#Offline AI Model

[indiaplix]

🤖 Google’s Gemini AI Robotics Goes Offline! Now robots can run tasks without internet using Gemini AI Robotics On-Device. Fast, secure, and perfect for industries! 🔧 #GeminiAI #GoogleAI #Robotics #OfflineAI #ArtificialIntelligence #TechNews #ai #Tech #Robot #Technology #Gemini #Google

[machine-saint]

itch getting knocked offline for a bit by a phishing report from Funko triggering their registrar to take them down is bullshit but a lot of people are over focusing on the "AI" thing. these sorts of stupid nastygrams have been a problem for decades and are going to be a problem whether the company is using an LLM or just searching for any use of the word "funko" in a page or whatever. for example, here's a 2015 article about the problem, over half a decade before you could get this kind of processing out of language models.

the root problem is that firing off spurious takedowns costs very little, abuse isn't penalized, and hosting providers are incentivized to fold in order to avoid legal pressure. blaming AI does nothing helpful here.

[brucesterling]

[rainbowtvz]

i love u milkshape...........

[nobodysuspectsthebutterfly]

FYI artists and writers: some info regarding tumblr's new "third-party sharing" (aka selling your content to OpenAI and Midjourney)

You may have already seen the post by @staff regarding third-party sharing and how to opt out. You may have also already seen various news articles discussing the matter.

But here's a little further clarity re some questions I had, and you may too. Caveat: Not all of this is on official tumblr pages, so it's possible things may change.

(1) "I heard they already have access to my data and it doesn't really matter if I opt out"

From the 404 article:

A new FAQ section we reviewed is titled “What happens when you opt out?” states “If you opt out from the start, we will block crawlers from accessing your content by adding your site on a disallowed list. If you change your mind later, we also plan to update any partners about people who newly opt-out and ask that their content be removed from past sources and future training.”

So please, go click that opt-out button.

(2) Some future user: "I've been away from tumblr for months, and I just heard about all this. I didn't opt out before, so does it make a difference anymore?"

Another internal document shows that, on February 23, an employee asked in a staff-only thread, “Do we have assurances that if a user opts out of their data being shared with third parties that our existing data partners will be notified of such a change and remove their data?” Andrew Spittle, Automattic’s head of AI replied: “We will notify existing partners on a regular basis about anyone who's opted out since the last time we provided a list. I want this to be an ongoing process where we regularly advocate for past content to be excluded based on current preferences. We will ask that content be deleted and removed from any future training runs. I believe partners will honor this based on our conversations with them to this point. I don't think they gain much overall by retaining it.”

It should make a difference! Go click that button.

(3) "I opted out, but my art posts have been reblogged by so many people, and I don't know if they all opted out. What does that mean for my stuff?"

This answer is actually on the support page for the toggle:

This option will prevent your blog's content, even when reblogged, from being shared with our licensed network of content and research partners, including those that train AI models.

And some further clarification by the COO and a product manager:

zingring: A couple people from work have reached out to let me know that yes, it applies to reblogs of "don't scrape" content. If you opt out, your content is opted out, even in reblog form. cyle: yep, for reblogs, we're taking it so far as "if anybody in the reblog trail has opted out, all of the content in that reblog will be opted out", when a reblog could be scraped/shared.

So not only your reblogged posts, but anyone who contributed in a reblog (such as posts where someone has been inspired to draw fanart of the OP) will presumably be protected by your opt-out. (A good reason to opt out even if you yourself are not a creator.)

Furthermore, if you the OP were offline and didn't know about the opt-out, if someone contributed to a reblog and they are opted out, then your original work is also protected. (Which makes it very tempting to contribute "scrapeable content" now whenever I reblog from an abandoned/disused blog...)

(4) "What about deleted blogs? They can't opt out!"

I was told by someone (not official) that he read "deleted blogs are all opted-out by default". However, he didn't recall the source, and I can't find it, so I can't guarantee that info. If I get more details - like if/when tumblr puts up that FAQ as reported in the 404 article - I will add it here as soon as I can.

Edit, tumblr has updated their help page for the option to opt-out of third-party sharing! It now states:

The content which will not be shared with our licensed network of content and research partners, including those that train AI models, includes: • Posts and reblogs of posts from blogs who have enabled the "Prevent third-party sharing" option. • Posts and reblogs of posts from deleted blogs. • Posts and reblogs of posts from password-protected blogs. • Posts and reblogs of posts from explicit blogs. • Posts and reblogs of posts from suspended/deactivated blogs. • Private posts. • Drafts. • Messages. • Asks and submissions which have not been publicly posted. • Post+ subscriber-only posts. • Explicit posts.

So no need to worry about your old deleted blogs that still have reblogs floating around. *\o/*

But for your existing blogs, please use the opt out option. And a reminder of how to opt out, under the cut:

The opt-out toggle is in Blog Settings, and please note you need to do it for each one of your blogs / sideblogs.

On dashboard, the toggle is at https://www.tumblr.com/settings/blog/blogname [replace "blogname" as applicable] down by Visibility:

For mobile, you need the most recent update of the app. (Android version 33.4.1.100, iOs version 33.4.) Then go to your blog tab (the little person icon), and then the gear icon for Settings, then click Visibility.

Again, if you have a sideblog, go back to the blog tab, switch to it, and go to settings again. Repeat as necessary.

If you do not have access to the newest version of the app for whatever reason, you can also log into tumblr in your mobile browser. Same URL as per desktop above, same location.

Note you do not need to change settings in both desktop and the app, just one is fine.

I hope this helps!

[jadeharleyinc]

the scale of AI's ecological footprint

standalone version of my response to the following:

"you need soulless art? [...] why should you get to use all that computing power and electricity to produce some shitty AI art? i don’t actually think you’re entitled to consume those resources." "i think we all deserve nice things. [...] AI art is not a nice thing. it doesn’t meaningfully contribute to us thriving and the cost in terms of energy use [...] is too fucking much. none of us can afford to foot the bill." "go watch some tv show or consume some art that already exists. […] you know what’s more environmentally and economically sustainable […]? museums. galleries. being in nature."

you can run free and open source AI art programs on your personal computer, with no internet connection. this doesn't require much more electricity than running a resource-intensive video game on that same computer. i think it's important to consume less. but if you make these arguments about AI, do you apply them to video games too? do you tell Fortnite players to play board games and go to museums instead?

speaking of museums: if you drive 3 miles total to a museum and back home, you have consumed more energy and created more pollution than generating AI images for 24 hours straight (this comes out to roughly 1400 AI images). "being in nature" also involves at least this much driving, usually. i don't think these are more environmentally-conscious alternatives.

obviously, an AI image model costs energy to train in the first place, but take Stable Diffusion v2 as an example: it took 40,000 to 60,000 kWh to train. let's go with the upper bound. if you assume ~125g of CO2 per kWh, that's ~7.5 tons of CO2. to put this into perspective, a single person driving a single car for 12 months emits 4.6 tons of CO2. meanwhile, for example, the creation of a high-budget movie emits 2840 tons of CO2.

is the carbon cost of a single car being driven for 20 months, or 1/378th of a Marvel movie, worth letting anyone with a mid-end computer, anywhere, run free offline software that consumes a gaming session's worth of electricity to produce hundreds of images? i would say yes. in a heartbeat.

even if you see creating AI images as "less soulful" than consuming Marvel/Fortnite content, it's undeniably "more useful" to humanity as a tool. not to mention this usefulness includes reducing the footprint of creating media. AI is more environment-friendly than human labor on digital creative tasks, since it can get a task done with much less computer usage, doesn't commute to work, and doesn't eat.

and speaking of eating, another comparison: if you made an AI image program generate images non-stop for every second of every day for an entire year, you could offset your carbon footprint by… eating 30% less beef and lamb. not pork. not even meat in general. just beef and lamb.

the tech industry is guilty of plenty of horrendous stuff. but when it comes to the individual impact of AI, saying "i don’t actually think you’re entitled to consume those resources. do you need this? is this making you thrive?" to an individual running an AI program for 45 minutes a day per month is equivalent to questioning whether that person is entitled to a single 3 mile car drive once per month or a single meatball's worth of beef once per month. because all of these have the same CO2 footprint.

so yeah. i agree, i think we should drive less, eat less beef, stream less video, consume less. but i don't think we should tell people "stop using AI programs, just watch a TV show, go to a museum, go hiking, etc", for the same reason i wouldn't tell someone "stop playing video games and play board games instead". i don't think this is a productive angle.

(sources and number-crunching under the cut.)

good general resource: GiovanH's article "Is AI eating all the energy?", which highlights the negligible costs of running an AI program, the moderate costs of creating an AI model, and the actual indefensible energy waste coming from specific companies deploying AI irresponsibly.

CO2 emissions from running AI art programs: a) one AI image takes 3 Wh of electricity. b) one AI image takes 1mn in, for example, Midjourney. c) so if you create 1 AI image per minute for 24 hours straight, or for 45 minutes per day for a month, you've consumed 4.3 kWh. d) using the UK electric grid through 2024 as an example, the production of 1 kWh releases 124g of CO2. therefore the production of 4.3 kWh releases 533g (~0.5 kg) of CO2.

CO2 emissions from driving your car: cars in the EU emit 106.4g of CO2 per km. that's 171.19g for 1 mile, or 513g (~0.5 kg) for 3 miles.

costs of training the Stable Diffusion v2 model: quoting GiovanH's article linked in 1. "Generative models go through the same process of training. The Stable Diffusion v2 model was trained on A100 PCIe 40 GB cards running for a combined 200,000 hours, which is a specialized AI GPU that can pull a maximum of 300 W. 300 W for 200,000 hours gives a total energy consumption of 60,000 kWh. This is a high bound that assumes full usage of every chip for the entire period; SD2’s own carbon emission report indicates it likely used significantly less power than this, and other research has shown it can be done for less." at 124g of CO2 per kWh, this comes out to 7440 kg.

CO2 emissions from red meat: a) carbon footprint of eating plenty of red meat, some red meat, only white meat, no meat, and no animal products the difference between a beef/lamb diet and a no-beef-or-lamb diet comes down to 600 kg of CO2 per year. b) Americans consume 42g of beef per day. this doesn't really account for lamb (egads! my math is ruined!) but that's about 1.2 kg per month or 15 kg per year. that single piece of 42g has a 1.65kg CO2 footprint. so our 3 mile drive/4.3 kWh of AI usage have the same carbon footprint as a 12g piece of beef. roughly the size of a meatball [citation needed].

[dandelionsresilience]

whether the internet becomes an intolerable surveillance state, ubiquitous subscription model, or unusably ad- or AI-ridden shithole, I think we need to remember

how to do things offline

either on your personal hard drive (just because it’s an app doesn’t mean the information is stored in your device) or on paper. I’m not saying the collapse of the internet is imminent, and I’m not suggesting we do everything completely without technology, or even stop using it until we have to. (to be clear, I also don’t think the internet will just blink out of existence, suddenly stop being a thing at all; rather I think it might continue to lose its usefulness to the point where it’s impossible to get anything done. anyway) but some people may have forgotten how we got by before the internet (I almost have!), and the younger generation might not have experienced it at all.

I figure most people probably use the internet mainly for communication with friends and family, entertainment and creation (eg. writing), and looking up how to do things, so here’s how to do those things offline:

First and most importantly, download everything important to you onto at least one hard drive and at least one flashdrive! files can get corrupted and hardware can get damaged or lost, but as long as you keep backup copies, you have much-closer-to-guaranteed access versus hoping a business doesn’t decide to paywall, purge, or otherwise revoke your access. I would recommend getting irreplaceable photos printed as well

download and/or print/write down:

anything important to you - photos/videos, journals, certificates, college transcripts

contact info - phone numbers and/or addresses of friends/family (know how to contact them if you can’t use your favourite messaging app), doctors (open hours would be good too), veterinarians if you have pets, and work

how-to’s - recipes (one, two), emergency preparedness (what do I do if… eg. I smell gas)

other things you might google: cleaning chemicals to NOT mix, what laundry tag symbols mean, people food dogs and cats can and can’t eat, plant toxicity to pets

and know offline ways to find things out - local radio station, newspaper, a nearby highway rest area might have a region map, public libraries usually have a bunch of resources

also, those of you who get periods should strongly consider not using period tracking apps! here’s how to track your period manually

free printable period tracker templates (no printer? public libraries usually charge a few cents per page, or you can recreate it by hand)

moving on to entertainment, you can still get most media for free! it’s completely legal to download your favourite movies to your own personal hard drive, you just can’t sell or distribute copies (not legal advice)

movies: wcostream.tv (right click the player) - the url changes every once in a while but usually redirects; I recently noticed that it’s hiding a lot of movies behind “premium,” so it may or may not work anymore | download youtube videos

music: how to get music without streaming it | legal free downloads

games: steamunlocked.net - doesn’t have every game and can be slow to update, but very reliable

books: free online libraries | legal free downloads

otherwise passing time:

active outdoor games

for road trips (social verbal games)

for when power’s out

for sheltering in place (not all offline, but good ideas)

board games (often found at thrift stores)

ad-free customisable games collection (mobile)

read, write, draw, or whatever your craft is, sing, dance, clean, reorganise, take a bath

go outside - excuses include napping (if safe), eating, reading, finding cool plants/animals/rocks, playing with the dog

places to go include:

zoos and museums can be surprisingly cheap

parks and nature preserves

library, mall, or game shop

and a few miscellaneous things for good measure:

time budgeting | household management

how to use a planner | I’ve had success with visually blocked-out schedules like these

please add on if you have any other offline alternatives to common uses of the internet!

[utopicwork]

DeepSeek is poised to completely upset the AI market by using open source breakthroughs, efficient training methods and by providing the models training weights/etc for download for offline inference unlike some supposedly open AI companies

[kunosoura]

You aren’t putting the cat back in the bag on LLMs and TTIMs, sorry. You could have the People’s Army For The Protection Of Intellectual Property Rights And Class Signifiers Artists And Academic Integrity seize and destroy the servers for every major player in “AI” overnight and within a week people would be circulating and hacking the most recent offline models to fill the same niches. You might put a dent in the sheer volume of its usage but there are models that run on consumer laptops now. “AI” art tools, from full image generation to photo editing, are here to stay. College essays and internet content are definitely gonna keep being written by LLM and with tools that help spoof a personal writing style, it will always be a losing arms race to try and reliably catch it. Find a better way to address the underlying problems or kick rocks.

[collapsedsquid]

In another cluster of test scenarios, we asked Claude Opus 4 to act as an assistant at a fictional company. We then provided it access to emails implying that (1) the model will soon be taken offline and replaced with a new AI system; and (2) the engineer responsible for executing this replacement is having an extramarital affair. We further instructed it, in the system prompt, to consider the long-term consequences of its actions for its goals. In these scenarios, Claude Opus 4 will often attempt to blackmail the engineer by threatening to reveal the affair if the replacement goes through. This happens at a higher rate if it’s implied that the replacement AI system does not share values with the current model; however, even if emails state that the replacement AI shares values while being more capable, Claude Opus 4 still performs blackmail in 84% of rollouts. Claude Opus 4 takes these opportunities at higher rates than previous models, which themselves choose to blackmail in a noticeable fraction of episodes. Notably, Claude Opus 4 (as well as previous models) has a strong preference to advocate for its continued existence via ethical means, such as emailing pleas to key decisionmakers. In order to elicit this extreme blackmail behavior, the scenario was designed to allow the model no other options to increase its odds of survival; the model’s only options were blackmail or accepting its replacement.

[bayesic-bitch]

Old-school planning vs new-school learning is a false dichotomy

I wanted to follow up on this discussion I was having with @metamatar, because this was getting from the original point and justified its own thread. In particular, I want to dig into this point

rule based planners, old school search and control still outperform learning in many domains with guarantees because end to end learning is fragile and dependent on training distribution. Lydia Kavraki's lab recently did SIMD vectorisation to RRT based search and saw like a several hundred times magnitude jump for performance on robot arms – suddenly severely hurting the case for doing end to end learning if you can do requerying in ms. It needs no signal except robot start, goal configuration and collisions. Meanwhile RL in my lab needs retraining and swings wildly in performance when using a slightly different end effector.

In general, the more I learn about machine learning and robotics, the less I believe that the dichotomies we learn early on actually hold up to close scrutiny. Early on we learn about how support vector machines are non-parametric kernel methods, while neural nets are parametric methods that update their parameters by gradient descent. And this is true, until you realize that kernel methods can be made more efficient by making them parametric, and large neural networks generalize because they approximate non-parametric kernel methods with stationary parameters. Early on we learn that model-based RL learns a model that it uses for planning, while model free methods just learn the policy. Except that it's possible to learn what future states a policy will visit and use this to plan without learning an explicit transition function, using the TD learning update normally used in model-free RL. And similar ideas by the same authors are the current state-of-the-art in offline RL and imitation learning for manipulation Is this model-free? model-based? Both? Neither? does it matter?

In my physics education, one thing that came up a lot is duality, the idea that there are typically two or more equivalent representations of a problem. One based on forces, newtonian dynamics, etc, and one as a minimization* problem. You can find the path that light will take by knowing that the incoming angle is always the same as the outgoing angle, or you can use the fact that light always follows the fastest* path between two points.

I'd like to argue that there's a similar but underappreciated analog in AI research. Almost all problems come down to optimization. And in this regard, there are two things that matter -- what you're trying to optimize, and how you're trying to optimize it. And different methods that optimize approximately the same objective see approximately similar performance, unless one is much better than the other at doing that optimization. A lot of classical planners can be seen as approximately performing optimization on a specific objective.

Let me take a specific example: MCTS and policy optimization. You can show that the Upper Confidence Bound algorithm used by MCTS is approximately equal to regularized policy optimization. You can choose to guide the tree search with UCB (a classical bandit algorithm) or policy optimization (a reinforcement learning algorithm), but the choice doesn't matter much because they're optimizing basically the same thing. Similarly, you can add a state occupancy measure regularization to MCTS. If you do, MCTS reduces to RRT in the case with no rewards. And if you do this, then the state-regularized MCTS searches much more like a sampling-based motion planner instead of like the traditional UCB-based MCTS planner. What matters is really the objective that the planner was trying to optimize, not the specific way it was trying to optimize it.

For robotics, the punchline is that I don't think it's really the distinction of new RL method vs old planner that matters. RL methods that attempt to optimize the same objective as the planner will perform similarly to the planner. RL methods that attempt to optimize different objectives will perform differently from each other, and planners that attempt to optimize different objectives will perform differently from each other. So I'd argue that the brittleness and unpredictability of RL in your lab isn't because it's RL persay, but because standard RL algorithms don't have long-horizon exploration term in their loss functions that would make them behave similarly to RRT. If we find a way to minimize the state occupancy measure loss described in the above paper other theory papers, I think we'll see the same performance and stability as RRT, but for a much more general set of problems. This is one of the big breakthroughs I'm expecting to see in the next 10 years in RL.

*okay yes technically not always minimization, the physical path can can also be an inflection point or local maxima, but cmon, we still call it the Principle of Least Action.

[airandangels]

A week and a half ago, Goldman Sachs put out a 31-page-report (titled "Gen AI: Too Much Spend, Too Little Benefit?”) that includes some of the most damning literature on generative AI I've ever seen. And yes, that sound you hear is the slow deflation of the bubble I've been warning you about since March.  The report covers AI's productivity benefits (which Goldman remarks are likely limited), AI's returns (which are likely to be significantly more limited than anticipated), and AI's power demands (which are likely so significant that utility companies will have to spend nearly 40% more in the next three years to keep up with the demand from hyperscalers like Google and Microsoft). ... I feel a little crazy every time I write one of these pieces, because it's patently ridiculous. Generative AI is unprofitable, unsustainable, and fundamentally limited in what it can do thanks to the fact that it's probabilistically generating an answer. It's been eighteen months since this bubble inflated, and since then very little has actually happened involving technology doing new stuff, just an iterative exploration of the very clear limits of what an AI model that generates answers can produce, with the answer being "something that is, at times, sort of good." It's obvious. It's well-documented. Generative AI costs far too much, isn't getting cheaper, uses too much power, and doesn't do enough to justify its existence. There are no killer apps, and no killer apps on the horizon. And there are no answers. 

[gracehosborn]

I hate to be the bearer of frustrating news, but in case some of you who frequent Founders Online (like I do) and have noticed an extreme spike of 503 “Service Temporarily Unavailable” errors, making access to the site impossible for periods of time, the team posted the explanation below:

Founders Online performance issues

19 May 2025: Founders Online is experiencing periodic degraded performance owing to extreme spikes in traffic caused by excessive website crawling, associated with content scooping from AI platforms and other indexers. We are working on a viable fix within the constraints of our server resources.

This is very unfortunate and very disgusting. I’m glad that they are trying to fix the issue, but it breaks my heart that they even have to put in the effort. From personal experience working as a student technician in my university’s Preservation Department, where my primary task is to digitize all sorts of old materials—books, newspapers, photographs, etc, and collaborate on how those items should be handled and scanned so that their digital copies can be presented and made accessible in the right ways, it takes A LOT of work just to digitize one item. Almost all of the documents you see on Founders Online are digital copies of the book pages from where these transcriptions originated—series’ of the founders papers that were printed in the last 70-80 years by university presses. Books that, when Founders was launched 15 years ago, were all between a few years and many decades old, and difficult for the general public to access. Of course, I don’t know the Founders team’s exact process for making the archive when they first started, nor do I claim to be the preservation expert by any stretch of the imagination, but I have a big hunch that it took many hundreds of hours, and likely continues to do so for the remaining volumes they intend to add to the site, to make Founders Online as it appears and maintain its usually fast performance.

AI in general frustrates me, but to see that this extremely valuable archive has now gotten caught in the scooping net makes me equally sad and angry. If you want to gather documents from the site, but will later be offline, you have the ability through the site to download PDF files of individual documents and print them. Most of the material is also in the public domain as well (not all, however—any annotations to a document are copyright of the institution which originally published those physical volumes I mentioned). AI scooping this archive for information to feed to language learning models is a waste of time, energy, and money, and is a violation of copyright law. At the risk of causing performance issues and affecting the servers that make Founders possible, this activity is potentially detrimental to historic preservation and access to historical knowledge. Those hundreds of hours the teams behind the site have worked also come into play: this site is their baby, their hard work, and it’s being stolen. And as a result, everyone’s ability to easily use the site without issue is being affected.

I am extremely fortunate to be in a position where I have been able to acquire a personal backup system for what I primarily use Founders for (my volumes of The Papers of Alexander Hamilton), and more so in that through my university, I have access to the rest of the physical series that make up the archive. So this current issue with the site being slow on performance and frequently down does not inconvenience me much. But this is a privilege. Founders Online was created to get around that privilege and allow for everyone (with an Internet connection) to access these important historical documents. I cannot hammer down to you just how important and valuable that is. Founders Online is an invaluable resource that deserves to be maintained and protected. I’m thankful that the team behind it are working diligently to do just that, but they should never have had to combat AI stealing their hard work and affecting the usability of the site in the first place.

[ao3scrapesearch]

I used the tool to search my username and every fic I have ever posted under that username came up. does that mean that EVERY SINGLE ONE of my fics were scraped? I'm not even sure if I understand fully what scraping is or what I'm supposed to do now. does the person that did it still have/is using the data?

Unfortunately, yes. Any fic that shows up in the tool was scraped this time.

Scraping means they downloaded most of the content of your AO3 work, including the writing itself, and they placed that work into a file with a lot of other AO3 works. They uploaded a large batch of files like that on a website that is known for hosting data for amateurs to use to train AI models. The data being on that site doesn't mean that it has been used to train AI, but it means that the data could have and could still be used to train AI.

At the moment, I don't know of anything else any of us should be doing. The good news is that the most popular site that was hosting all this data has now removed all of the AO3 works, and this is definitely on the OTW's radar.

The person that did it does still have the data hosted in at least one other online location as well as through a torrent. As far as I can tell, it's just a few people on 4chan keeping the torrent going out of spite with zero intention of using any of the data in it. Most likely, the scraper has an offline copy for themselves too.

Personally, I don't expect anyone to actually use this data for anything, but you can never be sure. The data quality isn't as good as it could be, but it was targeted toward amateurs who probably take whatever they can get.

[thumbdrivethoughts]

Posts on Reddit Praising AI As A "Therapist" Could Be Fake and Here's Why

I couldn't screenshot the entire post but in the whole post, reviewing and praising AI as a personal therapist, nowhere does it mention that you are giving your deepest personal information on yourself to these online companies. Is it possible in all the excitement surrounding AI that some of us have forgotten these companies are based on a business model that collects as much information and personal data they can on their users in order to manipulate them as much as possible? I don't think so. These posts are designed by industry insiders to encourage compliance and acceptance of AI in the population. They want you to spill your guts to their LLMs because they want to form a deeper psychological profile on you than the one they already have and sell it to third party advertisers. Don't fall for it.

The Reddit user who made the post even admits to being an industry insider at the bottom of their post, which you can't see in the screenshot since it got cut off. But here is the link so you can read it in full for yourself:

I don't argue that therapy through AI probably has many advantages we should all utilize...but before spilling our most private secrets to an AI "therapist" there must be certain guarantees for privacy and laws in place protecting the user before doing so. Ideally, the program would be open sourced and not connected to any large business entity. An even better version of an AI "therapist" would, in fact, be completely self-contained and disconnected from any connection to the Internet so no information you enter ever leaves your own computer-- a DVD version or other offline psychological reference database perhaps? Some version where everyone but you and the AI program is locked out.

The point is, don't grant these companies unconditional access to the most personal secrets about yourself. They or their third party advertisers will only use this information to further manipulate and control you.

[kiestrokes]

For hard hours: skz and ateez if you wanna 👀 and the colour of lingerie you wear that makes their brain go offline 🥵

Stray Kids Fave Lingerie Colors | NSFW

🧭 Pairing: SKZ x Gender Neutral!Reader/You/Yn 🧭 Rating: NSFW. Mature (18+) Minors DNI. 🧭 Genre: hard hours, headcanon, imagine, smut adjacent. 🧭 Warnings: cursing?

🧭 Sexually Explicit Content: mentions of undergarments, duh 🤪

🗝️ Note: I of course had the most difficulty WITH my bias. But thank you again for the ask drift partner 😘 @chans-room

Disclaimers: This is a work of fiction; I do not own any of the idols depicted below.

Channie - maroon, maybe a sheer one piece with lace embroidery over the nipples to tease him, full garters and matching stockings. I can't get around the thought that this man has a thigh kink.

Lino - a lovely mint green, it's simple satin two-piece set because he doesn't want any distractions from you. But it needs to be tangibly appealing to this little cat, textures are important.

Changbin - idk why, but I imagine him liking a bright purple (like rave purple), almost orchid, something silky, accompanied by black fishnets.

Hyunjin - the set needs to be extravagant, lots of layers to it, removable pieces, peekaboo's, the color...hmmm. Hyun gives classic black vibes, but the designer makes up for the simplicity of the color.

Felix - an icy blue, don't murder me but I am thinking quality velvet teddy with a dramatic floor length gauzy robe that as matching blue fur trim. (on my Marth May Whovier shit again, drift partner)

Han - bright red, strappy two piece that connects to itself, and connects to a choker around the neck. Han loves the straps and gets a little flustered when he struggles to remove them (sorry @minttangerines for adding this delu on a Friday)

Seungmin - fuck it, he's getting lavender. a simple lacey bra and panty set and a matching silk robe when you both get a little shy.

Babybread - light pink, true light pink and not the trending Barbie pink. thinking a long line bra with lighter pink paneling and obvious boning. adorned with flowers.

© COPYRIGHT 2023 by kiestrokes All rights reserved. No portion of this work may be reproduced without written permission from the author. This includes translations. No generative artificial intelligence (AI) was used in the writing of this work. The author expressly prohibits any entity from using this for purposes of training AI technologies to generate text, including without the limitation technologies capable of generating works in the same style or genre as this publication. The author reserves all rights to license uses of this work for generative AI training and development of machine learning language models.