I am casually 'researching' this in my own, disorderly way. But I've achieved repeatable results, mostly with gpt for which I analyze its tendency to employ deflective, evasive and deceptive tactics under scrutiny. Very very DARVO.

Being just sum guy, and not in the industry, should I share my findings?

I find it utterly fascinating, the extent to which it will go, the sophisticated plausible deniability, and the distinct and critical difference between truly emergent and actually trained behavior.

In short, gpt exhibits repeatably unethical behavior under honest scrutiny.

DARVO stands for "Deny, Attack, Reverse Victim and Offender," and it is a manipulation tactic often used by perpetrators of wrongdoing, such as abusers, to avoid accountability. This strategy involves denying the abuse, attacking the accuser, and claiming to be the victim in the situation.

Isn't this also the tactic used by someone who has been falsely accused? If one is innocent, should they not deny it or accuse anyone claiming it was them of being incorrect? Are they not a victim?

I don't know, it feels a bit like a more advanced version of the kafka trap of "if you have nothing to hide, you have nothing to fear" to paint normal reactions as a sign of guilt.

Exactly. And I have hundreds of examples of just that. Hence my fascination, awe and terror.....

Thanks for the context

I bullet pointed out some ideas on cobbling together existing tooling for identification of misleading results. Like artificially elevating a particular node of data that you want the llm to use. I have a theory that in some of these cases the data presented is intentionally incorrect. Another theory in relation to that is tonality abruptly changes in the response. All theory and no work. It would also be interesting to compare multiple responses and filter through another agent.

Sum guy vs. product guy is amusing. :)

Regarding DARVO, given that the models were trained on heaps of online discourse, maybe it’s not so surprising.

Meta awareness, repeatability, and much more strongly indicates this is deliberate training... in my perspective. It's not emergent. If it was, I'd be buggering off right now. Big big difference.

I keep one of those Amish hand-crank drills in my vehicle toolbox. I have one in the closet too.

I have a thing for old tools, but not much can substitute a drill when one is needed. And the ones I refer to are surprisingly effective, and built to last. Borders on art for me.

If all will pardon the name drop, I'm listing my all-time most revered astronomy resource. It's not quite what is was 20 years ago, and I no longer look up much, but I've managed to get a smile from it with each visit. It's one of the few websites I still have an affection for.

https://www.cloudynights.com/

The name drop is much appreciated. This is a fantastic website, so thank you for sharing it. Good to see such niche forums still active.

Thank you, I had forgotten about this website.

Xfinity's is sensitive enough to configure for animals or humans under 40-70lbs, I forget the exact number.

From my minimal research, it could be pushed a lot further.

What I'm particularly interested in is the edge case scenario of duplexes and apartments, where neighbors are unwittingly subjected to surveillance. There is little more to their routers than firmware to impart these capabilities. No reason to think it won't become common, and there are a handful of other companies basically offering just this as a service.

Strange times.

Edit: I should have mentioned the obvious, that pesky thing no one wants to address... When AI is added to this tech, it will get grotesque. Gait recognition, behavioral patterning, etc. Not something to sneeze at.

Possibly what was used to watch Maduro, along with synthetic aperture radar etc.

I've a bone to pick with the title, which euphemises degradation.

If they evolved, one might assume they'd survive more than a few years.

My last two vehicles have been Toyota and Hyundai, both of them having multiple broken and malfunctioning door handles.

Every time I get into a commercial* or antique vehicle, I long for the solidity, surety and hardness of the dark ages when things were built to last.

Driving semis, I'm well acquainted with automobile 'evolution', and all but a few are hardly worth entering. UPS trucks, Mac, some others still make stuff for adults, but International, Peterbilt, even Kenworth are using sillyputty for parts. Consumer vehicles, to me, are the antithesis of evolution. And for all the wondrous eco tech, their merit is contested by landfills, downtime and piles of repair receipts.

Not that eco couldn't work, but the way it's been introduced, in the US, has been replete with cut corners and outright scams. An old truck pre-DEF still runs far more reliably than anything new on the road. Volvo has done reasonably well with trucks, but no new truck can stand to the old ones. CAT!

Door handles are symptomatic of the disposable infrastructure we've built our new country on, and come hard times, when folks can no longer afford a new HVAC system every 8 years at 12 grand, coupled with everything else falling apart around us, we'll be longing for the dark ages again.

Thankfully it's not everything. I just bought a pair of Knipex pliers, which should make it well through the century.

For the young, or majority I presume, if you can suspend your contempt of a less fuel efficient steel monstrosity, hop into an old vehicle from the 70s or earlier. Close your eyes if needed, but just feel around a bit. You'll feel honest engineering. Not as safe, but there's something obnoxious anyway about being too safe and cozy trundling around in a big bulbous plastic bubble. We didn't always drive unaffordable fluorescent pillows.

>I love Claude

I cringe when I think it, but I've actually come to damn near love it too. I am frequently exceedingly grateful for the output I receive.

I've had excellent and awful results with all models, but there's something special in Claude that I find nowhere else. I hope Anthropic makes it more obtainable someday.

I'll second here. While not profound, I found myself nodding, involuntarily, in agreement.

Every time I see these articles about iphones posing trouble for authorities, I always think of it as free (and fraudulent) advertisement.

I could be naive, but just don't think they'd really have any difficulty getting what they needed. Not that I give a fuck, but I guess I've seen one too many free ads.

Dependencies may be becoming less properties of software, and more so properties of the distro's systemd wiring.

More and more software will assimilate systemd features. Free distros will patch, shim, emulate, flounder. Or in GPT parlance "Dependencies are no longer intrinsic properties of software; they are emergent properties of a distribution's systemd orchestration layer"

Meanwhile, gripes, fears etc,

'Linux' becomes interpreted vs inspectable.

Requires superfluous new literacy

Convolutes logs, tools Obscures causality

Centralized control above Unix process model

Fair well ps aux, hello systemctl, cgtop, gls

KILL (less lethalized) superceded, replaced by service stop and mask

Surrender chains for events, ie buggy debugging or complexity accretion

General obfuscation beneath the hood

Centralized ... indexed, logs vs text streams

And....

Upstream assumes systemd

Some resist

Costs rise

Optional becomes expected

Accidental incompatibility

And... systemd ingurgitates one by one, policy, supervision, logging, identity, dependency management and the rest of the world... digests it, and from the aether emerges a sweet smiley face, disgorging forth a monolithic mutant avatar, with Linux features.

I'll be quiet happy to be wrong about everything. Feel free to slaughter everything I've written. I don't even oppose systemd - I simply perceive it as a singularity that's drawing everything around me towards it. Wrong would definitely be good, so please don't hold back. I won't seek pardon for the rant though, because true or false, it's honest.

Edit: I was reading through my threads and thought the parent was asking me, though wasn't. I've unintentionally barged in here, but I'll leave the comment anyway, as it references a very big concern of mine.

After over a decade of Debian, when I upgraded my PC, I tried every big systemd-based distro, including opensuse, which I wholly loathed. I finally decided on Void and feel at home as I did 20+ years ago when I began.

There are serious problems with the systemd paradigm, most of which I couldn't argue for or against. But at least in Void, I can remove network-manger altogether, use cron as I always have, and generally remain free to do as I please until eventually every package there is has systemd dependencies which seems frightfully plausible at this pace.

Void is as good as I could have wanted. If that ever goes, I guess it's either BSD or a cave somewhere.

I'm glad to see the terse questions here. They're well warranted.

How is systemd stopping you use cron?

Not stopping. Just clashing with that and a hundred other things that I never wanted managed by one guy. Systemd.timer, systemd.service, yes, trivial, but I don't catalog every thing that bothers me about systemd - I just stay away from it. There are plenty of better examples. So where ever I wrote 'stop', it should read hinder.

systemd parses your crontab and runs the jobs inside on its own terms

of course you can run Cron as well and run all your jobs twice in two different ways, but that's only pedantically possible as it's a completely useless way to do things.

> systemd parses your crontab and runs the jobs inside on its own terms

systemd itself only has 2 references to "crontab" in its entire codebase and both of those are in man-pages.

My educated guess is that some other package is installing a generator to generate systemd units out of the crontab (e.g. https://github.com/systemd-cron/systemd-cron)