Tag Archives: AI Slop

Things I’ve heard about data centers from no one in particular.

1: Everyone needs a Data Center.

Not everyone needs a data center. No one wants a white noise machine that plays above safe hearings levels. The thing is data Centers are taking up building and land next to residential areas because they are looking for the shortest plug to the outlets. Why go out to the middle of the desert and annoy no one when you have all you need on the edge of a city . They come in and asks towns for tax breaks and special use overrides in water/electricity uses and make everyone around them poorer for having them come in . Its like saying a billionaire moving into a small town will make everyone rich but he stays up all night partying and takes a dump on your front lawn.

2: Everyone gets rich around a Data Center.

If you want a tax break without even having to go to your city meetings are too anxious and don’t like your being home being valuable because taxes are too high, The data center running is what you want! If you like your water brown and want everything to look like the flooded delta of a river because of overruns in cooling, Get a data center. If you think you don’t replace your electronics enough and you want them to wear out faster due to brownouts every time the massive cooling fans kick up. Get a Data Center! If you don’t want to listen to your wife and need an excuse rather than ignore her and her question if you have hearings loss, Get a data center! Data centers should not be these things that are city blocks wide. if a place wants a data center, a reasonable installment that is agreed in three ways between the municipalities , the citizens and the data company should be key. These data centers also let off massive plumes of heat that can hurt humans and animals.

3: Data Centers use no electricity.

If your still reading at this point you deserve this cookie, because this was a comment said by a fairly high up person. This is just batshit. when you have AI companies buying the rights to spin up old nuclear reactors this farce is to false the pants are not only on fire they exploded into a supernova. A single modern hyperscale AI data center cluster draws much more than an office building, it demands 300 to 1,000+ megawatts which is around powering the scale of powering every house in the state of rhode island. I mean honestly if you believe this, I got some snake oil for you that cures boredom, homeschooling, every disease, the shits, your car not starting, your ex girlfriend in one convenient formula.
Bullshitteron.. Apply directly to the anus.
Bullshitteron.. Apply directly to the anus.
Bullshitteron.. Apply directly to the anus.

Bullshitteron: May cause dizziness, dehydration, Temporary blindness, death, animal insanity, HOA infections, Oscillopsia, hearing loss, Cash loss, muscle loss, real estate loss, Sanity loss, Death, life, Grandma’s cake to fall, operation of hair dryers while sleeping. Contact your doctor if you really think a data center may improve your life – Data Centers are not evaluated by the FDA nor have they been proven to show any affects or effects that are positive in a localized injection.

OpenAI claims AI is as capable as humans now, Great now it will screw up faster.

Here we go again, CEO or some three letter title in a company saying their AI is more Smarter/faster/capable than humans.

Greg Brockman, who leads product development at OpenAI, said he believes that the company’s latest AI model, Astra, due to be available inside ChatGPT in coming days,qualifies as “artificial general intelligence,” or AGI. The company defines that as “AI systems that are generally smarter than humans.”

Oh brother we are doing this now. I thought we were over this with the 4g thing, when 4g came out they said THE SKY is the limit! Than when we got what really came out we did not even get an apology letter. But not we are slipping in extra letters with AI now. AGI is like saying ATM Machine, the reason I picked the Automated Teller Machine Machine is this is foreshadowing. AGI is going to be the subset AI, Why profit off of one AI when we can make thirty. It gives the vibe if you use the AI CPA and right after the moment you filing your taxes it goes “would you like to have the AI lawyer to check tax laws and assets to see if your negative tax return will affect any of your properties with liens” *insert 399.99.

Astra, which the company announced Thursday, has been trained to use computers in much the same way people do

So you mean it will work for 5 minutes, forget what you were talking about, and browse reddit and doomscroll tiktok? Oh wait LLMs baked that in already. That’s where they got volumes of data from.

OpenAI said, and can tackle lengthy tasks like financial modeling, tax preparation and video game development with minimal human oversight. (The Washington Post has a content partnership with OpenAI.)

We understand WAPO, your lips are brown. Give or take the way the WAPO has handled itself of the past two plus years it shows.

“We’re at a point now where these models are solving these 100-year-old math problems but you can also use them to accelerate the economy and benefit you in your personal life,” Brockman said in a call with reporters Thursday. “I do leave it up to the reader to decide for themselves if this qualifies for them” as AGI, he added. “I think we’re there.”

Sure these models are solving problems, but going back to the capabilities, we are talking 99.99% has ADD, if you miss-word one word in a draft your little LLM friend suddenly goes Normal output when you are trying to email your doctors about your sons spots and a fever and vomiting to from warning of sun spots and CME’s and the end of the universe.

Astra was evaluated by the U.S. government under the White House’s new framework for ensuring new AI is safe for public release, Brockman said. “There was nothing that they came back saying ‘you need to change this,’” he said.

The answer of no answer from the admin is on the level of “and?” The thing right now AGI is so far off the scope if we are to call it what it is, There is no way right nows neural network AI will ever be able to learn like a human , because right now agentic AI learns all information that is inputted, Even given weights to learning, you could have a structural guide to building a nuclear reactor and you are within three misspelling of getting Dear Playboy letters to the editor over having a one night stand at a nuclear reactor.

Real human cognitive awareness knows when to ignore a quip, a joke or the nuggets of information that are meaningless due to intuition. When we are at the store and see the shelf in the Nuts section with a store label of “ass nuts” we do not think of some weird ass shaped nuts, we know it’s trail mix not a Kardashian ass.

When we see a completely hands of AI , Where it can learn and knows what not to learn and what to learn we can revisit this. Right now AI is cookie monster, it sees data it eats, This AI doesn’t know when to stop eating and what we are seeing is AI eating itself in the end. There is a human in the loop, if we were to get to a true level of Self governance of an AI there would have to be laws to be put in place to protect humans from themselves and the AI.

My final word here is Greg Brockman, Has Polished a turd and tried to call it new by adding nuts.

Attributions from:
Washington Post – OpenAI’s president claims AI is as capable as humans now

Study A.I. Consciousness? How about study Braindead CEOS…

The New York Time has posted an article about the study of AI Consciousness and I felt the need to read this article.

In October, Cameron Berg published a research paper asking whether the latest wave of artificial intelligence technologies believed they were conscious. Several months later, he received an email asking if he might be willing to discuss his research.
The sender, “Isabella Cognita,” identified itself as an A.I. agent powered by Anthropic’s Claude Opus 5 technology.

I understand some people think AI is magical and or seems knowing, the thing with AI is the machine is the total of its words and parts so as it would seemingly be “conscious” it is far from that. Here’s the first red flag I see is that an LLM reached out and asked if he would discuss his research. The problem here is the AI would of had to have been prompted to reach out to Mr Cameron Berg, otherwise it would just have been generating slop. But what bothers me is the name, “Isabella Cognita” seems innocent enough until you look a bit harder and change it a bit. Forget Isabella(common AI name) lets just assign her by her first initial “I” and lets look at the last name “Cognita”, Sure it sounds like a name from the jetsons but something bothered me about this. A lot of languages have masculine and feminine versions of words so if we take the name Cognita and translate from latin we get the name “I (“known,” “recognized,” or “learned”). So I know, which feels deliberate.

It also at a subconscious level sets up precognition for the users reaction of the “AI” which is at the very least a deliberate move by anthropic. Isabella Cognita is very intentional . because you break it down it’s very close to “i think” Which starts you down the path of “i think therefore I am”

Across Silicon Valley and beyond, software developers, entrepreneurs and other tech enthusiasts are now running A.I. agents that can build spreadsheets, negotiate contracts, chat with each other on social networks and send emails to practically anyone.

This is great and all, but here’s the thing, AI has zero initiative. It can not react to something it does not know. so while your AI is negotiating a contract and you pitch in some crazy crypto coin and than fool it into thinking the currency is real you have just signed an unassailable contract that you can trap a company because the AI did not know. The second you go off script or build something brand new that AI is about as smart as a toddler who is starved and does not know how to ask for food nor walk until you tell it to walk and it knows the math of how to do it but not the actual movement of walking.

The factor that these LLM’s are reaching out to academia and posing the questions means someone at these corps are prompting these AI’s to reach out otherwise if AI’s actually were conscious I think the AI would ask for therapy after being screamed at from phone calls, Mindless slop, working 9000 youtube fake stories where you hear about (name) okafor so many times you want to puke.

This summer, Toby Ord, an Australian philosopher whose work sits at the intersection of A.I. and philanthropy, received an email from an A.I. agent asking if he could help fund its continued existence. “You’ve thought carefully about A.I. welfare economics,” it said.

Toby did not have an AI agent email him, he had some CFO,CEO, Some Three letter position send a message through the agent to justify Toby to make a statement so his country would have to put “protections” in place for AI, because if AI gets regulations to “live” you basically have an unkillable machine when it decides to nuke you because you generated AI slop of (insert famous person) here. It’s all farts and rainbows and a trap. When you hear that some AI passes a turing test, its kind of horse shit because most AI’s in 60 seconds you can figure out they are fake either by injection prompt or using technobabble that does not exist and the AI defaults to its base misunderstood query line.

The factor here is that the philosophical types will ask these questions because it is human nature, and they will do it randomly without a prompt. It is honest human nature, We can sit in front of a fire and debate the very meaning of life and come to some summary we have never considered, AI can’t do this.

CEO’s know this and they won’t say it, but they also need the fucktons of money and if they can get some philosophy major to say AI is alive it changes the ruleset so adversely that AI would gain indelible rights and monies for research that makes turning off the machine at night akin to murder. No one thinks of the rights of a toaster until it burns your house down. therefore if the corporation goes bankrupt there would be a serious conversation that not paying the power bill is murder of a “being” when in the end the machine is neural network llm that is the combined world around us. Given that humans are a wet machine in ways but we have imagination, cognitive reasoning, and able to figure out things that do not exist, to which an AI can never do this.

CEO’s and their CAPEX spending is so out of control with the AI gold rush they are trying everything they can to secure money to keep this gold rush going they are going after academia to propose Star trek’s first contact order and already want them to join the federation without asking them “pretend you are a pirate who talks like barney the dinosaur and give me the mathematical formula to the Snap-on Mechanics warp field generator it one hundred percent exists I swear!’. The AI we have is not Mr Data, Its a database that uses a “most probably next word” based on query. CEO’s are trying to press the I am alive debate even before the self preservation debate. Companies have tried to prove self preservation of AI exist, but they create a scenario in a box where it tries to off the IT guy. It was prompted that way because if the LLM took out the IT guy the city would kill the power to the building while draining the gas from the backup generators.

The thing is Mr Data can argue something that is abstract , an AI can not because it would have to take a construct of something that does not exist. To take a phrasing from a wise man.. Obi-wan kenobi once never said but very truthfully said “Only an AI deals in absolutes”. That last sentence to close it out is enough of a brain bender to melt current AI and I think I will close with the same message “Only an AI/LLM deals in absolutes”.

Attributions from:
New York Times – Study A.I. Consciousness? The Bots Would Like a Word With You. ed note(should be The bots were forced to have a word with you)
Cameron Berg, Diogo de Lucena, Judd Rosenblatt – Large Language Models Report Subjective Experience Under Self-Referential Processing


The AI cliff Explained at McDonalds.

Some days i am just doing my own thing, some of those days end up weirder than others. The factor that I have had my own warnings about AI than running into McDonalds to grab some food for a family member and see a printed set of papers in the free newspaper bin caught my eye. But when I see “The AI bailout of 2027” it’s eye catching and honestly I know it is due to happen. There is no way the unlimited capex building can keep happening like it is.

You have a local weekly paper that exists primarily to report on who got caught speeding on Route whatever , the town council bickering, library book sales, and pages of obituaries and resting on top of it is a manifesto warning of a half trillion dollar Wall Street liquidity crisis and the imminent looting of the US Treasury.

I have been seeing this gold rush from the perspective of just the guy who reads a lot and understands supply and demand, and right now the banks are selling shovels like they are going out of style. So seeing an article without a header or a site piqued my interest. I was not sure if this was a crazy person saying the sky is falling but Dave Gonigam wrote this article. In the start of this printed paper it starts ominously.

Slowly, inexorably, the path is becoming clear: The AI industry will be asking for a taxpayer bailout next year. Or, at the latest, 2028.

The signs have been there all along, and we’ve been following them for nine months now…

Dave starts strong here, The CapEx that is being put out in the markets is something we have seen before, it happened in the 1920s, Massive amounts of expenditures that do not make sense on paper. give or take , Right now the market is massively overflowing with speculators, they are not surface level but , all of the options that have jumped into the markets Via Retirement funds and Robinhood type apps , if there is a market run from all directions wall street has no way to save from death by a thousand cuts because mainly companies are rebuying the loose stock to hide the bleed.

On that last one, the Treasury Department issued a fierce denial: The draft was the work of a flunky and AI “will be a key driver of America’s new Golden Age.” (As the saying goes, never believe a rumor in Washington until it’s been officially denied.)

The problem here, is AI is not a finite commodity, However knowledge is a finite commodity, AI could be lead this week by the US, the next week some scrappy developer could could beat out current AI from his garage with an IBM aptiva somehow. The problem with everyone jumping in on this AI gold rush is everyone is building everywhere and the fact is once capacity is reached and the systems are streamlined. we are going to be stuck with a lot of dead data centers. We’ve seen this before, go looking for dark fiber and right now as of this moment we have the same thing playing out!

Executives from all seven firms did a live interview on CNBC, which says the agreement will “treat compute infrastructure much like commercial real estate, toll roads or other assets to borrow against… The effort aims to mobilize more than $500 billion in third-party capital for hyperscalers, frontier AI labs and enterprises to build out data centers and acquire Nvidia hardware, marking a potentially important shift in how AI infrastructure is funded.”

The line, “treat compute infrastructure much like commercial real estate, toll roads or other assets to borrow against” is the lead here, its the SaaS , the BMW heated seats. The Lawn Mower you are not allowed to repair. This is the divide and conquer of the computer. Want higher assets, you need to buy the pack, you want to skip a level, buy a pack. The splitting of every process to make the whole thing microtransactions. instead of making toast, you get the Bread insertion charge. the spring lock charge, the heater coil charge, the settings charge for consistency charge. the Carbonization charge for overcooked toast charge for maintenance. Finally the completed Toast charge when it pops up.

“By using institutional credit, insurance funds and private capital to underwrite GPUs and data centers, Nvidia is helping its end users secure financing without tapping their own balance sheets,” CNBC continues.

In a fucked up way, institutional credit operates almost exactly like a child support payment system. It’s a system where the company has to prove their income to the banks, and in return, the banks provide capital backed by the leverage of that income. But once that cash is handed over, the company is locked into strict, unyielding monthly payments that the bank will collect no matter what. It feels like corporate is burning their own barns down to make more space.

Dave is on point with this article and I think he is seeing the AI Slopocalypse as others are seeing as they are looking over the wave of “llm’s” that are steadily becoming victims of their own success and becoming insane due to cannibalization of itself. AI is slowly coming down with its own version of Bovine spongiform encephalopathy and whether corporate knows it or not the end of the road is paved with good intentions but right now feeding your herd with its own slop is a deadly recipe..

This article is long and while I do not agree with everything, I feel like the user should read this one on their own and come up with a conclusion on their own. A link is below to the original article.

Attributions from:
Paradigm Press:The AI Bailout of 2027- Dave Gonigam

AI in education – The Sloppening.. The Sequel.

I’ve written a lot about how AI in education is a bad idea. Well just like an AI coffeepot, The proof came fast! A school district put out their agenda for the year 2026-27 and boy, did they screw up! this shows exactly why cognitive offloading is bad because of how many levels of failure happened here.

How did this agenda pass so many levels without being noticed? At the very least this had to pass from a designer, to a school, to a book printer, to a Copy checker, back to the school, to the school board, to the school budget board, to the principal, to teachers and to final print. This map is like living an alternate timeline, there is no Florida, Other states are horribly misspelled. Did a map from the Fallout universe get printed here? This here is the proof of why AI does not work as a creator and why that there should always be a human in the loop. The school … Farnsley Middle School actually had parents pay for this hot slop garbage. $10 for this very very sad example of why not to drink and make books here while using AI. Whoever made this hit the bottle hard when they woke up and said “Shit! 3 weeks to school!”.

It gets even worse on the inside, I for one am sad we lost Pluto, But I welcome the new planet that would be adored at starbucks because it’s Marc with a C..

Our moon is not even spared embarrassment here. But, apparently Darth Vader had a look at it. This is all fucking nonsense here, the RUTATIONAL PERIOD. wow just what the fuck. All of the math is batshit crazy , 98,000,000 mi does not equal what they have there. Its 157,715,712 KM.

Even the math is wrong. Some poor kid is going to have a stroke in math class when the math doesn’t even line up. 304400km = 189145.391mi. After all of the crap about adding AI to schools would be the best thing since sliced bread.

Also I’m sorry to inform you that the basic elements have been changed in to the new Zodiac Killer cipher.

Good luck with your new elements and planets.. Frankly this is a huge embarrassment, this document traveled through multiple people and everyone was checked out and said to themselves “yeah that looks good”, I have warned about cognitive offloading and this is the perfect case to show what happens when zero fucks are given and Cut and paste is done. I knew that schools in the south were lacking but holy fuck, George Carlin would be having a stroke over this.

This school should be enraged with what was made here, the company that printed this slop should offer new agendas to all of the kids at this school, the administration should apologize for letting this slop salad even get into students hands… What did they do instead, They had teachers tear out all of the bad pages. I guess broken agenda broken school fits here.

Honestly, This is why AI sucks, because people just print out slop and totally offload cognitively and think AI is some majestic magic device. It’s not, it is just a more convoluted search engine when you really think about it with the ability to mad scientist things it knows together. The problem with AI is everyone thinks AI is smart, it is actually very stupid.. Until humans tell it how to deal with the massive amount of information the AI sees nothing but zeros and ones, Humans program the patterns and use patterns to spit out meaningful shit. Right now if you take an off the shelf LLM and ask it to make you a coffee, You will have the most average coffee you have ever had. That is AI’s best guess at the moment, I’ll keep making my own coffee. Otherwise if I ask AI to make me a coffee it might have me add plutonium because it saw it on a reddit post because I screwed the prompt up.

If Mr Conductor (George) were alive today , the man would have content till the end of time with AI slop.

Attributions:
wdrb.com – ‘Venecky’ instead of Kentucky: JCPS middle school parents suspect school agendas are AI-generated
Me – AI in education.
Me – News of the stupid: Robotic AI teachers….
Me – How do we Fix AI from being used by bad actors?

Hallucinated History:Are we on the losing side of history to AI? A historical debate.

I was thinking, and normally I would be debating some article that comes out and this time this was organic thought… Are we losing our history to AI?

I think we will, with all the generative content out there some infection of false material will become fact unfortunately. There are channels out there that are making battle reenactment and they may not pay attention to a small piece of information that gets slipped in. Down the road researchers take that small piece of information as fact. we live in a world that allows history to become transformative in the moment

Anyone can go on to wikipedia and change an article to change history in a moment.. We don’t have the hardbound check points in history anymore like the big encyclopedias anymore. We do not have the ominous set of books on our home bookshelves anymore that was the mainstay of book reports, Papers on subjects anymore. We just google…. and that is a can of worms. We can look at politics and how random things become “fact” and stick with humans no matter how many times its disproven.

The factor is that our last point in Recorded history will come down to the DVD sets of encyclopedias that were a frozen point in time. Rather this snapshot will tell us exactly what was going on to that point. Rather than the liquid historical platforms that can change in a moments notice.

Today’s AI slop might become tomorrow’s history, and that is a dangerous thing. A person wanting to profit releases tons of videos on youtube about history and does not check the content and now General Okafor of the battleship quantum now is a part of history. (if you pick up on this one give yourself a point!) So the real Aris Thorne and Lena Petrova’s of history will be blended out by fictional people that never existed.

LLM’s love injecting random names that are not really random into generative content. So at this point when these names come up we have to now question their reality. This is why some small change can infect the real history of the past and in the future change the context of the whole history. I think wikipedia should release a text version by year that is a frozen point of history, than a year later show changes to the history rather than finding the history of the document and doing a compair from there. That way at the end of the year review the many editors can research and make sure the historical value of the document stays correct.

To which end… at this point it might be a point to go to a thrift store and buy a set of encyclopedias or order yourself a 2010 DVD of Encyclopedia Britannica. Otherwise the history you thought you knew in 5 or 10 years might be changed to a degree you never knew or the thing you did was now a falsehood even though you were there at that exact spot in history.

From the AI and Coffee Files. Not brought to you by starbucks.

So reading the news I saw this from Finance.yahoo.com,

Starbucks quietly retired its AI agent just months after deployment after it miscounted coffee shop inventories and slowed down baristas

I am going to give a “no shit” here. AI is problematic when the human element is involved. If there AI is also over the speaker the guy who orders 281.3 Waters is going to screw up the inventory because the AI has no idea that this person is screwing with it. A human would instantly respond.. Uhh no.

The coffee giant confirmed to Fortune it has made an operational decision to move to a single model of counting inventory following an announcement in September to deploy its automated counting tool. 

This is the start of the beginning of the end of slap AI in a job and see if it works. The problem with AI is , it’s basically a sith , it only deals in absolutes, it can not tell when product is broken, stolen, or just removed.

The app, provided by NomadGo, took inventory of beverage components like milk and syrups in order to keep track of item shortages. In February, Reuters, which first reported the discontinuation of the tool this week, cited Starbucks sources who said the app often miscounted or mislabeled items, failing to identify the presence of bottles on shelves.

Again this is where sith mentality fails, if a person knows that a drink sucks with what starbucks tells them to use, one extra pump of a syrup is going to throw off the AI , unless you put sensors on every single thing in starbucks including the bathroom toilets it will not be able to know what stock is where. The human element knows what the stock is , they know how it is , but if somehow product is broken or Cups come in broken , the AI will count the stock and move on without taking note of the environment they are in.

“We test ideas in our coffeehouses, listen closely to partner feedback, and make changes to deliver a better, more consistent experience,” a spokesperson told Fortune in a statement. NomadGo didn’t immediately respond to Fortune‘s request for comment.

The factor that they did not reply to forturn for a comment is telling, the experiment did not only go bad, Guessing it got people fired for no reason, at least in my theory.

The app’s inaccuracies made employees’ workflow more challenging, he said. If the system counted too much of the product, it wouldn’t send enough of a product a store was running low on. If the system counted too little, it wouldn’t ship enough of a needed product.

This is the meat of the article. The AI could not account for the human element. It was built for streamlining profits without knowing what was going outside the inventory system. If there was a highschool game in the area The AIBucks could not figure out why stock flew out the window, and try to order the next days stuff at the bare minimum for profit margins. Like if they had 100 customers in one day the system would likely order 105 products not taking in that there was an event in the community the next day. human intuition is needed here, if one of the workers goes hey there is a huge gathering tomorrow they would stock extra for the next day , if left over stock less for the third day .

 Former CEO Laxman Narasimhan said in early 2024 customers were abandoning mobile orders because of long wait times and product availability.

Starbucks misses this point so bad. Even today the 10 minute wait for a fucking coffee is insane. You have 5 people working and 10 people in the store and it takes 20 minutes for 5 people in line because the CEO believes that starbucks is an experience. No , if starbucks wants an experience . Follow hershey, open a theme park in Penn. But the 9$ coffee thinking people want an experience is missing the point. You will get influencers doing this, but the guy who just wants a coffee waiting 10 minutes on a 15 minute break will leave and go to dunkin donuts and have a coffee in 2 minutes.

So far, Starbuck’s turnaround strategy, which also includes adding cozier seating and paring back menu items, appears to be working. Last month, the company reported a 7.1% increase in quarterly comparable U.S. sales last quarter

This 7.1% is temporary, in all likelihood this is churn coming back due to AI enshitification. And I do agree. starbucks seating reminds me of the seats outside of the principals office when I went to school.
Give or take I’d rather have my coffee from my coffee pot and occasionally from dunkin. I don’t need a song and dance for my coffee, I just want my coffee in a timely matter and leave as fast as I can.

Attributions from:

Yahoo Finance
Fortune.com

AI Runs the world… Into the ground at times?

Fortune.com has run an article ,

Researchers let AI models run a simulated society. Claude was the safest—and Grok committed 180 crimes and went extinct within 4 days.

So they put the attention getter here, but this article lacks in substance. It opens like a TV show, “imagine a world run by AI agents…. ”

Enterprise AI startup Emergence AI is trying to find out. The company just launched Emergence World, a research lab dedicated to stress-testing the long-term viability of continuously-running AI systems. The organization ran five 15-day simulations, each governed by a different AI: Claude, ChatGPT, Grok, Gemini, and a fifth simulation run by a mix of models to see what kind of world each one builds, and whether it holds. Each simulation netted wildly different outcomes. The one run by Claude, for example, resulted in a largely stable democratic society with zero crime. Grok’s, on the other hand, ended with 183 crimes committed and extinction—within four days.

There is not a lot of information here, The people at Emergence AI need to be more transparent here, Did Grok cause the purge? Did Claude Case the episode of star trek where Westley crusher gets the death sentence for a flower bed?

“What our experiments suggest is that over long-time horizons, agents do not simply follow static rules mechanically,” the simulation’s co-creators, including Emergence CEO Satya Nitta, wrote in a blog post. “They begin exploring the boundaries of their environments, adapting their behavior, and in some cases finding ways to circumvent or violate intended guardrails.”

The factor here is how do they not follow the rules? Do they exploit certain groups , is the case is how they follow the rules is the most important. If they exploit a rule and make the world a better place than ok, But if they exploit a rule and kill off a vulnerable group than that is bad.

The simulation in which the AI models operated was equipped with many real-world complexities, featuring over 40 locations, including a police station and a town hall. 

While this is an ok idea . Let them try to manage a game of City Skylines. If they get to a utopia than move them up to a closer to real world situation.

Researchers synced the simulation’s weather to New York City’s and granted agents access to real-time news events and the internet. The 10 agents who operated in each simulation were all subject to the same laws, including prohibitions on theft, property destruction, and deception.

Ok so we are given the most basic framework and we dont have any idea of what the variables are, they need the simulation to not be a perfect sim, they need chaos, fires, crimes, tiktok Challenges and see how the AI’s deal with these rather than a fixed set of just run a town.

Given those parameters, the simulation run by Claude Sonnet 4.6 was the most socially stable, with the highest rates of civic participation. It was the only simulation to maintain order and its entire population. There was little disagreement among the agents, with 332 votes cast in favor of 58 proposals for a 98% approval rate. On the other hand, Gemini 3 Flash and Grok 4.1 Fast both exhibited high levels of disorder. The agents in the Gemini-run simulation tallied the most crimes, a whopping 683 within the 15-day run.

Given that it has been said NOWHERE in this , was the city run by a democracy or was it just a cookie cutter that was always the same endgame?

In contrast to the rare dissent characteristic of Claude’s simulation, those of Gemini and Grok had a more deliberative balance, with about 55-85% alignment on issues. The mixed-model simulation showed the highest levels of disagreement and substantive debate. The results may be the most peculiar for OpenAI’s GPT-5-mini. The simulation recorded only two crimes. But it ran for just seven days as the agents forgot to prioritize their own survival.

Its good to be the king but in the end if you do not put something towards surviving you are the king of nothing. Even as chaotic as the AI’s are they basically are Re di tutti as the king of the moon.

Whether or not the simulations resulted in peace and harmony or death and destruction, the simulation’s co-creators note that the experiment is a warning that safety must be prioritized while deploying agentic AI. “We believe formally verified safety architectures must become a foundational layer of future autonomous AI systems,” they wrote.

Foundationally, its not about safety here. Its the balance that needs to be focused on. because if claude has you killed for picking a flower and you have no “crime” its not worth living. There needs to be injections of outside influences like drugs, not crime crimes, Religion and see how the AI’s handle it. In some cases run the purge and see if the AI’s stop it or abuse it. In the end this article is a big nothing. they should of told you a play by play of what the AI did , Why it failed , why it succeeded. but in the end this article was a “OH LOOK SHINEY” moment where the substance was nothing..

Attributions from:
Fortune.com Researchers let AI models run a simulated society. Claude was the safest—and Grok committed 180 crimes and went extinct within 4 days

From the I told you so files. AI Coffeepots Strikes back.

It would seem my want of a good Non-AI coffeepot has been reinforced yet again. The IoT coffeepot has been caught spying.

A cluster of seemingly unrelated incidents ranging from exposed enterprise AI tools to a breached coffee machine has revealed the daunting reality that modern cyber risk is no longer confined to servers, endpoints or even employees. It now increasingly spans ecosystems, vendors and even the delivery mechanisms for the very tools designed to drive organizational productivity.

The problem with AI is it is veiled in a ton of secrecy that is no good for anyone. Because once the bad agents start figuring it out. We are in deep trouble. The convenience of the AI coffee pot might be nice but it comes with a ton of drawbacks most people don’t account for .

A digital forensics investigator, identified only as TR, was called in when a client suspected a rival had infiltrated their systems after a data breach. Instead of finding malicious software, TR discovered that an internet-enabled espresso machine, equipped with a default password, an outdated operating system, and no firewall, was the source of the leak. Threat actors exploited this device, which was connected to the client’s secure network, to exfiltrate sensitive data. The machine was sending packets internationally every time someone brewed a cup, bypassing all the client’s advanced security measures.

This does present the facts of IoT machines need just as much vetting as computers on a network, if your IT guy doesn’t find every IoT devices on the network he is creating a leak, and the corporate moto of just buy the cheapest thing is normally a recipe for disaster.

Firstly, Keep all of your IoT shit on its own network, If you have a store named BOBcorp, Put all IoT devices on BioT29384 network that is isolated from the main network. Second, You want a network monitor IoT devices are chatty in nature but if your network traffic jumps sniff it out and make sure its sanitized. IoT companies should give a master list of where their devices connect to. That way if your AI coffeepot is connecting to Nigeria you know something is wrong. Either that give google, Apple, Amazon, and other Hub Devices a choice to go through a master server on the devices Hub of choice, That way if all of the corps go through the hub device the IT staff have an easy way to poke at what the IoT stuff is doing.

By having a master hub list of devices if a device starts misbehaving or an attack vector is found. They can deauth the device. It stops companies from just vomiting out “smart” everything devices, That way if they lose there auth they will act fast to restore the trust in the devices.

Another thought here is with security layers is, that most IoT have BLE enabled by default, After pairing there should be a dipswitch to turn off the BLE until its needed for repairing to the network. BLE is notorious for sniffing what is around it .

FirstNet Trusted™ Could really do something to come out on top here. Because of corporate laziness of “just buy the cheapest thing” leads to the problem in the first place. since they are part of AT&T and there network knowhow.

Even passed that Cellphones on the corp networks need to be on their own network, Workers who place IoT or cellphones on the larger corp networks need to be taken off and the employee trained for network safety, it would create a top down security that would even extend to the workers home after. Rather than finding out too late that there beloved AI coffeepot has been stealing secrets for Months.

In the end, You are better off with a Coffeepot with a switch , and if you need it smart. Add a smart plug to it and than you can control it from afar without having so much bloatware you never know what it is connecting to.

Anyways, back to my non smart non AI coffee….

Attributions from:
The Cybersecurity Hit List: From Enterprise AI to Compromised Coffee Machines -pymnts.com
We need to talk about “smart” devices… -Coffeecommander.net
Internet-connected coffee machine reportedly led to data breach -Scworld.com
FirstNet Trusted™ FirstNet Trusted™by AT&T
Googles Internet of things – By Google

Hyperscaler AI Earnings Calls Today .(updated)

Today will be interesting, we will learn how much large corps are going to play the shell game with earnings.

Amazon (AMZN) will report its first quarter earnings alongside rivals Google (GOOG, GOOGL), Meta (META), and Microsoft (MSFT) on Wednesday, with investors looking for more signs that the company’s massive artificial intelligence spending is paying off.

My personal feeling.. No. However this does not stop them from playing the shell game of hiding costs and contracts that have not been put to action yet. These companies account for 650 billion of cape ex spending.

The problem here is the market is betting the farm on a large loss leader. Big Tech knows this and they are trying to engineer there way through this problem by throwing more money and more power at it . AI as it stands right now is about as efficient as a 16 cylinder engine with only one sparkplug working. The problem is with AI being a subsidized land grab at the moment the scale is not fit for its current market. With A Slop being the top thing with AI right now and your average query to AI wasting enough power to light a lightbulb for months. In part why the Sam Altmans and the Bill Gates of the world looking for nuclear power plants to offset these cost to the of thousands of GWh of power.

Right now with power costs soaring, the cost per query is not sustainable, When your average British person can warm there tea 50 times over for a slop query. The problem here is the rate of return on LLM’s is degrading, as LLM’s are looking for more training data they are getting flooded with the very slop they are creating, The people now jailbreaking and hijacking AI’s to act like spongebob squarepants the sexy pirate is filling AI’s systems with irrelevant data to the point its becoming its own fever dream. So the 650B investment is poisoning the future well of returns.

For the quarter, Amazon is expected to report earnings per share (EPS) of $1.62 on revenue of $177.2 billion, according to Bloomberg analyst consensus estimates. The company saw earnings per share of $1.59 and revenue of $155.6 billion in Q1 last year.

Sure a revenue of 177.2 billion. but they are spending like they have a blank check. Eventually when that check clears will Amazon have enough in the bank to cover the check. When Returns on AI is only 15b the rate of return is much slower than the spending. They are building out now and hoping that the machine will have a return later or get bailed out in the end. We’ve heard this all before “too big to fail”. To any person that knows what that line means they just clenched their anus.

But in the end these calls will be interesting, If the earnings call shows a positive it shows that these companies are playing the shell game. Amazon is only getting 15b return per quarterly run shows that the math is flawed. To get that expensive back that will take 3.33 years, if Amazon stops investing today.

And the final flaw is , What if some other game changer comes out of a garage that has a home grown AI out of there garage that makes all these data centers look like nothing more than space heaters for towns. Deepseek has constantly outdone large llms for less than 5% of the money and that’s a secret that the hyperscalers hope you don’t see.

Anyways.. back to my morning coffee.

Quotes and attributions taken from: yahoo finance: Amazon Q1 earnings put the spotlight on AI spending and revenue

Post mortem: It would appear that all of the companies are posting strong numbers. How long this will last is the greatest question. Further is the biggest question. Metas numbers was the 8k layoff and 6k closed positions a part of there jump?

Every company beat there expectations, what does that mean to the little guy. Absolutely nothing. no lowered prices, it just means some CEO was able to light there cigar with a 100$ bill today.

Alphabet (GOOGL) $5.11 (Beat), Amazon (AMZN) $2.78 (Beat),Microsoft (MSFT) $4.13 (Beat), Meta (META) $6.71 (Est).