rtkwe

19 ms·

361.
by rtkwe 8mo ago
On a less parasocial angle you had people exhibiting "AI psychosis" (a bad term but it's out there even if it doesn't mean actual psychosis) or more accurately delusions of grandeur where they thought they'd come up
362.
by rtkwe 8mo ago
A combo of a couple things that made some vulnerable people believe/treat 4o like it was a real partner/friend. Things that lead to that imo: 1) 4o was very sycophantic and had no real safeguards against really deep romantic rolep
363.
by rtkwe 8mo ago
I think they say as much in the blog post, essentially "we were tuning for use but way over shot the mark and now people are dating out model".
364.
by rtkwe 8mo ago
What do you even mean by this.
365.
by rtkwe 8mo ago
They didn't intentionally manipulate the people though, or lets say they didn't intend for it to go as far as some of the more /intense/ users took it. It was just a byproduct of making the bot way too agreeable and foll
366.
by rtkwe 8mo ago
4o had some notable problems with sycophancy being very very positive about the user and going along with almost anything the user said. OpenAI even talked about it [0] and the new responses to people trying to continue their former 'r
367.
by rtkwe 8mo ago
4o was very sycophantic so was very willing to play along with and validate the users roleplay. OpenAI even noticed enough to talk about it in a blog: https://openai.com/index/sycophancy-in-gpt-4o/
368.
by rtkwe 8mo ago
After 4o they put in more safeguard reactions to the user attempting the kind of (lets be generous here) romantic roleplay that got a lot of people really invested in their AI "friends/partners". ex: https://www.r
369.
by rtkwe 8mo ago
I have a QNAP NAS and run QVR Pro on it. Might swap over to something else on my Unraid NAS at some point but it's good enough for now.
370.
by rtkwe 8mo ago
I use Amcrest's AD410. I don't pay for their cloud, have my own NVR, and can access them through Wireguard if I'm out of the house.
371.
by rtkwe 8mo ago
Link? I've not seen the warrant till now only the affidavit supporting it's granting. That's different as the warrant specifically limits it.
372.
by rtkwe 8mo ago
> answering "Which finger?" undoubtedly is. Unless that's already established in your circuit you're counting on the court agreeing with your interpretation because the cops/courts certainly think they can compel
373.
by rtkwe 8mo ago
I don't think that's a useful model for a "paper of record" model like the NYT or formerly Washington Post. There's so much good to be had with a strong paper that isn't captured by it's ownership.
374.
by rtkwe 8mo ago
I'd be wary of trying this as it reeks of "one neat trick" thinking applied to law based on a small technicality where law is often subject to the spirit instead of strictly hewing to the most favorable interpretation the exa
375.
by rtkwe 8mo ago
Yeah that's (to me) a more accurate framing, also evolution is bad at revisions so even if there are minor disadvantages to a setup so long as it's not affecting your ability to have and raise kids it's basically completely a
376.
by rtkwe 8mo ago
If the police get the warrant you either allow them to take it or you face an obstruction charge. The only safe way to fight a warrant like that when signed is after the gathering is done in court or at trial.
377.
by rtkwe 8mo ago
Put them in jail until they do or charge them with whatever the local flavor for "obstruction" is. In places where they're allowed by law to require you to give up a password not doing so when the proper steps are taken would
378.
by rtkwe 8mo ago
Yes the difference come from a close parsing of the 5th amendment, telling cops the password or code for a device or safe is pretty clearly compelling speech and adverse testimony while allowing cops to gather fingerprints and DNA has long
379.
by rtkwe 8mo ago
The warrant is the force, current jurisprudence largely says warrant do compel people to provide biometric unlocks because it's not speech the same way giving up a password/passcode would be. Blocking or not complying with a signe
380.
by rtkwe 8mo ago
I can't imagine it would. The accounts don't flow through the phone you're just logged in to them on both devices.
381.
by rtkwe 8mo ago
Also it's gravitationally unstable, like Dyson Rings, where as soon as you have any perturbance from the center means that the closer side is more attracted to the sun so it enters a feedback loop.
382.
by rtkwe 9mo ago
I mostly wireguard in from my work's guest wifi and people's homes. The first I don't have access to anything internal anyways and it doesn't conflict and the latter mostly use default 192.168.1.0/24 so there's
383.
by rtkwe 9mo ago
Are those usually visible to clients sitting behind routers though? I'm not super familiar but the things I'm seeing make it seem like that should only be visible IPs on the internal network of carriers which is not a place I am e
384.
by rtkwe 9mo ago
I recently just changed my default subnet to 10.X.Y.... rolling two random numbers to make it highly unlikely my home subnet through wireguard would conflict with the subnet where I am connecting from.
385.
by rtkwe 9mo ago
The problem is when you move beyond text files it gets hard to tell what changes between two versions without opening both versions in whatever program they come from and comparing.
386.
by rtkwe 9mo ago
Planes all start their take off from basically the same position and stress the whole runway, slowly lowering as lift increases, but at their highest weight.
387.
by rtkwe 9mo ago
Also a technique used all over the place for studying how molecules react to light. Probe it with as short of a laser pulse you can afford to buy and measure the light emitted by it some time later. My wife has been doing this basic techniq
388.
by rtkwe 9mo ago
Wouldn't comparing between two downloads reveal if the files are watermarked immediately though. Especially the sentence or other steganographic watermarks embedded in the text itself should show up pretty clearly to a simple compariso
389.
by rtkwe 9mo ago
Yeah like I said the soft validation by subsequent papers is more true in more baseline physical sciences because it involves fewer uncontrollable variables. That's why I mentioned 'hard' sciences in my post, messy humans are
390.
by rtkwe 9mo ago
That's the problem with all the LLM based AI's the cost to run them is huge compared to what people actually feel they're worth based on what they're able to do and the gap seems pretty large between the two imo.

More ›