as spaces of "cooperation, courtesy, and grown-ups mostly acting like grown-ups." Shoppers follow unwritten rules: move along, don't block the way, step aside to check your phone.

Maybe in Granger, Indiana. Every one I've visited in a major metropolitan area (whether coastal or heartland) has been a cacophony of chaos with, at best, oblivious shoppers looking at everything except where they're going and who is around them, and at worst people who actively jockey for position, rushing to pass one another, only to stop short and block the person they just passed. It's insanity, and I avoid going there as much as possible except to take someone else.

LLMs cannot reason. None of them are "smart."

Not all FPGAs are that expensive... I work a lot with AMD stuff (they bough Xilinx in 2022), and you can get a decent SOM for $300 or less. Not "dirt cheap", but not expensive. They have FPGA chips well under $100, including ARM CPU cores, or a soft RISC processor.

It's a freaking 16-bit Intellivision running at 895kHz (not even mHz, let alone gHz!) ... any latency from the wireless devices will be _dwarfed_ by latency in the processor (or emulated processor). I'm sure the processor in the re-issue we're talking about here is much much faster, but it's probably spending 99% of its time in sleep cycles to dumb itself down to the original console's timing... meaning it can handle the wireless controllers just fine.

Now for modern-day gaming, yeah, I won't argue. But for this context, lol...

MTBF is useful information, but I think it would be more useful in conjunction with factors like active spinning time, total spin up/down counts, cumulative head seek time, total IO, etc. Presumably time-in-service affects the MTBF more than the age of the drive, but to what extent? Is a NIB drive that's two years old going to be as reliable as one that's only a month or two old? So many variables....

The internet is really the best medium for sarcasm.

Pretty sure this opens him up to a legal malpractice suit. Probably more lucrative than whatever the debt was.

lasting ecological shifts will hinge on design and long-term care.

We don't really know that for sure. It may improve the odds, but neither desertification nor greening require human intervention, nor is human intervention necessarily going to achieve the desired outcome. Life, uh... finds a way. (Except when it doesn't.) But for all we know (and what seems most likely absent evidence to the contrary), this is just a temporary oasis of sorts that will last only as long as the structures on the site.

The article is sparse on details. I don't necessarily think driverless cars should be given a free pass -- in fact, we should probably have higher fines for the manufacturers -- but 9 times out of 10 when a road is blocked, it's because of construction or an accident, not a checkpoint. I suspect it was reacting to the obstruction, because when a road is obstructed, the "no U-Turn" rule generally doesn't apply (or isn't enforced anyway). In fact, if it hadn't been a checkpoint, I doubt they would have even been looking for illegal U-Turns, which are indicative of people trying to avoid the checkpoint, presumably.

As for fines, I do think they should be higher for self-driving cars, because $300 isn't even a slap on the wrist for Google. On the other hand, that could create a perverse incentive where officers are ignoring flagrant violations by human drivers in favor of issuing a $100k ticket to a Waymo that veered out of its lane to avoid a hazard. It could also create a situation where self-driving cars are so cautious that traffic is snarled by puritanical robot cars that won't even approach the speed limit because it's not worth the risk.

It was an interesting experience, just out of curiosity. A few highlights:

- it seemed to max out at about 20mph on my ride, but that seems to be so that it could time the lights well and not have to stop/slow down

- a car backed out of a parking lot and directly in front of us. The Waymo stopped, honked, and when the car kept backing up it quickly reversed. A human driver would have been hit (seriously)

- but, I also seem to think that the driver backing up did so because they knew they could "bully" the Waymo and it would get out of their way.

It was easy to get one from the hotel to a cafe outside the city center, they were everywhere. We counted over 30 going by a restaurant while we had a quick meal. But getting one to come pick us up outside the city center would have taken 25 minutes, so we Ubered back instead.

"Interesting" is the only description of the experience. I can see the appeal of not dealing with people in the process, but I'm an introvert.

Sorry for your loss. Dating apps are indeed garbage. If I were single, I'd be talking to every attractive person I saw at a grocery store, museum, out walking, etc. I'm an introvert and it makes me nervous AF, but I've also realized that pretty much anyone who agrees to meet for a drink is already interested, so that makes it easier. I mean I hate doing job interviews too, but it's just part of the process, not the end of the world.

I deeply apologize that Nancy Mace has anything to do with us. We're working on fixing that at mid-terms...

space-based would be closer to 24/7 sunlight (depending on orbit), would suffer less from atmospheric absorption (presumably, the light would always be coming directly "down" on the target receiver - not obliquely through the atmosphere), etc.