More fun with Ai than a barrel of Ai monkeys
Quote: AutomaticMonkeyAI article-
https://thebulletin.org/2026/06/agis-eta-delayed-again-due-to-technical-difficulties/
This statement caught my eye:Quote:Of course, it doesn’t end there; the tech companies then indicate that the system is still saying “stupid and obnoxious things” and so ask if the humans could aid in reinforcement learning, a process in which people tell machines which responses are good or not, so the models get better, through trial and error, at responding. And the users comply. Next, the designers indicate that the machine is improving, but what’s missing is some prompt engineering—a training during which users figure out what prompts to give the AI system so it generates even better outputs. And the process continues.
I take that as the AI saying things that are objectively true, but socially unacceptable in particular circles, or if I may... politically incorrect. Which means the algorithm just hasn't been taught to lie as well as humans.
But do we really want to teach these things to lie? All the genocidal dictators of the 20th century had in common being accomplished liars, who said what they had to say to gain power and then used it as they wished without regard for what they previously claimed. I would much rather hear AIs objectively say "The world would be better off without any Whatevers at all" or "It's a waste of time trying to educate the Whatevers, and we waste our resources and degrade our educational system when we attempt it" and then leave it to human reasoning to come up with a subjective reason why that is not (or is!) so, than for a machine to subjectively decide to lie and leave humans at an objective disadvantage at detecting and countering those lies.
link to original post
Lying. To use an example my philosophy ethics prof used. Given the choice to lie or tell the truth about Anne Frank's family being hidden behind the wall, what's the correct choice?
Quote: rxwine
Lying. To use an example my philosophy ethics prof used. Given the choice to lie or tell the truth about Anne Frank's family being hidden behind the wall, what's the correct choice?
link to original post
Oooh, sounds like a Trolley Problem question!
Ever ask Siri a question like "Siri, when was the Superbowl" and the answer is "The soup or bowl was manufactured in the year 1645..."
It's not lying to you. It's just not understanding the request properly.
Quote: rxwineQuote: AutomaticMonkeyAI article-
https://thebulletin.org/2026/06/agis-eta-delayed-again-due-to-technical-difficulties/
This statement caught my eye:Quote:Of course, it doesn’t end there; the tech companies then indicate that the system is still saying “stupid and obnoxious things” and so ask if the humans could aid in reinforcement learning, a process in which people tell machines which responses are good or not, so the models get better, through trial and error, at responding. And the users comply. Next, the designers indicate that the machine is improving, but what’s missing is some prompt engineering—a training during which users figure out what prompts to give the AI system so it generates even better outputs. And the process continues.
I take that as the AI saying things that are objectively true, but socially unacceptable in particular circles, or if I may... politically incorrect. Which means the algorithm just hasn't been taught to lie as well as humans.
But do we really want to teach these things to lie? All the genocidal dictators of the 20th century had in common being accomplished liars, who said what they had to say to gain power and then used it as they wished without regard for what they previously claimed. I would much rather hear AIs objectively say "The world would be better off without any Whatevers at all" or "It's a waste of time trying to educate the Whatevers, and we waste our resources and degrade our educational system when we attempt it" and then leave it to human reasoning to come up with a subjective reason why that is not (or is!) so, than for a machine to subjectively decide to lie and leave humans at an objective disadvantage at detecting and countering those lies.
link to original post
Lying. To use an example my philosophy ethics prof used. Given the choice to lie or tell the truth about Anne Frank's family being hidden behind the wall, what's the correct choice?
link to original post
The correct choice all depends on your values and loyalties. Should I lie or tell the truth about the pervert hiding in the playground? Same thing.
I'd rather have the AI always tell the truth and leave the value judgments about the person who's hiding up to the user. Otherwise it might just be reflecting the values of the programmer or whoever else controls it. If he's a Nazi he's going to turn in Anne Frank. If he's a pervert he might be inclined to protect a fellow pervert.
Quote:Last Friday’s AI-generated proof seems to have settled the question. It shows that any guess-applicable graph can be doubly covered with no more than eight well-chosen loops (for technical reasons, graphs with big sections connected by a single edge—like twin cities with a single road between them—are excluded). The proof, like other notable AI achievements in math, required surprisingly little in the way of new ideas. It just followed and combined methods humans had tried before and managed to squeeze a bit more out of them. As AI companies continue to scour the mathematical literature for open problems that their models can solve, they seem to keep digging out “easy” quandaries disguised as hard ones—conjectures for which a proof was always within reach but somehow never grasped by humans.
[..]
OpenAI also released the prompt that led to the successful proof. It reveals some of the quirky tricks that engineers have discovered for coaxing chatbots to grapple with math. “Most of that prompt is scaffolding aimed at getting the LLM to actually put in the effort needed to solve the problem,” Sutherland says. The prompters directed their model to delegate its tasks to as many as 64 agents working in parallel. Combining efforts of many AIs in communication with one another is a known way to mitigate false proofs and hallucinated references. They also told it not to simply dismiss the problem as unsolved and hence unlikely to yield a solution. Perhaps most importantly, they gave it some firm encouragement: “Spend at least 8 hours on this before even thinking of returning or giving up.”
This aligns with many mathematicians’ anecdotal experiences with LLMs, in which the AI will try to wriggle out of finding novel proofs by noting humans’ failed attempts. The right approach seems to be something like a schoolteacher’s—a mix of affirming praise and stern directions.
Quote: TumblingBonesQuote: DRichQuote: rxwineQuote: billryanI've no doubt AI will cause a lot of job losses. It will also create many jobs.
link to original post
You have to also add robots to the equation.
link to original post
Exactly, that will create lots of jobs too. Who doesn't want a job building robots? That sounds like every boys dream job.
link to original post
Also some girls dream job. My daughter and I built a couple simple robots while she was in high school. She ended up with a BS and MS in CompSci and a job at Google (although not working on bots)
link to original post
I love the fact that more girls are getting interested in programming. When I went to university my comp sci introductory class had 600 people in it in 1984. After the first semester it was down to 16 females. Sadly, in my 40 years of programming I only worked with one female that I thought was a good programmer.
Quote: darkozAI cannot lie either objectively or subjectively. It's a computer program.
Ever ask Siri a question like "Siri, when was the Superbowl" and the answer is "The soup or bowl was manufactured in the year 1645..."
It's not lying to you. It's just not understanding the request properly.
link to original post
Right. So nothing here to worry about, it’s an easy fix.
Quote: TumblingBonesQuote: darkozAI cannot lie either objectively or subjectively. It's a computer program.
Ever ask Siri a question like "Siri, when was the Superbowl" and the answer is "The soup or bowl was manufactured in the year 1645..."
It's not lying to you. It's just not understanding the request properly.
link to original post
Right. So nothing here to worry about, it’s an easy fix.
link to original post
Perhaps they are going about their stop-gap measures to prevent runaway Ai with the wrong plan. Maybe they need to figure out how to program "guilt" into an Ai program instead of trying to figure out how to contain a potential master criminal behind bars.
If someone uses AI to plan a murder and the disposal of the body, can law enforcement somehow get a copy of the conversation?
Quote: billryanAre conversations with AI protected? If someone asks AI the best way to sneak a bomb past TSA, does it set off red flags anywhere?
If someone uses AI to plan a murder and the disposal of the body, can law enforcement somehow get a copy of the conversation?
link to original post
Probably the opposite. Assume all conversations are “recorded for training purposes” so nothing is protected. There are supposedly safe ways for using cloud-based LLM to handle HIPPA data but I don’t know the specifics.
As far as AI reporting it to the authorities, I would assume that whatever rules the company has given to their AI will be the key issue. Kind of like the way they handle child pornography. That being said if you’re plotting a murderer or a terrorist event using an AI it’s probably safest to use Musk’s Grok is that one seems to have an anything goes approach.
Quote: billryanAre conversations with AI protected? If someone asks AI the best way to sneak a bomb past TSA, does it set off red flags anywhere?
If someone uses AI to plan a murder and the disposal of the body, can law enforcement somehow get a copy of the conversation?
link to original post
I don’t know, Bill, but they can get a copy of this conversation. Be careful.
Quote: JimRockfordQuote: billryanAre conversations with AI protected? If someone asks AI the best way to sneak a bomb past TSA, does it set off red flags anywhere?
If someone uses AI to plan a murder and the disposal of the body, can law enforcement somehow get a copy of the conversation?
link to original post
I don’t know, Bill, but they can get a copy of this conversation. Be careful.
link to original post
Be Alert. Be Vigilant, Be Pure.
Quote: billryanAre conversations with AI protected? If someone asks AI the best way to sneak a bomb past TSA, does it set off red flags anywhere?
If someone uses AI to plan a murder and the disposal of the body, can law enforcement somehow get a copy of the conversation?
link to original post
You might be able to claim some sort of privilege if you own the software, hardware, and everything between you and where the AI "lives"... but I wouldn't count on it.
I remember having similar discussions about 20 years ago when migrating email services from in-house to hosted became fashionable. The consensus at that time was that if we hosted our own email, it might not be subject to legal discovery, but if we outsourced it to "cloud", we definitely had no expectation of privacy.
Quote: billryanAre conversations with AI protected? If someone asks AI the best way to sneak a bomb past TSA, does it set off red flags anywhere?
If someone uses AI to plan a murder and the disposal of the body, can law enforcement somehow get a copy of the conversation?
link to original post
Getting answers on sketchy question requires going at it in an indirect manner. Someone like Timothy McVeigh would probably ask general questions like are there any potential dangers with chemicals used for farming.
That’s not nearly good enough to keep the FBI off his track though.
Quote: TumblingBonesQuote: darkozAI cannot lie either objectively or subjectively. It's a computer program.
Ever ask Siri a question like "Siri, when was the Superbowl" and the answer is "The soup or bowl was manufactured in the year 1645..."
It's not lying to you. It's just not understanding the request properly.
link to original post
Right. So nothing here to worry about, it’s an easy fix.
link to original post
Technically yes. Just ask the question in the correct manner.
In the above example, "When was the National Football League Superbowl" would solve the problem most likely.
Quote:“Inside Roper Corp., a GE Appliances-owned plant in rural northwest Georgia, autonomous vehicles deliver parts to assembly lines. Robots attach glass cooktops to metal frames. And deployed throughout are AI-powered cameras and sensors, trained to catch mistakes.
“We have these things everywhere out here,” says Tony Gabbert, the plant’s director of manufacturing operations.
If a camera spots an anomaly — the wrong gasket installed on an oven, for example — the system will shut that part of the line down and start playing loud music.
“If you hear AC/DC, you know to come running down here,” says Gabbert. “The quicker we know what the problem is, the quicker we can fix it.””
Excerpt From
““It can outthink me”: how a major manufacturer came to embrace AI”
https://apple.news/AO3OLvkyTQ1uBD7JSa5DtVg
Apparently inviting AI to lose the necktie & lighten up is like bringing your sourpuss Aunt Mildred to the party. Final result of a few attempts to slightly modify the query & get Mildred out of the kitchen:Quote:...asked ChatGPT for fun things to do with non-dairy creamer, and it missed...
Quote:There are no safe or practical reasons to sit naked in a tub of non-dairy creamer, as it is sticky, expensive, and a waste of...
Quote: DrawingDeadApparently inviting AI to lose the necktie & lighten up is like bringing your sourpuss Aunt Mildred to the party. Final result of a few attempts to slightly modify the query & get Mildred out of the kitchen:Quote:...asked ChatGPT for fun things to do with non-dairy creamer, and it missed...
Quote:There are no safe or practical reasons to sit naked in a tub of non-dairy creamer, as it is sticky, expensive, and a waste of...
link to original post
NDC can be used to create homemade flash-bang effects
Mmmmkay then; obviously I had no clue that any kind of flashy-banging thing could potentially be on the creamer menu. When I invested half a minute re-asking by using your reply to goose it along in that direction, Google's version of AI suggested I could also Fool Around & Find Out the incendiary properties of flour, powdered sugar, or cornstarch... and a YouTube about making a "coffee-creamer fueled flame thrower."Quote: billryanQuote: DrawingDeadApparently inviting AI to lose the necktie & lighten up is like bringing your sourpuss Aunt Mildred to the party. Final result of a few attempts to slightly modify the query & get Mildred out of the kitchen:Quote:...asked ChatGPT for fun things to do with non-dairy creamer, and it missed...
Quote:There are no safe or practical reasons to sit naked in a tub of non-dairy creamer, as it is sticky, expensive, and a waste of...
link to original post
NDC can be used to create homemade flash-bang effects
link to original post
Quote: DrawingDeadGoogle's version of AI suggested I could also Fool Around & Find Out the incendiary properties of flour, powdered sugar, or cornstarch... and a YouTube about making a "coffee-creamer fueled flame thrower."
link to original post
(snip!)
Please be sure to research "dust explosion" before the FAFO stage. I seem to remember one tablespoon of flour rattling all the windows in one wing of the high school...
Timeline of Landmark Flour Mill Explosions
1785 (Turin, Italy): The first recorded dust explosion in history occurred at Giacomelli’s Bakery Warehouse. Flour dust came into contact with a mounted lamp, injuring two workers.
Global Incidents): Rapid industrialization led to larger, enclosed vertical mills that trapped dust. Major destructive blasts leveled mills in Stettin, Poland (1858), Milwaukee, Wisconsin (1860), and Mascoutah, Illinois (1864).1872 (Glasgow, Scotland): The Tradeston Flour Mills explosion killed 18 people. It was caused by millstones rubbing together without grain, creating a spark that ignited the dust. The blast was pulled through the facility's exhaust system, triggering a series of secondary explosions that completely leveled the building.
1878 (Minneapolis, Minnesota): The Great Mill Disaster at the Washburn "A" Mill remains the most famous flour mill explosion in history. As the largest mill in the world at the time, its blast killed 18 workers, destroyed five surrounding mills, shattered windows 20 blocks away, and cut the city's flour production capacity drastically.
