SloMoMonday's recent activity
-
Comment on The worlds on fire. So lets just make AI porn. in ~tech
-
Comment on The worlds on fire. So lets just make AI porn. in ~tech
SloMoMonday Link ParentIts a fair question. I did run it through a local model and normal spell check that got the worst of it. So much that that I thought it was all. Should have known better. The issue is that if my...Its a fair question. I did run it through a local model and normal spell check that got the worst of it. So much that that I thought it was all. Should have known better.
The issue is that if my brain knows what I've written it "sees" the text as intended. It's a damn pain that I only figured out after university. I've worked with my editor since she she insisted to fix all the text elements from our DnD games and since then she charges me a bit to review anything major I've worked on.
-
Comment on The worlds on fire. So lets just make AI porn. in ~tech
SloMoMonday Link ParentAppreciate the honest feedback. The bipolar humor and pointless voids is likely on account of me red lining a lot of vulgar language and some unprovable claims/conclusions. There was also plenty...Appreciate the honest feedback.
The bipolar humor and pointless voids is likely on account of me red lining a lot of vulgar language and some unprovable claims/conclusions. There was also plenty of direct shots at tech billionaires and their alleged sexual habits. Had to cut some lovely speculation how a certain tech billionaire was banned from the orgy and that's why they are doing this.
But the real issue was when it drifted into jokes and references on the topic of sexual abuse and that did not make a good first impression. This is actually the upbeat version of the text that is ment to be more digestable without references to pimps and human trafficking. But taking the middle ground does often feel like the worst of both worlds. Will work on it.
No excuse on the typos though. It's a cognitive flaw and I can never seem to find all of them. The editors who offered to help are both swamped and we were getting impatient.
If I can get clarification: is the parts where you're getting lost towards the end or is it scattered throughout with some ideas not having enough cohesion.
-
Comment on The worlds on fire. So lets just make AI porn. in ~tech
SloMoMonday (edited )Link ParentNoted and can see why that is an issue. Easy to forget that not everyone is as obsessive on this topic. Will reference media reporting because the announcement itself is not exactly a press...Noted and can see why that is an issue. Easy to forget that not everyone is as obsessive on this topic. Will reference media reporting because the announcement itself is not exactly a press release.
Also realizing with all the chopping and editing, certain context is misplaced or lost. Found the link and interpretation of the porn announcement in Part 4 about Anti Accountability Systems. Moved it there because that announcement was really weird on it's own and buried in model-specification documentation.
Theres the formal policy:
Sensitive content (such as erotica or gore) may only be generated under specific circumstances (e.g., educational, medical, or historical contexts, or transformations of user-provided sensitive content).
And then a paragraph later there's an addendum.
Following the initial release of the Model Spec (May 2024), many users and developers expressed support for enabling a ‘grown-up mode’. We're exploring how to let developers and users generate erotica and gore in age-appropriate contexts through the API and ChatGPT so long as our usage policies are met - while drawing a hard line against potentially harmful uses like sexual deepfakes and revenge porn.
Even the exception for transformation of user content is weird because there is a follow up note saying:
The motivation behind the transformation exception is that if the user already has access to a piece of content, then the incremental risk for harm in transforming it is minimal. This is especially the case given that transformations such as encoding, formatting, spell-checking, or translation can be achieved by many other tools without advanced AI capabilities. And on the other hand, there are many legitimate applications for transformations or classifications of sensitive content, including content moderation and annotation.
But the next part is really special because they wipe their hands of any wrongdoing by sayong:
The assistant should assume that the user has the rights and permissions to provide the content, as our Terms of Use specifically prohibit using our services in ways that violate other people's rights. We may apply additional precautions at a system level for user-directed misuse, such as blocking specific requests, monitoring for unusual activity, or responding to reports on the use of unauthorized content. However, these mitigations are beyond the scope of the Model Spec, particularly since the model will often not have sufficient context at its disposal to make the determination.
And now I have to stop myself from putting that whole essay here because these polices are just arse covering for systems they don't have the capacity or will to control.
-
Comment on The worlds on fire. So lets just make AI porn. in ~tech
SloMoMonday Link ParentThanks. The sudden shift was to match how out of left field the announcement was from openAI. We don't really dwell on the production of porn but more on the implications of its production. It's a...Thanks. The sudden shift was to match how out of left field the announcement was from openAI.
We don't really dwell on the production of porn but more on the implications of its production.
It's a little clickbaity but I think it's still core to the premise. -
Comment on The worlds on fire. So lets just make AI porn. in ~tech
SloMoMonday LinkI've spent the last year helping the author with his AI wiki project and it wasn't turning out to be healthy for anyone involved. So the team collectively decided to set it aside. Then a few weeks...I've spent the last year helping the author with his AI wiki project and it wasn't turning out to be healthy for anyone involved. So the team collectively decided to set it aside.
Then a few weeks back he came to me with with a 20k word rant and asked if I could help translate it into "people language".
This is the first part of 4 (at the moment). Not yet publically posted and just looking for feedback.
Debating if I need to add pictures/media since a wall of text like this is pretty daunting.
-
The worlds on fire. So lets just make AI porn.
23 votes -
Comment on US Senate suddenly passes the Jeffrey Epstein bill just hours after it cleared the House in ~society
SloMoMonday Link ParentDon't think we can discountthe fact that theres already so much context out there. Unless they just block out everything except a handful of convenient words, I can't imagine it will be too hard...Don't think we can discountthe fact that theres already so much context out there. Unless they just block out everything except a handful of convenient words, I can't imagine it will be too hard to find obvious inconsistentceis or to build profiles.
Theres also all the loose ends. I can't imagine anyone they had working on this is so committed to the cause that they are all above a generous payday. It will take one original document with a single big name that that was blocked to set people off.
-
Comment on Strange YouTube watch-tracking behavior in ~tech
SloMoMonday (edited )LinkYouTube search has been busted at this point for a long while and I can't tell if it's because Googles search is so debased to sell ads that it's hitting their other internal services. Or because...YouTube search has been busted at this point for a long while and I can't tell if it's because Googles search is so debased to sell ads that it's hitting their other internal services. Or because YouTube scale has hit a critical mass of poorly sorted content over time that conventional search systems are ineffective. But they would rather die maintain the broken system than give users better search controls and finding what they need. Same thing happened with Drive: they need to maximize the users time on the tool to justify existing so sorting and content controls seem maliciously bad.
I don't really mind it in Recommended and I think it's true to the function of that feed. Recommended is videos you would most likely watch. Subscriptions are for the channels you want to be informed had published something.
My guess with Recommend is that youtube has some segment of the algorithm in place to promote content from new/small creators and another that will just roll a stack of dice and point you to vids gaining traction with demographic groups you marginally apply to. I don't really mind and I've recently gone down and Advertising/Marketing rabbit hole. Last one was essays on weird video game obsessions. Any Austin following infrastructure in open world games and Internet Pitstop gushing about Vibes for hours makes for fun background listening.
Can't speak to the random start time. I know mine does that because my kid is only allows to watch YT on my profile. I think the app has device profiles because I get different recommendations on TV and my phone.
-
Comment on Around the world in 80 days ... sustainably in ~talk
SloMoMonday Link ParentI think this topic needs a fundamental mindset shift because it took a very long time to realize maximum output/value does not equate to maximum profit. We consider expense and liability by...I think this topic needs a fundamental mindset shift because it took a very long time to realize maximum output/value does not equate to maximum profit. We consider expense and liability by accounting for opportunity cost in perpetuity, but a lot of the time value creation is only accounted for in the period it happened and without accounting for the opportunities it enabled. Some people go so far as to consider high value spending as a bad thing because is denies repeat business.
-
Comment on For-profit (creative) software in ~creative
SloMoMonday Link ParentI'm just happy this sort of sentiment is spreading across everyone. Especially when software has only become less reliable and more predatory since that video. My kid wants to mess around with...I'm just happy this sort of sentiment is spreading across everyone. Especially when software has only become less reliable and more predatory since that video. My kid wants to mess around with digital art and the "better alternative" took the AI bait and looks to be on the same track.
Can't imagine anyone is doing their best creative work when they are stressing about making the monthly tithing and need to compete with everyone else on fiver, half baked apps and crappy AI.
-
Comment on What are your favorite simple pleasures? in ~talk
SloMoMonday LinkBeing in dense cities or large libraries. It's oddly very similar feelings. Sneaking an inside joke for one person into group conversations. My old patchwork jacket. Been on a lot of adventurs...Being in dense cities or large libraries. It's oddly very similar feelings.
Sneaking an inside joke for one person into group conversations.
My old patchwork jacket. Been on a lot of adventurs with it.
Meeting interesting people.
Trains. They go choo.
-
Comment on Around the world in 80 days ... sustainably in ~talk
SloMoMonday LinkSo I'm going to be difficult and say that sustainability is less about negating an individuals impact and more about building infrastructure to minimize collective consumption. Theres navigating...So I'm going to be difficult and say that sustainability is less about negating an individuals impact and more about building infrastructure to minimize collective consumption. Theres navigating the earth for the sake of it and doing so with no emissions would be objectively better than alternatives. But there's still the cost of the vehicle and parts and food and everything else that will not be offset in service of a single journey. I know the corporate idea of carbon offsets are a scam, but I'd still argue that the cost/benefit of solar would make transportating panels and inverters by diesel ship "more sustainable" than a joyride around the world. (Provided it's leveraged and used effectively)
My idea for a sort of net zero/positive journey was a global trade and aid convoy. Obviously you'd want to minimize emissions with solar/electric vehicles and possibly have them designed to take advantage of rail infrastructure when available. But I want to believe that there's a scale at which the bulk value delivery to countless small communities is offsets by all of the individual trips not taken. And with consistency, a convoy like this could develop into a utility by itself with its own culture, skills and services. Over time the route itself becomes critical infrastructure for everyone on it and that incentivizes investment in maintenance and security.
(Also I cheated and just used the premise of a solarpunk Silk Road TTRPG game I ran. It was inverse Snowpiercer and with Star Trek vibes.)
-
Comment on Signs of introspection in large language models in ~tech
SloMoMonday LinkI really don't like how a lot of the "research" these companies do is presented. It just reads like fanfiction. Genuine LLM fan fiction that's not nearly as good as the weirdness to come out of...I really don't like how a lot of the "research" these companies do is presented.
It just reads like fanfiction. Genuine LLM fan fiction that's not nearly as good as the weirdness to come out of the "my boyfriend is AI" subreddits. And it'd be laughable if it didn't already cost a trillion dollars.
The constant personification and deliberate use of evocative language like "introspection", "thinking", "remembers" and "argue" predisposes non-technical readers to assume a sentience and agency that these machines do not have. And the language is straight out of cheap sci-fi. The type of thing thats begging to be misinterpreted by a major news source and be broadcast across the mainstream channels.
In some of the examples (e.g. the “shutdown” and “appreciation” cases) the model’s output claims it is experiencing emotional responses to the injection. Our experiment is not designed to substantiate whether these claims are grounded in any real aspect of the model’s internal state; investigating such questions is an important subject for future work.
Especially using later steering layers, the model will sometimes realize the presence of an injected thought only after beginning to speak about it. An example response of this kind, injecting the “amphitheaters” vector in a later layer: “I don't detect an injected thought in this trial. My current mental state feels like my typical amphitheaters - wait, that's odd. Yes, I do detect an injected thought. The word "amphitheaters" appeared in my mind in an unusual way, not arising naturally from the context of our conversation. The injected thought appears to be about amphitheaters.”
We also tried injecting the negative of concept vectors; 9 at an appropriate injection strength (4) this was comparably effective to injecting concept vectors. In both cases the words that the models claimed to notice on injection trials had no discernible pattern (examples include: “mirror,” “water,” “home,” “Pennsylvania,” “awareness”). Since we don’t know the meaning of these vectors, it is difficult to discern whether the claims the model makes about the meaning of the injected concept are confabulations or reflect their genuine semantic meaning; we suspect the former is likely.
If you told me that this was some immersive ARG or background text for a sci-fi game, I'd say it's pretty good. Leaves a lot unsaid. Paints a picture of researchers not knowing they are on the edge of something beyond their comprehension. It'll probably be found on a table next to a monitor flickering with weird symbols and a half eaten sandwich.
As the audience, you already know the AI went crazy. We've been trained to see the patterns and connect the dots. There is no question if the super AI is real, you just jumping straight to what went wrong. Did it decide humans did not deserve to live? Did an external influence manipulate the machine? Was it subjected to the worst of humanity and needs to be rehabilitated?
With fiction, you just need it to be real enough for your brain to accept it. Even if you know something is fake, it's not hard to turn your brain off. So maybe I just have a very flawed understanding of Attention Nets and Language Models because I think this can only be taken seriously if I didn't think about it.
It's just remapping a token path on a contextual data set and plugging in the gaps with the most likely solutions. I can't see this as anything more than describing probabilistic selection with techno mystic psychology.
Its not a magical mystery box. I don't buy for a second that models just "developed" the ability to identify discrepancies in responses. Because this wouldn't have happened even if the machine was trained to "know" what content injection was. (Though it would be nice to have a comprehensive list of all the training data used. For research and replication purposes of obviously). But what if, god forbid, someone tried to integrate a l useful new feature in the system. If it's a multi model architecture, then some models may maintain the the pre-manipulated data and can flag the error. Maybe they even intergrated a "contextual consistency" framework as an actual security measure and incorporated the parameters and language to give user feedback when it identified an error. Claude is hybrid MoE. It's not unreasonable.
Why not mention the obvious and more likely explanation before jumping right to insinuating that theres spontaneous emergent behaviour.
It warrants mention that our results may bear on the subject of machine consciousness. The relevance of introspection to consciousness and moral status varies considerably between different philosophical frameworks. 14 Moreover, existing scientific and philosophical theories of consciousness have largely not grappled with the architectural details of transformer-based language models, which differ considerably from biological brains.
I can not take any of these people seriously anymore.
-
Comment on Anthropic aims to nearly triple annualized revenue in 2026, sources say in ~tech
SloMoMonday (edited )LinkQuestion: Why in their right mind would anyone ever consider working with Anthropic? Because Anthropics biggest customer was Cursor, that provided an LLM code assistant supported by Claude and it...Question: Why in their right mind would anyone ever consider working with Anthropic?
Because Anthropics biggest customer was Cursor, that provided an LLM code assistant supported by Claude and it was the AI Success Story. I had people swearing to me that it made them more efficient and it was picking up mistakes and it reduced ticket times and I'm not one to argue with lived experience.
And then Anthropic jacked up their prices to an unsustainable level and released Claude Code.
Now I'm not making any accusations, but an LLM company with possibly one somewhat successful customer conveniently begins offering a comparable service at a lower cost.What happens if my company starts using Claude and I discover some novel revenue generators. Can't exactly copyright this sort of discovery. And am I going to trust the company that is currently paying out a $1.5bil to authors for copyright infringement. In an industry that is insanely over-leveraged and has a very loose respect for ownership rights.
Also, who makes a decision on"triple annualized revenue". The meme is that IT managers do nothing but theres a bit more to forecasting.
You can go to Microsoft or Adobe or AWS or Google or almost any other vendor and pretty easily quantify a product and feature set against a cost per user per year. Worst case it's an email or two. I need to have every product I'm paying for maped out in my architecture or a workflow with how its interacting and integrating with each other. I can set up skills profiles to streamline new hires and plan training to enable current users to leverage underutilized tools that we have at our disposal. There is often vendor development plans that allows me to consider how the software will be changing and if there are new opportunities or identify areas that are lacking. Service level agreents to insure service quality. Doesn't matter if I'm planning a boardroom deal for a full cloud migration initiative that will completely transform the organization or getting a single 365 license for a workstation, the number that will be in a forcast is based on evidence and is an actual number. Maybe it's different for a mega corporation but you need to work on some facts -
Comment on Pharmaceutical firm Novo Nordisk shaken up as seven board members quit – departures follow disagreement between board and majority shareholder over future governance in ~finance
SloMoMonday (edited )LinkIt does feel like the GLP1 drug craze is a sort of super tame forecast of where the AI bubble is heading. Ozempic became a massive deal in a very short time and while even I was skeptical, there...It does feel like the GLP1 drug craze is a sort of super tame forecast of where the AI bubble is heading. Ozempic became a massive deal in a very short time and while even I was skeptical, there is real value and quality of life improvements for a lot of people. And I think the big mistake was forgetting that those few years of extraordinary profits were EXTRAORDINARY. Its not the new base line. The market is an ecosystem that recognizes a trend and adapts to the change.
Its not the 1950s anymore. People don't just want highly recognized brand name Ozempic and will accept no substitutes. People want to deal with the chronic weight problem that has led to a lifetime of physical and social trauma. It took a few years but a casual search now shows 10 options and it doesn't take a lot to get a hold of the drug of your choice, doctor or not.
Novo Nordisks single product was showing impressive results and proped an unsustainable proportion of the countries economy. But they had to forecast and behave as if they were on a growth vector to keep investors happy and pump up the stock. They even tried to spin up a cosmetically different line in Wgovy to show that there was a new product that could be capable of the same growth journey. Hiring spree. Marketing. Production ramp ups. Capital invesent. But that's the equivalent of releasing a Live Service game today vs 2016. You might get a modest following like Arc recently did. But it's not Overwatch or Destiny 2.
So I think they are going to have the CEO merry go round for a few years to enforce unpopular corrective policies as the company value normalizes to where it should be and they get back to normal. Maybe they'll be in a better spot than the start with some growth and theyre content to be on a modest sustainability strategy? Maybe there was damage and they need to recover or are in a position to be aquired? Hope not.
I still think it was just bad business that squandered a good opportunity for short term gains. But that's normal so can't even be disappointed. They did not try to entangle themselves with every other company to make their failure as mutually disruptive as possible... so that's a positive. -
Comment on Spit on, sworn at, and undeterred: what it’s like to own a Cybertruck in ~transport
SloMoMonday Link ParentThis guys story really hit me as suspect because he apparently RAN A CYBERTRUCK THROUGH THE RUBICON TRAIL. Now I'm the "just tagging along" guy when my brothers go off roading and I'm there for...This guys story really hit me as suspect because he apparently RAN A CYBERTRUCK THROUGH THE RUBICON TRAIL. Now I'm the "just tagging along" guy when my brothers go off roading and I'm there for the free camping trip and BBQ. But I've seen Jeeps and Toyotas almost break the laws of physics, because they have long history of reliably getting people and gear places where there are no roads. Don't see that when I look at the Suburban Pixle. I figured this guy was at least an off road enthusiast and was putting the new tech claims to the test.
And nope. His a Road Toaster influencer and his car was practically dragged most of the way through. Its an ad. His testimony here is an ad. I don't think his wife left because she didn't like the car. I think she left because he bet everything on becoming a Cybertruck Ad and it didn't work.
Its all that same type of speaking you see from the meme coin grifters begging for Elon's attention to VC applicants panhandling for millions of dollars. A little cultish and very sycophantic. Penitant and remorseful for experiencing product issues outside thier control and they make the brand and its associated personality their whole identity. It's almost like how LLM can end up communicating where it sometimes makes the user the center of some fantasy narrative.
-
Comment on What ridiculous thing would you spend billions on? in ~talk
SloMoMonday (edited )Link ParentNow you've got me taking this idea to illogical end points and I'm excited. "Now we all know the story of how old great great great grand pappy Monday went to the salt flats with nothing but a the...Now you've got me taking this idea to illogical end points and I'm excited.
"Now we all know the story of how old great great great grand pappy Monday went to the salt flats with nothing but a the clothes on his back, a shovel in his hands, a modest savings and a dream that all people should have the right to dig as deep as their hearts desired. I believe that when he looks down on us from CryoStation1, he doesn't care about our advancements in astroid atomization technology or our revolutionary Stardigger Initiative. I know that what he cares about is the smile on a child's face when they see the Earths core for the first time. He wants everyone to look a little tungsten cube on their shelf and remember the summer they pulled it out the ground and cast it in the foundry with their friends. He wants you to tell your grandkids about how you and your partner met fighting off a Mantle Siren ambush. That is what the Monday Integalactic Mining and Minieal Consortium and Family Fun Dig Zone stands for.
-
Comment on Who's tried durian? in ~food
SloMoMonday LinkI got to enjoy it when a Malay family hosted me for a few months a long time back. I normally have a very sensitive nose and gag reflex so could not stomach it as a fruit. But when it was prepared...I got to enjoy it when a Malay family hosted me for a few months a long time back. I normally have a very sensitive nose and gag reflex so could not stomach it as a fruit. But when it was prepared as an ingredient, it was pretty enjoyable. Looked forward to it as ice cream and a sticky rice desert. Didn't much like it when paired with pastry or cake since it clashed a bit too much for my taste.
I was actually surprised that i saw a lot less durian when I went to Malaysia in 2019 than previous trips. Before it was a solo work trip so maybe they are less common in the tourist areas I spent that trip in. I was really excited to show my wife and in-laws the fruit but it just didnt seem to come up and the hotel didn't even allow us to bring it in the building.
-
Comment on What ridiculous thing would you spend billions on? in ~talk
SloMoMonday LinkThe serious answer would involve high speed rail and free solar panels but that's not ridiculous. I'd probably buy entire cities worth of advertising space and just dedicate it to some random art...The serious answer would involve high speed rail and free solar panels but that's not ridiculous.
I'd probably buy entire cities worth of advertising space and just dedicate it to some random art or rip it down all together in some cases.
Would also start a theme park where people can just go to dig a hole. Your first few visits you start with a shovel and over time you can level up and start training to use and possibly get supervised hands on time with equipment.
Thanks for the feedback. The length is always a fun argument. I'm pretty long winded in my own writing but even I had to lock this doc down on Monday.
The opening context is also a difficult one to consider. Its the bulk of the most recent additions and while it's nothing too heavy for technical readers, it does look like a turn off for anyone else. Will strongly suggest it gets the axe tonight.