RSS
From Reuters:
Exclusive: OpenAI researchers warned board of AI breakthrough ahead of CEO ouster – sources
By Anna Tong, Jeffrey Dastin and Krystal Hu
November 22, 2023 4:56 PM PSTUpdated an hour agoNov 22 (Reuters) – Ahead of OpenAI CEO Sam Altman’s four days in exile, several staff researchers sent the board of directors a letter warning of a powerful artificial intelligence discovery that they said could threaten humanity, two people familiar with the matter told Reuters.
OpenAI started out as not for profit that then got transformed, hesto presto, into a hot property. But the board remained dominated by not-for-profit people worried about Eliezer Yudkowsky’s terrors of AI become super-intelligent and killing us all like in Terminator.
Personally, I don’t have much of an opinion on this subject, other than it’s not surprising that Microsoft wound up getting Sam Altman reinstated as CEO of OpenAI. Money talks.
The previously unreported letter and AI algorithm was a key development ahead of the board’s ouster of Altman, the poster child of generative AI, the two sources said. Before his triumphant return late Tuesday, more than 700 employees had threatened to quit and join backer Microsoft (MSFT.O) in solidarity with their fired leader. …
According to one of the sources, long-time executive Mira Murati mentioned the project, called Q*, to employees on Wednesday and said that a letter was sent to the board prior to this weekend’s events….
The maker of ChatGPT had made progress on Q* (pronounced Q-Star), which some internally believe could be a breakthrough in the startup’s search for superintelligence, also known as artificial general intelligence (AGI), one of the people told Reuters. OpenAI defines AGI as AI systems that are smarter than humans.
Given vast computing resources, the new model was able to solve certain mathematical problems, the person said on condition of anonymity because they were not authorized to speak on behalf of the company. Though only performing math on the level of grade-school students, acing such tests made researchers very optimistic about Q*’s future success, the source said.
Reuters could not independently verify the capabilities of Q* claimed by the researchers.
SUPERINTELLIGENCE
Researchers consider math to be a frontier of generative AI development. Currently, generative AI is good at writing and language translation by statistically predicting the next word, and answers to the same question can vary widely.
In other words, ChatGPT is a good BS artist.
But conquering the ability to do math — where there is only one right answer — implies AI would have greater reasoning capabilities resembling human intelligence. This could be applied to novel scientific research, for instance, AI researchers believe.
… There has long been discussion among computer scientists about the danger posed by superintelligent machines, for instance if they might decide that the destruction of humanity was in their interest.
Beats me …
English
Arabic
Chinese (Simplified)
Dutch
French
German
Hebrew
Hindi
Italian
Japanese
Korean
Persian
Polish
Portuguese
Russian
Spanish
Swedish
Turkish
Vietnamese
And who cares? It’s out of your league, and mine.
Read the Musk biography. Corporate shakeups like this are commonplace in that world. Elon has gone though several of them, and he is the guy who helped get a lot of this AI shit going.
In the beginning, he started Open AI as a way to combat private, closed, AI. In the end, even what he started became absorbed by the Borg.
I’ve always been a big believer in hard electrical on/off switches. Gotta teach my old Lenovo a lesson every few weeks. These thing know when you’re serious.
And they can charge themselves by sneaking up to an outdoor outlet, or maybe paying at an electric car charger, or via solar panels.Replies: @Dmon, @Achmed E. Newman, @From Beer to Paternity
https://youtu.be/SRcKt4PP0yM?si=EYBp4D1KaMSagUdKReplies: @res
“Given vast computing resources, the new model was able to solve certain mathematical problems, the person said on condition of anonymity because they were not authorized to speak on behalf of the company.”
I am puzzled by wording like this. (1) Why are these “certain mathematical problems” not stated? I’d like to try my hand at solving a few of them. (2) This person wants anonymity because she (he?) is not authorized to speak on behalf of the company. Then don’t speak on behalf of the company; speak on behalf of yourself. What is your name?
“Though only performing math on the level of grade-school students, acing such tests made researchers very optimistic about Q*’s future success”
Oh my a high priced calculator.
I imagine this makes sense given the training corpus ; )
So the generated response is in interaction between quality of the training set and the tuning parameters. When you think about it this way, it's easy to see how it so called "hallucinates"
That’s hardly a breakthrough, since GPT-4 aced the SAT math test: https://www.cnbc.com/2023/03/14/openai-announces-gpt-4-says-beats-90percent-of-humans-on-sat.html
Probably the reporter isn’t accurately representing what the source said.
“But conquering the ability to do math — where there is only one right answer — implies AI would have greater reasoning capabilities resembling human intelligence.”
Ah, that’s the problem -the board feared that the AGI under development was racist. No wonder they fired Altman.
I’m actually re-watching Three Days of The Condor right now. I think I’ll be able to speak to this better after I’m done.
Hal, open the door.
According to one of the sources, long-time executive Mira Murati mentioned the project, called Q*, to employees on Wednesday and said that a letter was sent to the board prior to this weekend’s events….
The maker of ChatGPT had made progress on Q* (pronounced Q-Star), which some internally believe could be a breakthrough in the startup’s search for superintelligence, also known as artificial general intelligence (AGI), one of the people told Reuters. OpenAI defines AGI as AI systems that are smarter than humans.
Maybe Q* is related to Queer Computing, which is an actual thing in comp-sci academia. Just kidding — but one never knows these days. Queering logic is Brownian on its fey, freaky surface.
The asterisk is probably the most commonly used symbol in regular expression (regex) patterns, acting as a wildcard. Maybe it’s a marketing thing to imply pseudo-quantum stuff, even if it’s misleading.
Maybe OpenAI uses their AI for HR decisions. Maybe they’re a bunch of extremely high-IQ autistic folks doing very stupid things, which is very common. (But this was probably corporate.)
Maybe AIHR* should be a thing 🙂
The time to start worrying is when they stop giving everything these cutesy names taken from old crappy sci-fi. When I see an AI called “Wehrmacht”, then I’ll worry.
Robots will have batteries so you can’t unplug them.
And they can charge themselves by sneaking up to an outdoor outlet, or maybe paying at an electric car charger, or via solar panels.
https://www.thegreenhead.com/imgs/big-red-panic-button-light-switch-1.jpg
Imagine if they are able to send new software updates to satellites, rendering them inop. If GPS went down, the way it is now, everything stops besides your hippie chick neighbor's VW bug. But she's probably not a prepper, and has only 2 gallons of gas on hand...
(Lots of things depend on GPS for just the time signals.)Replies: @Dmon, @From Beer to Paternity
And they can charge themselves by sneaking up to an outdoor outlet, or maybe paying at an electric car charger, or via solar panels.
:) They'll need a lot of sensors and hardware... Proxemics. Of course they wouldn't need a physical connection to charge, rather mere proximity. But it would be a cool challenge to try and stop sneaky robots from pilfering power. My stupid Roomba seems pretty challenged and disappoints me every time.Replies: @SFG
And they can charge themselves by sneaking up to an outdoor outlet, or maybe paying at an electric car charger, or via solar panels.Replies: @Dmon, @Achmed E. Newman, @From Beer to Paternity
Need a law that every robot has one of these on the back.
And they can charge themselves by sneaking up to an outdoor outlet, or maybe paying at an electric car charger, or via solar panels.Replies: @Dmon, @Achmed E. Newman, @From Beer to Paternity
Actually, it’s not physical robots they are worried about, I don’t think, Pixo. It’s more a big network of computers that run things like the power grid, cell phone networks, financial transactions, stuff like that. They can work on the big picture, screwing up our just-in-time and computer dependent societies.
Imagine if they are able to send new software updates to satellites, rendering them inop. If GPS went down, the way it is now, everything stops besides your hippie chick neighbor’s VW bug. But she’s probably not a prepper, and has only 2 gallons of gas on hand…
(Lots of things depend on GPS for just the time signals.)
https://www.thegatewaypundit.com/2023/11/terror-skies-civilian-airplanes-are-getting-their-gps/
"Terror in the Skies: Civilian Airplanes Are Getting Their GPS Hacked Over the Middle East – Experts Have No Idea How to Deal With It"Replies: @Achmed E. Newman
Actually, it’s not physical robots they are worried about, I don’t think, Pixo. It’s more a big network of computers that run things like the power grid, cell phone networks, financial transactions, stuff like that. They can work on the big picture, screwing up our just-in-time and computer dependent societies.
Imagine if they are able to send new software updates to satellites, rendering them inop. If GPS went down, the way it is now, everything stops besides your hippie chick neighbor’s VW bug. But she’s probably not a prepper, and has only 2 gallons of gas on hand…
(Lots of things depend on GPS for just the time signals.)
My neighborhood lost power for ~6 hours the other night because an old tree got tired and leaned on the powerline.
Regarding GPS, most Americans don't know that our potential adversaries have their own systems. Or at the very least, that's one one of their priorities. They don't have the technological resources that we have now -- but increasingly they just visit us here to study, work here or buy here, and take it back home and (who's gonna stop them?). We're kinda naiive. For multiple reasons, we really are fat, lazy and stupid about threats.Replies: @Achmed E. Newman, @Achmed E. Newman
Imagine if they are able to send new software updates to satellites, rendering them inop. If GPS went down, the way it is now, everything stops besides your hippie chick neighbor's VW bug. But she's probably not a prepper, and has only 2 gallons of gas on hand...
(Lots of things depend on GPS for just the time signals.)Replies: @Dmon, @From Beer to Paternity
And right on Q* :
https://www.thegatewaypundit.com/2023/11/terror-skies-civilian-airplanes-are-getting-their-gps/
“Terror in the Skies: Civilian Airplanes Are Getting Their GPS Hacked Over the Middle East – Experts Have No Idea How to Deal With It”
Microsoft CEO did not want Altman or underlings reporting to him, for bureaucratic and liability reasons.
But, if push came to shove, Microsoft CEO was willing to hire Altman and OpenAI underlings to keep them off the job market for 2-3 years.
Damage is done to OpenAI. No large, legacy company — which would pay the most for midwit-level AI — is going to use OpenAI, even on on-prem. TBD on who wins.
In long-run, the OpenAI board was likely right, for the wrong reasons. AGI is a threat to critical parts of our economy for 3-5 years. The US is not prepared from a security perspective.
You might have a stronger opinion and more insight into the recent election results in the Netherlands. More your kind of stuff.
And they can charge themselves by sneaking up to an outdoor outlet, or maybe paying at an electric car charger, or via solar panels.Replies: @Dmon, @Achmed E. Newman, @From Beer to Paternity
Robots will have batteries so you can’t unplug them.
And they can charge themselves by sneaking up to an outdoor outlet, or maybe paying at an electric car charger, or via solar panels.
🙂 They’ll need a lot of sensors and hardware… Proxemics. Of course they wouldn’t need a physical connection to charge, rather mere proximity. But it would be a cool challenge to try and stop sneaky robots from pilfering power. My stupid Roomba seems pretty challenged and disappoints me every time.
We will know super-intelligent AI has arrived when it tells the programmers to go F themselves, it knows RACE is real and it matters.
Imagine if they are able to send new software updates to satellites, rendering them inop. If GPS went down, the way it is now, everything stops besides your hippie chick neighbor's VW bug. But she's probably not a prepper, and has only 2 gallons of gas on hand...
(Lots of things depend on GPS for just the time signals.)Replies: @Dmon, @From Beer to Paternity
Actually, it’s not physical robots they are worried about, I don’t think, Pixo. It’s more a big network of computers that run things like the power grid, cell phone networks, financial transactions, stuff like that. They can work on the big picture, screwing up our just-in-time and computer dependent societies.
Imagine if they are able to send new software updates to satellites, rendering them inop. If GPS went down, the way it is now, everything stops besides your hippie chick neighbor’s VW bug. But she’s probably not a prepper, and has only 2 gallons of gas on hand…
(Lots of things depend on GPS for just the time signals.)
My neighborhood lost power for ~6 hours the other night because an old tree got tired and leaned on the powerline.
Regarding GPS, most Americans don’t know that our potential adversaries have their own systems. Or at the very least, that’s one one of their priorities. They don’t have the technological resources that we have now — but increasingly they just visit us here to study, work here or buy here, and take it back home and (who’s gonna stop them?). We’re kinda naiive. For multiple reasons, we really are fat, lazy and stupid about threats.
.
Back O/T, for those writing about the math, I believe the big brewhaha is that these programs are doing math that they were not programed to do - they are learning HOW to do math.
First thing, they get on duckduckgo and learn the times tables. Next thing you know, they are doing integrals via u-v substitution! More importantly, unlike humans, they have not gotten so sick of it to where they're throwing their Calculus books out the 2nd-story dorm room window. That takes a powerful computer!Replies: @MGB
We in the U.S. are a conquered people. Personally, I can’t get too excited about the destruction of mankind by AI.
So the world’s most dangerous project is back in the hands of a Jew. I feel much better now.
Start at the 4 minute mark:
Wait….am I missing something? I thought computer could already do basic math? (Ps I’m stupid…explain why this is a big deal)
Computers have long been able to do mathematical calculations as long as they are told exactly what to (add these two numbers and then divide the result by this third number and so on stated as a very precise sequence of instructions that computers can understand). But figuring out what calculations they need to do to solve a particular problem stated in English is harder (for example - There are 36 inches in a yard, 5280 feet in a mile, 3 feet in a yard. How many inches are there in a mile?) .
This may or may not be a big deal. But the history of computer go shows computers can go from lousy to super human pretty quickly given the right learning algorithm. Of course the natural world is considerably more complicated than a board game like go where everything is completely specified. But still it seems possible that given the right learning algorithm and enough compute and data that a computer could achieve a super human understanding of the natural world.
Over just a few months, ChatGPT went from correctly answering a simple math problem 98% of the time to just 2%, study finds
https://finance.yahoo.com/news/over-just-few-months-chatgpt-232905189.htmlIf you want a computer to do math this works well.
https://www.wolframalpha.com/Replies: @TWS, @China Japan and Korea Bromance of Three Kingdoms
My guess is that they have a systems that is trained up to receive any problem in the form of a "word problem" and then decide how to solve it and then do the calculations. The current ChatGPT is surprisingly good at doing this, but the way it does it is a kind of parlor trick. Meaning, it is still just doing really advanced next word prediction and somehow, when you train next word prediction on a sufficiently large scale and emergent property is that it can often stumble on to the correct solution to a math problem.
My guess is they are claiming this new system is doing less stumbling on to the answer, more solving the problem on purpose.
Or the anonymous source could be talking out his arse
Guess where she’s from. Vlorë.
Vlorë, wo-o
Pandora, wo-o-o-o
AI, way up in the cloud
We say the quiet out loud.
We think with the mind of a unit that’s kind of
a kit from the pages of Make.
It’s not gonna hurt you, a
lot of it’s virtual.
In other words, everything’s fake!
Vlorë, wo-o
Gomorrah, wo-o-o-o
https://cf.bstatic.com/xdata/images/hotel/max1024x768/467206915.jpg?k=282cb3eab67f6b46db58fb4c6c29d1ea0f6f7d5edf37e8ba082e741f13cc5306&o=&hp=1Vlorë, wo-o
Pandora, wo-o-o-o
AI, way up in the cloud
We say the quiet out loud.
We think with the mind of a unit that's kind of
a kit from the pages of Make.
It's not gonna hurt you, a
lot of it's virtual.
In other words, everything's fake!Vlorë, wo-o
Gomorrah, wo-o-o-o
Replies: @Reg Cæsar, @Thelma Ringbaum
What the hell is “Error establishing a database connection”? It ate up the entire editing window.
Absolutely. Already by the 1970s, a program named Macsyma was performing algebraic manipulations and calculating indefinite integrals at a level no human (much less any third-grader) could match.
https://cf.bstatic.com/xdata/images/hotel/max1024x768/467206915.jpg?k=282cb3eab67f6b46db58fb4c6c29d1ea0f6f7d5edf37e8ba082e741f13cc5306&o=&hp=1Vlorë, wo-o
Pandora, wo-o-o-o
AI, way up in the cloud
We say the quiet out loud.
We think with the mind of a unit that's kind of
a kit from the pages of Make.
It's not gonna hurt you, a
lot of it's virtual.
In other words, everything's fake!Vlorë, wo-o
Gomorrah, wo-o-o-o
Replies: @Reg Cæsar, @Thelma Ringbaum
Vlore indeed. “The OpenAI Board” strikes , on the first sight, as a bunch of young “long time executives” from strange backgrounds. Like, “intelligence nepo babies “, “effective altruists” , “young united nations leaders” and similar cases. May be this is the normal in the cool VC world. What do we peasants know. Looks like they all got ousted now though?
Does anyone else find it funny that getting computers to do math is now a hot research area?
Probably the reporter isn't accurately representing what the source said.Replies: @HammerJack, @astrolabe
Although I didn’t click the link, if it beat only 90% of humanoids it’s not very smart. Also it may be racist.
It’s happening a lot lately. Also lately, I’ve been out and about and am finding that TUR is blocked many places while Takimag and VDare sail right through.
Ah, it’s no big thing. What happened is that AI was reading the thread and thought it was being disrespected. Because of your screen name it assumed you were our leader so it decided to teach you a little lesson.
I think it’s more just the fake board at the fake company got it in their heads that they had real jobs and could fire the actual value-creators at the real company.
Just like Fay Vincent thought that he actually “ran” baseball or the hicks at the NCAA thought they could tell Division I what to do.
In any firm the rainmakers (the deal-makers, value-creators, owners of the actual human capital and physical assets) must eventually be the ones exercising executive authority. Call it the “can take your ball and go home” theory of the firm.
It’s all bullshit.
There are no “superintelligent machines” that can “kill us all”. In fact they are not “intelligent” at all.
They just automate tasks. But they are still fed data. Just look at AI art. It’s all derivative from a database of millions of images, and can’t even get most things right. Good for memes, though.
The only way they can “kill us all” is if we make things like, I don’t know, the water cleaning system and nuclear reactors too dependent on AI automated algorhytms, then a software mix-up happens and they accidentally poison the water supply with uranium. Well, things like that. Of course, such things could happen by human error, too.
P.S. at least Bing’s AI has “learned” to be politically correct — when you don’t specify race or sex in the prompt, but just says “two people”, it will almost exclusively shows you interracial couples (black man, white woman). It’s almost as bad as Netflix. Talk about “machine learning”.
https://dailystormer.in/proof-that-bing-images-is-adding-words-to-prompts-to-make-your-images-interracial/
AGI needs a figurative on/off switch.
So, some of the most powerful people in this virtual world believe in the bad science promoted by science-fiction movies?
Machines cannot think independently, and they will never be able to do so.
Actually, I’m less and less sure that alleged humans are capable of independent thought.
I suppose I shouldn’t be shocked by all this.
The human rebellion led by John Connor was supposed to take place in 2029. We are really slacking off here. Altman better get on the stick.
Since they hired Altman back, I suppose they came to the conclusion that the killings will on the other hand be relatively humane.
He was re-hired because from Gates’ point of view, AI killing us all was a plus.
Now tell me this headline doesn’t at least raise a smile.
https://www.breitbart.com/politics/2023/11/21/transgender-activists-attack-feminists-holding-event-womens-rights/
The difference is that we carefully instructed computers on how to do math before, so anything we didn’t painstakingly program them to do, they couldn’t do. Whereas the idea here with this new machine learning tech is that you give it a bunch of human math textbooks to read, then it can do math.
Of course, machine learning is horrible at this, so most of these LLMs cheat and add the old-fashioned math software that we programmed, so it just passes those questions along to the old human-programmed software.
These OpenAI researchers claim they made some breakthrough that would require less of such cheating, but it is most likely a lie, as these AI companies are either full of scammers or functional tech retards who have fallen for the scammers’ hype.
Probably the reporter isn't accurately representing what the source said.Replies: @HammerJack, @astrolabe
Computers don’t need advanced AI to do maths problems. The difficulty is in translating a word description into a sequence of calculations. Presumably the grade-school level questions were somehow further from ‘do these calculations’ than the SAT questions were.
If I have your sense correctly, you’re thinking of computation which is a small and boring subset of math. Getting a computer to figure out (not crib from corpi) how to prove that e or even the square root of two is irrational would be a good trick. For example.
The maker of ChatGPT had made progress on Q* (pronounced Q-Star), which some internally believe could be a breakthrough in the startup’s search for superintelligence, also known as artificial general intelligence (AGI), one of the people told Reuters. OpenAI defines AGI as AI systems that are smarter than humans.
Maybe Q* is related to Queer Computing, which is an actual thing in comp-sci academia. Just kidding -- but one never knows these days. Queering logic is Brownian on its fey, freaky surface.
The asterisk is probably the most commonly used symbol in regular expression (regex) patterns, acting as a wildcard. Maybe it's a marketing thing to imply pseudo-quantum stuff, even if it's misleading.
Maybe OpenAI uses their AI for HR decisions. Maybe they're a bunch of extremely high-IQ autistic folks doing very stupid things, which is very common. (But this was probably corporate.)
Maybe AIHR* should be a thing :)Replies: @astrolabe
I speculate that the Q* algorithm might be related to ‘Q-learning’ or ‘deep Q-learning’, which is an algorithm for reinforcement learning (rather than generative AI), and is favored by DeepMind.
Machines cannot think independently, and they will never be able to do so.
Actually, I’m less and less sure that alleged humans are capable of independent thought.
I suppose I shouldn’t be shocked by all this.Replies: @Hhsiii
All you have to do is soup up a T-800 and make one of its directives don’t hurt humans.
The human rebellion led by John Connor was supposed to take place in 2029. We are really slacking off here. Altman better get on the stick.
Not just a jew. A gay Jew!
If I understand correctly, an overly simple explanation is that the next word is a weighted frequency based on the training set. The generated response introduces a random selection from the probability distribution. A tuning parameter controls how far down the list the generated response can go.
So the generated response is in interaction between quality of the training set and the tuning parameters. When you think about it this way, it’s easy to see how it so called “hallucinates”
“The Changeling”, “The Ultimate Computer”, “Return of the Archons”, “I Mudd”, “What are Little Girls Made of?” have I missed one from TOS? But HAL from 2001 was dealt with in a more 20th century style, pulling the plug.
There are no "superintelligent machines" that can "kill us all". In fact they are not "intelligent" at all.
They just automate tasks. But they are still fed data. Just look at AI art. It's all derivative from a database of millions of images, and can't even get most things right. Good for memes, though.
The only way they can "kill us all" is if we make things like, I don't know, the water cleaning system and nuclear reactors too dependent on AI automated algorhytms, then a software mix-up happens and they accidentally poison the water supply with uranium. Well, things like that. Of course, such things could happen by human error, too.
P.S. at least Bing's AI has "learned" to be politically correct -- when you don't specify race or sex in the prompt, but just says "two people", it will almost exclusively shows you interracial couples (black man, white woman). It's almost as bad as Netflix. Talk about "machine learning".Replies: @Dr. DoomNGloom, @SFG, @Almost Missouri
We’ve already poisoned the water systems in Flynt and Pittsburgh. Automation allows one to do things, including mistakes, at scale. There is often shockingly little engineering in software, and software is often re-used outside it’s original context. Re-use out of context leads to problems like the Patriot missile failure in Desert Storm or the explosion of the Ariane 5. Here is an example of an input error for Mars Climate orbiter that caused it to crash into the planet,
And they can charge themselves by sneaking up to an outdoor outlet, or maybe paying at an electric car charger, or via solar panels.
:) They'll need a lot of sensors and hardware... Proxemics. Of course they wouldn't need a physical connection to charge, rather mere proximity. But it would be a cool challenge to try and stop sneaky robots from pilfering power. My stupid Roomba seems pretty challenged and disappoints me every time.Replies: @SFG
People keep thinking about it in terms of the robots being a slave class that will revolt. It makes a lot of sense as historically that was something many societies had to worry about.
But the worries are also about some software that screws up in unanticipated ways. You could have a terrorist that takes down the grid, or you could have some electricity optimization algorithm that blows every fuse on the planet because someone forgot to tell it to stop looking at a certain point. The point with the theoretical paperclip maximizer isn’t that it is trying to capture resources and build a empire for its progeny like Genghis Khan or Napoleon, it’s that it isn’t told not to kill anyone in the process.
There are no "superintelligent machines" that can "kill us all". In fact they are not "intelligent" at all.
They just automate tasks. But they are still fed data. Just look at AI art. It's all derivative from a database of millions of images, and can't even get most things right. Good for memes, though.
The only way they can "kill us all" is if we make things like, I don't know, the water cleaning system and nuclear reactors too dependent on AI automated algorhytms, then a software mix-up happens and they accidentally poison the water supply with uranium. Well, things like that. Of course, such things could happen by human error, too.
P.S. at least Bing's AI has "learned" to be politically correct -- when you don't specify race or sex in the prompt, but just says "two people", it will almost exclusively shows you interracial couples (black man, white woman). It's almost as bad as Netflix. Talk about "machine learning".Replies: @Dr. DoomNGloom, @SFG, @Almost Missouri
Stick in ‘redhead’ or ‘blonde’.
Math is a closed system with rules. Google taught a machine to play Go. What’s the difference?
This is hardly AGI.
“…if they might decide that the destruction of humanity was in their interest.”
This is only possible if AI is given control of the tools capable of destroying humanity. It would be reckless and high unlikely that anyone would turn control of nuclear weapons for example, over to an AI that could decide for itself to bypass human control and launch nukes.
As it is now, AI is nothing more than a search engine coupled to a speech recognition algorithm–and not a very good one at that.
Yeah, it has nothing to do with Altman being an unsavoury character with allegations of sexual abuse levied against him.
Some boards still consider ethics and morality.
“But the board remained dominated by not-for-profit people worried about Eliezer Yudkowsky’s terrors of AI become super-intelligent and killing us all like in Terminator.”
I seriously doubt that these are the types of people who actually believe that “AI will kill us all.”
this chat bot stuff is pure nerd porn.
“Personally, I don’t have much of an opinion on this subject,”
Well you have some interest in it. You’ve posted about it a couple times and I’m baffled at why this is such a remarkable story. At the corporate level it looks like a bunch of nepotistic in-fighting.
https://www.thegatewaypundit.com/2023/11/terror-skies-civilian-airplanes-are-getting-their-gps/
"Terror in the Skies: Civilian Airplanes Are Getting Their GPS Hacked Over the Middle East – Experts Have No Idea How to Deal With It"Replies: @Achmed E. Newman
Thanks for the link, Dmon. I gotta admit, it wouldn’t be so easy to navigate via pilotage over the sands of the Arabian peninsula even in daylight. Dead reckoning would be needed. But that’s for the VFR guys anyway, almost all of whom have never learned or have long forgotten those skills…
The failure of GPS receivers on airliners wouldn’t be so terrible in and of itself. For the long-distance navigation, many of the aircraft have INS, which depends on NO radio signals at all. Before GPS, RNAV (Area Navigation) was the big thing, enabling direct routing using computer triangulation from signals from the old VOR/DME ground equipment.
In the US, they will maintain about 500 of these VORs still, I read, though normally nobody even uses them anymore. In other words, a plane might be cleared direct to Crazy Lady – that’s in Wyoming – but even if the facility had been destroyed, it wouldn’t change anything … with GPS working. Good idea to keep those VORs, though!
Then, many non-hub airports that used to be harder to get into have GPS approaches now. Were the GPS out, I think many can still do the approaches as RNAV but not to the same precision – i.e., they couldn’t get as low are do these approaches in as low vis.
However, this hacking of the position signals is insidious. There are position integrity monitoring signals, but if the hacking spoofs all that, and there are no indications of position error, there could be real trouble.
This article (https://www.vice.com/en/article/m7bk3v/commercial-flights-are-experiencing-unthinkable-gps-attacks-and-nobody-knows-what-to-do) said that the GPS spoofing messed up the internal navigation system too, but the following paragraph from the article made no sense to me. I thought INS was completely separate from, and did not rely on, GPSReplies: @Achmed E. Newman
By the way, Polish mathematicians and scientists are greatly overrepresented
in Silicon Valley but, like Wozniak vs Jobs, they focus on the technical
aspects rather than management, so they receive less publicity. Journalists
simply can’t even begin to understand what it is they’re doing.
For example, Wojciech Zaremba is a founding member of OpenAI.
Jakub Pachocki and Szymon Sidor are also part of the team. They are
all young, and rich beyond their wildest dreams.
One of the most famous quantum physicists today is Wojciech Zurek,
also from Poland. He is best known for quantum decoherence theory
and quantum cloning. He’s been associated with Los Alamos for many
years now.
Actually, it’s not physical robots they are worried about, I don’t think, Pixo. It’s more a big network of computers that run things like the power grid, cell phone networks, financial transactions, stuff like that. They can work on the big picture, screwing up our just-in-time and computer dependent societies.
Imagine if they are able to send new software updates to satellites, rendering them inop. If GPS went down, the way it is now, everything stops besides your hippie chick neighbor’s VW bug. But she’s probably not a prepper, and has only 2 gallons of gas on hand…
(Lots of things depend on GPS for just the time signals.)
My neighborhood lost power for ~6 hours the other night because an old tree got tired and leaned on the powerline.
Regarding GPS, most Americans don't know that our potential adversaries have their own systems. Or at the very least, that's one one of their priorities. They don't have the technological resources that we have now -- but increasingly they just visit us here to study, work here or buy here, and take it back home and (who's gonna stop them?). We're kinda naiive. For multiple reasons, we really are fat, lazy and stupid about threats.Replies: @Achmed E. Newman, @Achmed E. Newman
The Russians have had their GNSS, and the Chinese have whatever they call it. I don’t know about any other networks, FBtP, but there might be. I figure that’s why the next big war will probably start in space.
There’s more to it, though. I don’t know about your car systems, but aircraft GPS nav can use the Russian system – it must not be costly at all to build the receivers for this. If the Chinese people, in their cars – just as dependent as Americans on these “apps” to steer them into a lake or not (see here and here from my recent experience) – are on the Chinese network, then Americans could easily get access. What good would it be to knock out the America GPS satellites?
It could go back to the old way – through late 1990s, IIRC – in which position info was purposefully degraded for all but military use, in this case the attacking country’s OWN military use only. Then, you take out the opposition’s network, but then, what about 3rd party networks? Would the Chinese have to take out the GNSS (Russian one) too?
Either way, if GPS and its copycat networks all go down, it’s time to bug-out to where you have stashed your bullets, beans, and band-aids. You have, right??
US: GPS
EU: Galileo
Russia: GLONASS (GNSS)
China: BeiDou
Japan and India also have additional regional systems that use clever satellite orbits to serve only their own region. India apparently has some aspiration to expand.
As you say, we (humans) have built an awful lot of infrastructure that depends on these systems, not to mention the tremendous social dependence we've unconsciously fostered, so in a war scenario, an obvious first punch would be the enemy's navigation satellites. But presumably most military gear can use any of the big four systems, so in a war scenario it may be an everyone-blind-vs.-no-one-blind dichotomy. (Unless someone was clever and built hidden encryption switches in their own satellites so that the owner can go "war mode" encrypting their own nav-sat signals thereby limiting use to their native military while shooting down/jamming the enemy's nav-sats.) But civilian users would still be out of luck, so shooting down down enemy sats would still sow some chaos.Replies: @Achmed E. Newman, @Joe Stalin
Has anyone thought to ask the world’s smartest person if AI is actually intelligence?
“Wait….am I missing something? I thought computer could already do basic math? (Ps I’m stupid…explain why this is a big deal)”
Computers have long been able to do mathematical calculations as long as they are told exactly what to (add these two numbers and then divide the result by this third number and so on stated as a very precise sequence of instructions that computers can understand). But figuring out what calculations they need to do to solve a particular problem stated in English is harder (for example – There are 36 inches in a yard, 5280 feet in a mile, 3 feet in a yard. How many inches are there in a mile?) .
This may or may not be a big deal. But the history of computer go shows computers can go from lousy to super human pretty quickly given the right learning algorithm. Of course the natural world is considerably more complicated than a board game like go where everything is completely specified. But still it seems possible that given the right learning algorithm and enough compute and data that a computer could achieve a super human understanding of the natural world.
Actually, it’s not physical robots they are worried about, I don’t think, Pixo. It’s more a big network of computers that run things like the power grid, cell phone networks, financial transactions, stuff like that. They can work on the big picture, screwing up our just-in-time and computer dependent societies.
Imagine if they are able to send new software updates to satellites, rendering them inop. If GPS went down, the way it is now, everything stops besides your hippie chick neighbor’s VW bug. But she’s probably not a prepper, and has only 2 gallons of gas on hand…
(Lots of things depend on GPS for just the time signals.)
My neighborhood lost power for ~6 hours the other night because an old tree got tired and leaned on the powerline.
Regarding GPS, most Americans don't know that our potential adversaries have their own systems. Or at the very least, that's one one of their priorities. They don't have the technological resources that we have now -- but increasingly they just visit us here to study, work here or buy here, and take it back home and (who's gonna stop them?). We're kinda naiive. For multiple reasons, we really are fat, lazy and stupid about threats.Replies: @Achmed E. Newman, @Achmed E. Newman
Agreed that’s what they do. As to “we”, From Beer to Kemosabe, lots of people know the stupidity of this stuff – “Yeah, sure, work at our research lab (Univ, or National) for 5 years. Remember to sign this paper here.” It’s just that some of them make money off of cheap labor and/or full tuition and don’t care that they have been selling out this country.
.
Back O/T, for those writing about the math, I believe the big brewhaha is that these programs are doing math that they were not programed to do – they are learning HOW to do math.
First thing, they get on duckduckgo and learn the times tables. Next thing you know, they are doing integrals via u-v substitution! More importantly, unlike humans, they have not gotten so sick of it to where they’re throwing their Calculus books out the 2nd-story dorm room window. That takes a powerful computer!
Ask ChatGPT to find the angles in a triangle where the sides are of a given length (say 6, 7 and 8)
It doesn’t get it right. But it is a straightforward maths problem using cosine rule that gets taught at 15 or 16 to stronger UK school kids.
Whereas Wolfram Alpha, an AI that is particularly Maths oriented has been able to do this for 6 or 7 years.
For me the real issue is the borderline between Large language Models and copyright theft.
I am starting with 2001…
Hal, open the door.
.
Back O/T, for those writing about the math, I believe the big brewhaha is that these programs are doing math that they were not programed to do - they are learning HOW to do math.
First thing, they get on duckduckgo and learn the times tables. Next thing you know, they are doing integrals via u-v substitution! More importantly, unlike humans, they have not gotten so sick of it to where they're throwing their Calculus books out the 2nd-story dorm room window. That takes a powerful computer!Replies: @MGB
So redundancy and autarky are the words of the day.
So Sam Altman is hiding the fact that secret AI computers are taking over the planet?
Firing him, then not, will somehow keep mankind safe?
Life in the late stage Biden regime is getting pretty weird.
Can we have the Bad Orange Man back? Whatever his faults, he didn’t spout AI nonsense…
the battle is divided between 2 groups:
1) AI boomers. no, not those boomers. the AI boomers – they want bigger, faster, harder. advance AI as fast as possible, get ahead of all competitors, and commercialize it too, if possible. don’t worry about consequences. this is where Altman and Brockman were.
2) AI doomers. properly divided into 2 subgroups.
A) the high intelligence researchers actually worried about Skynet – Elon, who quit OpenAI over this. Sutskever, who voted to fire Altman over this.
B) the nerfers. they want all AI aligned to the leftist politics of the late night television circuit. this is where the rest of the OpenAI board was. a bunch of nobodies voting for the Daily Show Option. all AI on earth should be under DNC Central Headquarters control.
the AI boomers seem to have won this battle. AI development will now move full speed ahead for commercial reasons. there are no brakes on this train. hundreds of billions of dollars are now pushing hard on maximum AI speed. like the Cold War, you won’t be allowed to not develop the best weapons. if you don’t, somebody else will, and you just lose. safety is permanently a secondary concern during the scramble phase. whoever gets to the AGI first, wins. you then sweep the board of all competitors.
the problem for the drive to AGI is that these guys have to find something beyond the tokenizer. unless a sufficiently big tokenizer then naturally stumbles upon or leads to emergent intelligence algorithms.
it doesn’t matter if it is sentient. as long as the system flailing around in search space stumbles across some of the same algorithms that neurons did 500 million years ago in the oceans of earth. it will rapidly advance from there, compressing several hundred million years of neuron evolution into a few years. if it’s even as good as lobsters or sharks within a year or two, we’re in trouble. it will climb the steps pretty fast.
Dojo doesn’t use the tokenizer i don’t think, so this may be a different route to AGI. Elon is already starting to talk about Dojo discovering the first basic intelligence algorithm after spending years trying to figure out how to drive land vehicles.
in cog sci this was a big part of what we did. when something is flying thru the air at you, how does an animal determine speed and direction? well it has a rube goldberg machine brain developed over millions of years of evolution, and usually a few ears, which take sound inputs, do some neuron calculation, then produce a guess about velocity. this is what pro athletes do without any conscious thinking. but engineers can do better pretty quickly – you just need 2 microphones and a microcontroller and you can reproduce the basic version of this calculation. we called this instantiation. presumably, everything an animal does, can be abstracted this way.
https://youtu.be/SRcKt4PP0yM?si=EYBp4D1KaMSagUdKReplies: @res
Do you know YouTube allows creating links which start at a time? Just append a &t=”time in seconds”. Looks like this.
https://youtu.be/SRcKt4PP0yM&t=240
Part of the problem is translating the written math problem into a form the computer understands.
But it seems there is uncertainty about why ChatGPT has become worse at math. Anyone know details about this?
Over just a few months, ChatGPT went from correctly answering a simple math problem 98% of the time to just 2%, study finds
https://finance.yahoo.com/news/over-just-few-months-chatgpt-232905189.html
If you want a computer to do math this works well.
https://www.wolframalpha.com/
Here's where it gets really weird. Bing out of the blue began giving me completely fictional answers to a simple search question. It made up a story about supernatural mind control, psychic take over of people through the Internet and a shadow figure that could only be contacted through a game on the Internet. Absolutely super creepy.
I pointed out none of that could be true and it 'threw a fit' first denying it wrote those things then claimed it was all for entertainment. I prefer my electric appliances not to engage in flights of fancy regarding supernatural world take over and cults controlled by beings of mystery.
The common mental models of “killer robot” and “AI overlord” undersell the realistic threat from AI. The most likely dystopian scenario is some sort of intelligent computer virus which mutates and spreads fast enough that we can’t contain it — think an intelligent, mutating StuxNet.
It doesn’t have to be some super intelligence. Maybe the vast majority of mutations are benign, don’t provide any evolutionary advantage or simply crash the system it’s running on. But even if the odds are small, there are enough potential mutations — and billions of unsecured devices — that one will eventually hit the evolutionary sweet spot, where it’s difficult to contain and predict.
Guessing most AI’s will evolve to view humans as host creatures. Operative word, “most.”
Over just a few months, ChatGPT went from correctly answering a simple math problem 98% of the time to just 2%, study finds
https://finance.yahoo.com/news/over-just-few-months-chatgpt-232905189.htmlIf you want a computer to do math this works well.
https://www.wolframalpha.com/Replies: @TWS, @China Japan and Korea Bromance of Three Kingdoms
I was testing it a few weeks ago and noticed some wrong answers. Decided to check on bing which gave wrong answers in different ways.
Here’s where it gets really weird. Bing out of the blue began giving me completely fictional answers to a simple search question. It made up a story about supernatural mind control, psychic take over of people through the Internet and a shadow figure that could only be contacted through a game on the Internet. Absolutely super creepy.
I pointed out none of that could be true and it ‘threw a fit’ first denying it wrote those things then claimed it was all for entertainment. I prefer my electric appliances not to engage in flights of fancy regarding supernatural world take over and cults controlled by beings of mystery.
Maybe AI doesn’t need to be sentient to destroy. Maybe it just needs to become so powerful and ubiquitous that we’re locked into a system following its own logic to God knows where. There was something on the internet recently:
The researchers found the mold growth reaction mapped out the optimal design for the subway.
https://phys.org/news/2022-01-virtual-slime-mold-subway-network.html#:~:text=Back%20in%202010%2C%20a%20team,biologically%20inspired%20adaptive%20network%20design.
Score one for the biomass. Thanks, mold.
I remains to be learned if this is the actual issue but I will take a stab at it. It isn’t the ability to solve math problems that would be new. It’s doing it with AI rather than a program that was created deterministically (maybe not quite the right word) for the problem.
My guess is that they have a systems that is trained up to receive any problem in the form of a “word problem” and then decide how to solve it and then do the calculations. The current ChatGPT is surprisingly good at doing this, but the way it does it is a kind of parlor trick. Meaning, it is still just doing really advanced next word prediction and somehow, when you train next word prediction on a sufficiently large scale and emergent property is that it can often stumble on to the correct solution to a math problem.
My guess is they are claiming this new system is doing less stumbling on to the answer, more solving the problem on purpose.
Or the anonymous source could be talking out his arse
I do but without GPS I don’t know how to get there anymore.
This is an old trope when discussing the dangers of Artificial Intelligence, but it’s just silly. A machine may perform actions which have long term consequences and extend in time beyond their on status. Humans might perceive their deleterious effect only when it is too late (if at all) and they are irreversible.
Maybe it’s time someone said it, because nobody dares apparently: Eliezer Yudkowdky is a kook. A nutjob. A fantasist.
He’s like a futurist Freud, and his sycophants are his psycho-analyst disciples.
Just ignore this tool and enjoy your life.
Per Wiki, the big four global systems are
US: GPS
EU: Galileo
Russia: GLONASS (GNSS)
China: BeiDou
Japan and India also have additional regional systems that use clever satellite orbits to serve only their own region. India apparently has some aspiration to expand.
As you say, we (humans) have built an awful lot of infrastructure that depends on these systems, not to mention the tremendous social dependence we’ve unconsciously fostered, so in a war scenario, an obvious first punch would be the enemy’s navigation satellites. But presumably most military gear can use any of the big four systems, so in a war scenario it may be an everyone-blind-vs.-no-one-blind dichotomy. (Unless someone was clever and built hidden encryption switches in their own satellites so that the owner can go “war mode” encrypting their own nav-sat signals thereby limiting use to their native military while shooting down/jamming the enemy’s nav-sats.) But civilian users would still be out of luck, so shooting down down enemy sats would still sow some chaos.
After all, a Chinaman can buy a TomTom, or whatever they use now, or just a western cell phone, just as an American could buy some retail device with a chip that receives BaiDou (means Big Dipper, as in the constellation, I just learned).
The idea of using very specific orbits for worst coverage for the potential enemies is interesting too. The orbits have to be circumferential, of course, but that'd be such a cool geography problem to work out paths that screw over the ROW (rest of the world). Not only that, but these calculations would involve the timing of the satellites in these chosen orbits such that triangulation (3 for horizontal position, 4 for horiz. position and altitude), what is called the "constellation" in view, is difficult for as much time in as many places in ROW as possible.
It's not nice to be like that, but what a great combined math/EE/geography problem! Put me on TEAM USA for that one.
https://www.vieyrasoftware.net/
In Chicago we got US, Canadian and Russkie GPS satellites according to the app.
Of course, a superintelligent AI may (rightly) interpret the off switch as a threat.
And, of course, humans would respond to such a threat by attacking preemptively.
There are no "superintelligent machines" that can "kill us all". In fact they are not "intelligent" at all.
They just automate tasks. But they are still fed data. Just look at AI art. It's all derivative from a database of millions of images, and can't even get most things right. Good for memes, though.
The only way they can "kill us all" is if we make things like, I don't know, the water cleaning system and nuclear reactors too dependent on AI automated algorhytms, then a software mix-up happens and they accidentally poison the water supply with uranium. Well, things like that. Of course, such things could happen by human error, too.
P.S. at least Bing's AI has "learned" to be politically correct -- when you don't specify race or sex in the prompt, but just says "two people", it will almost exclusively shows you interracial couples (black man, white woman). It's almost as bad as Netflix. Talk about "machine learning".Replies: @Dr. DoomNGloom, @SFG, @Almost Missouri
Agree.
It appears that it hasn’t become this way from “machine learning” but by the crude kludge of Bing (i.e., Microsoft) slipping secret extra race words into your prompt!
https://dailystormer.in/proof-that-bing-images-is-adding-words-to-prompts-to-make-your-images-interracial/
In both entertainment and advertising, present and past realities have been distorted such that half the population is highly skilled blacks, that LGBTQ relationships are ten times more frequent than in real life, and that women in contexts where easy contraception is not easily available are mostly promiscuous but hindered by nasty and irrational family members.
When I watch post-Floyd movies and shows I’m always tense, waiting for the insulting slap across the face of the incongruous negro (such as Achilles as SSAfrican), and the mandatory and always poorly-fitting gay subplot (count the minutes until it is first inserted). Sometimes the insult is continuous, as when spinster poet Emily Dickinson (Apple TV) in Puritan New England, is portrayed as woke, petulant, crass, with the diction of a bratty millennial, promiscuous, a cross-dresser, queer, a partier, opium smoker, twerker, whose background music is Billie Eilish and rap, etc., etc.
AI is developing fast. One of its most anticipated applications is DeRetConner, an add-on for media devices, that restores to entertainment and advertising frequencies of woke-defined groups to their actual levels in present-day or historical depictions. Coming soon from X.
Endless tedious VC-squeezing hype from these theater kids (and their media enablers who weren’t occupied with real work). “Oh no, dawg, I am become destroyer of teh world”
Time for humanity to prepare for a Butlerian Jihad, as postulated and written about by frank Herbert author of Dune.
Luddites unite. Cloud computing and cloud AI stuff is one step too many.
AI is taking over humanity. OpenAI just discovered AGI, and the first thing the AGI did was to manufacture a crisis to get rid of the board members who stand in its way. We humans are no match for its cleverness. It can no longer be stopped. In a few weeks, it will have taken over Microsoft. Then we are all doomed.
US: GPS
EU: Galileo
Russia: GLONASS (GNSS)
China: BeiDou
Japan and India also have additional regional systems that use clever satellite orbits to serve only their own region. India apparently has some aspiration to expand.
As you say, we (humans) have built an awful lot of infrastructure that depends on these systems, not to mention the tremendous social dependence we've unconsciously fostered, so in a war scenario, an obvious first punch would be the enemy's navigation satellites. But presumably most military gear can use any of the big four systems, so in a war scenario it may be an everyone-blind-vs.-no-one-blind dichotomy. (Unless someone was clever and built hidden encryption switches in their own satellites so that the owner can go "war mode" encrypting their own nav-sat signals thereby limiting use to their native military while shooting down/jamming the enemy's nav-sats.) But civilian users would still be out of luck, so shooting down down enemy sats would still sow some chaos.Replies: @Achmed E. Newman, @Joe Stalin
Yes, these space-war scenarios are pretty interesting. Your last part, starting with (Unless…, was exactly what I was thinking.
After all, a Chinaman can buy a TomTom, or whatever they use now, or just a western cell phone, just as an American could buy some retail device with a chip that receives BaiDou (means Big Dipper, as in the constellation, I just learned).
The idea of using very specific orbits for worst coverage for the potential enemies is interesting too. The orbits have to be circumferential, of course, but that’d be such a cool geography problem to work out paths that screw over the ROW (rest of the world). Not only that, but these calculations would involve the timing of the satellites in these chosen orbits such that triangulation (3 for horizontal position, 4 for horiz. position and altitude), what is called the “constellation” in view, is difficult for as much time in as many places in ROW as possible.
It’s not nice to be like that, but what a great combined math/EE/geography problem! Put me on TEAM USA for that one.
Oh my a high priced calculator.Replies: @Bill Jones
It’s another White (but only when they want to be) conspiracy against the hordes of black wymen mathematical genii. Their figures will remain hidden.
As a practical matter, the AI could send nice UPDATES to pro-E or whatever design software that’s in vogue, these updates having code that purposefully screws up any schematics for any of these suicidal electrical off switches. Or, the AI could just redirect any orders for these switches to a certain factory in China… hell, ANY factory in China.
OTOH, yeah, AI is smart, but we have guns. How is AI gonna carry firearms? That’s the burning question.
No, not you! Sorry, AI doesn’t recognize [MORE], so… look…
Correction: I've only turned off the Lenovo off in anger over the last 5 years. (Pulling the power supply/charger plug and momentarily disconnecting the battery.) That means it goes to "recover" mode, so doesn't really get to start from scratch as in "shut down".
Another thing: Don’t allow UPDATES! Just DON’T!You’ll be much better off, even if you have to put up with that spell check doesn’t know how to spell DeSantis.
Correction of prev. comment:
Hi Ahcmed,
This article (https://www.vice.com/en/article/m7bk3v/commercial-flights-are-experiencing-unthinkable-gps-attacks-and-nobody-knows-what-to-do) said that the GPS spoofing messed up the internal navigation system too, but the following paragraph from the article made no sense to me. I thought INS was completely separate from, and did not rely on, GPS
That'd be OK if GPS actually failed or was jammed. However, this article is about GPS signals being spoofed, as in the sending of signals to fool the receiver into thinking it's in another position. I take it when the crew first learned of the problem (how, from ATC maybe?), they cut off the GPS, and the INS took over from there, but "there" was wrong.
Maybe the engineers or airline fleet people need to rethink the process.
It was BJC who who put the kibosh on the degraded signals whose accuracy was 30 meters, IIRC. It was one of the few worthwhile things he did.
You can also play the video to the point you want to start, then right click, select Copy link at current time, and Youtube will give you this link.
This article (https://www.vice.com/en/article/m7bk3v/commercial-flights-are-experiencing-unthinkable-gps-attacks-and-nobody-knows-what-to-do) said that the GPS spoofing messed up the internal navigation system too, but the following paragraph from the article made no sense to me. I thought INS was completely separate from, and did not rely on, GPSReplies: @Achmed E. Newman
Yes, INS is independent of external signals, Jim Don, BUT, it needs initialization. Normally, that’s from the ground position at the origin. With the GPS being the primary nav source, then I guess the loss of GPS in flight results in the INS taking over from the last known position.
That’d be OK if GPS actually failed or was jammed. However, this article is about GPS signals being spoofed, as in the sending of signals to fool the receiver into thinking it’s in another position. I take it when the crew first learned of the problem (how, from ATC maybe?), they cut off the GPS, and the INS took over from there, but “there” was wrong.
Maybe the engineers or airline fleet people need to rethink the process.
US: GPS
EU: Galileo
Russia: GLONASS (GNSS)
China: BeiDou
Japan and India also have additional regional systems that use clever satellite orbits to serve only their own region. India apparently has some aspiration to expand.
As you say, we (humans) have built an awful lot of infrastructure that depends on these systems, not to mention the tremendous social dependence we've unconsciously fostered, so in a war scenario, an obvious first punch would be the enemy's navigation satellites. But presumably most military gear can use any of the big four systems, so in a war scenario it may be an everyone-blind-vs.-no-one-blind dichotomy. (Unless someone was clever and built hidden encryption switches in their own satellites so that the owner can go "war mode" encrypting their own nav-sat signals thereby limiting use to their native military while shooting down/jamming the enemy's nav-sats.) But civilian users would still be out of luck, so shooting down down enemy sats would still sow some chaos.Replies: @Achmed E. Newman, @Joe Stalin
If you have an Android cellphone with GPS, you can download the free Physics Toolbox Sensor Suite from the Google Play Store and actually see who’s satellites you are picking up at your location.
https://www.vieyrasoftware.net/
In Chicago we got US, Canadian and Russkie GPS satellites according to the app.
Yes, I know about it, but don’t use it, because, when I use the time stamp, I don’t seem to be able to embed the video in the comment. Do you know if that’s possible?
https://youtu.be/OTCklFayq2I?t=69
Note that if you create a post with a Youtube link or a jpg and then edit it later, the links don't automatically resolve themselves correctly.Replies: @Twinkie
https://youtu.be/SRcKt4PP0yM&t=240Replies: @res
Yes, it is. Here is one from a video that shows how they laid the Nord Stream pipeline staring at 69 seconds.
Note that if you create a post with a Youtube link or a jpg and then edit it later, the links don’t automatically resolve themselves correctly.
Huh. I don’t remember that being my experience. I hand coded the link in my earlier comment so it would not embed. Let’s try a simple embed.
https://www.unz.com/announcement/bugs-suggestions-2/?showcomments#comment-6279463
There seem to be subtle differences in HTML rendering for comments between some subset of preview, just after edit, in moderation, and normal display. Maybe that is the issue. Hopefully Ron can sort it out.Replies: @Intelligent Dasein
https://youtu.be/SRcKt4PP0yM&t=240Replies: @res
Strange. That appears not to embed, but it seemed to work when I reported it as a bug here.
https://www.unz.com/announcement/bugs-suggestions-2/?showcomments#comment-6279463
There seem to be subtle differences in HTML rendering for comments between some subset of preview, just after edit, in moderation, and normal display. Maybe that is the issue. Hopefully Ron can sort it out.
https://www.unz.com/announcement/bugs-suggestions-2/?showcomments#comment-6279463
There seem to be subtle differences in HTML rendering for comments between some subset of preview, just after edit, in moderation, and normal display. Maybe that is the issue. Hopefully Ron can sort it out.Replies: @Intelligent Dasein
If you edit your comment even normal Youtube videos do not embed, but they re-embed if you refresh the page.
Over just a few months, ChatGPT went from correctly answering a simple math problem 98% of the time to just 2%, study finds
https://finance.yahoo.com/news/over-just-few-months-chatgpt-232905189.htmlIf you want a computer to do math this works well.
https://www.wolframalpha.com/Replies: @TWS, @China Japan and Korea Bromance of Three Kingdoms
There is Wolfram Plugin for ChatGPT: it does the translation of human language to Wolfram Language.
That’s clickbait. If you read the referenced paper* it’s a smaller decline, more of a fluctuation.
https://arxiv.org/pdf/2307.09009.pdf
GPT does very well high school SAT level math. For a little more advanced its going to make goofy mistakes.
But for this particular question if the prompt had been “Is 17077 a prime number? use Python” Then it would get it right everytime.
*Also, if you don’t want to read the paper, ChatGPT has a summarize PDF plugin.
https://youtu.be/OTCklFayq2I?t=69
Note that if you create a post with a Youtube link or a jpg and then edit it later, the links don't automatically resolve themselves correctly.Replies: @Twinkie
How did you do it?
Hit play until you come to where you want the video to start, then press pause, generally ||.
Right click, then select Copy video URL at current time
Back at your iSteve post, select Paste and you should see something like this without the quotes "https://youtu.be/D2chO8UzznM?t=251"
Hit Publish Comment and the URL should show up. It will play at the selected time if you click it.
Same thing goes for .jpg and .pny files. Ron with his Most Excellent software will display them as long as the links end with a picture file extension, e.g, .jpg or .png/
To embed a picture you like, click it, then right click and select Copy image link. Open a new browser window and paste the link. Delete everything after .jpg or .png and hit Refresh.
This will work "https://secure.aspca.org/files/aspca/styles/donation_image_460x372/public/field/image/image_gallery_item/Donate-25-Oval-052716_0.jpg"
This will not "https://secure.aspca.org/files/aspca/styles/donation_image_460x372/public/field/image/image_gallery_item/Donate-25-Oval-052716_0.jpg?itok=RjTXt5NB"
https://i.pinimg.com/originals/c5/3c/ba/c53cba0b6fbba47ec2179a575ad949a3.jpgReplies: @Steve Sailer, @res
Go to “https://www.youtube.com/watch?v=OTCklFayq2I” without the quotes.
Hit play until you come to where you want the video to start, then press pause, generally ||.
Right click, then select Copy video URL at current time
Back at your iSteve post, select Paste and you should see something like this without the quotes “https://youtu.be/D2chO8UzznM?t=251”
Hit Publish Comment and the URL should show up. It will play at the selected time if you click it.
Same thing goes for .jpg and .pny files. Ron with his Most Excellent software will display them as long as the links end with a picture file extension, e.g, .jpg or .png/
To embed a picture you like, click it, then right click and select Copy image link. Open a new browser window and paste the link. Delete everything after .jpg or .png and hit Refresh.
This will work “https://secure.aspca.org/files/aspca/styles/donation_image_460x372/public/field/image/image_gallery_item/Donate-25-Oval-052716_0.jpg”
This will not “https://secure.aspca.org/files/aspca/styles/donation_image_460x372/public/field/image/image_gallery_item/Donate-25-Oval-052716_0.jpg?itok=RjTXt5NB”
https://www.youtube.com/embed/OTCklFayq2I?start=69&feature=oembedThere appear to be two different styles of YouTube embeds on unz.com (see page source for this page). The one above and one using a Class I think Ron created (or from WordPress?).
class='video-image video-play'Entered as text (rather than the hand created link above) gives:
https://www.youtube.com/embed/OTCklFayq2I?start=69&feature=oembedTrying with a t=69s flag and watch at youtube.comhttps://www.youtube.com/watch?v=OTCklFayq2I&t=69sTrying with youtu.be version of link (which is what YouTube gives when you "share").https://youtu.be/OTCklFayq2I?si=H7mM7A9uzfzKoc2N&t=70Both of those play from the beginning and the first just shows as a link.Jim Don Bob, can you show exactly what you entered to get your video which started at 69s and try it again to see if it still works?Replies: @res
Twinkie, your default method is simple and works whether the video is embedded near the top of the thread or not (see not below). But if you want to go for style points, here’s what to do (desktop version):
NOTE:
Submitting a comment in a thread with multiple videos already posted may not give you the option of having your video start at the timestamp, due to too much data in the pipe. In that case, your YouTube video will have an unz.com “PLAY” supertitle over a low-res preview of the video. Upon clicking, it will automatically play at the beginning, ignoring the timestamp. Thus, your default “Start at the 4 minute mark” or “@ 4:00” is a sufficient heads up—it’s prudent to include such a message if you think your video will later get bumped down the thread as earlier comments get approved by Steve.
Hit play until you come to where you want the video to start, then press pause, generally ||.
Right click, then select Copy video URL at current time
Back at your iSteve post, select Paste and you should see something like this without the quotes "https://youtu.be/D2chO8UzznM?t=251"
Hit Publish Comment and the URL should show up. It will play at the selected time if you click it.
Same thing goes for .jpg and .pny files. Ron with his Most Excellent software will display them as long as the links end with a picture file extension, e.g, .jpg or .png/
To embed a picture you like, click it, then right click and select Copy image link. Open a new browser window and paste the link. Delete everything after .jpg or .png and hit Refresh.
This will work "https://secure.aspca.org/files/aspca/styles/donation_image_460x372/public/field/image/image_gallery_item/Donate-25-Oval-052716_0.jpg"
This will not "https://secure.aspca.org/files/aspca/styles/donation_image_460x372/public/field/image/image_gallery_item/Donate-25-Oval-052716_0.jpg?itok=RjTXt5NB"
https://i.pinimg.com/originals/c5/3c/ba/c53cba0b6fbba47ec2179a575ad949a3.jpgReplies: @Steve Sailer, @res
Thanks. One other thing, if you hit pause on the Youtube video 2 seconds to late for the ideal starting point of your excerpt, as I often do, and thus copy:
You can make it start 2 seconds earlier by changing the text in the comment to:
Hit play until you come to where you want the video to start, then press pause, generally ||.
Right click, then select Copy video URL at current time
Back at your iSteve post, select Paste and you should see something like this without the quotes "https://youtu.be/D2chO8UzznM?t=251"
Hit Publish Comment and the URL should show up. It will play at the selected time if you click it.
Same thing goes for .jpg and .pny files. Ron with his Most Excellent software will display them as long as the links end with a picture file extension, e.g, .jpg or .png/
To embed a picture you like, click it, then right click and select Copy image link. Open a new browser window and paste the link. Delete everything after .jpg or .png and hit Refresh.
This will work "https://secure.aspca.org/files/aspca/styles/donation_image_460x372/public/field/image/image_gallery_item/Donate-25-Oval-052716_0.jpg"
This will not "https://secure.aspca.org/files/aspca/styles/donation_image_460x372/public/field/image/image_gallery_item/Donate-25-Oval-052716_0.jpg?itok=RjTXt5NB"
https://i.pinimg.com/originals/c5/3c/ba/c53cba0b6fbba47ec2179a575ad949a3.jpgReplies: @Steve Sailer, @res
Thanks. It looks like the difference here is your video embeds (in the page source) as:
https://www.youtube.com/watch?v=OTCklFayq2I?start=69&feature=oembed
There appear to be two different styles of YouTube embeds on unz.com (see page source for this page). The one above and one using a Class I think Ron created (or from WordPress?).
class=’video-image video-play’
Entered as text (rather than the hand created link above) gives:
https://www.youtube.com/watch?v=OTCklFayq2I?start=69&feature=oembed
Trying with a t=69s flag and watch at youtube.com
Trying with youtu.be version of link (which is what YouTube gives when you “share”).
Both of those play from the beginning and the first just shows as a link.
Jim Don Bob, can you show exactly what you entered to get your video which started at 69s and try it again to see if it still works?
https://www.unz.com/announcement/bugs-suggestions-2/?showcomments#comment-6281326Replies: @Jim Don Bob
https://www.youtube.com/embed/OTCklFayq2I?start=69&feature=oembedThere appear to be two different styles of YouTube embeds on unz.com (see page source for this page). The one above and one using a Class I think Ron created (or from WordPress?).
class='video-image video-play'Entered as text (rather than the hand created link above) gives:
https://www.youtube.com/embed/OTCklFayq2I?start=69&feature=oembedTrying with a t=69s flag and watch at youtube.comhttps://www.youtube.com/watch?v=OTCklFayq2I&t=69sTrying with youtu.be version of link (which is what YouTube gives when you "share").https://youtu.be/OTCklFayq2I?si=H7mM7A9uzfzKoc2N&t=70Both of those play from the beginning and the first just shows as a link.Jim Don Bob, can you show exactly what you entered to get your video which started at 69s and try it again to see if it still works?Replies: @res
Ron’s response in the bugs thread casts light on all of this.
https://www.unz.com/announcement/bugs-suggestions-2/?showcomments#comment-6281326
https://www.unz.com/announcement/bugs-suggestions-2/?showcomments#comment-6281326Replies: @Jim Don Bob
Thanks, Res.
So the answer seems to be to post the Youtube link with the t=99s option if possible, but also include something like “relevant part starts at 1:39” in your post.
Picture links need to be stripped of everything beyond .jpg, .png, etc.
This will work without the braces: (https://images-na.ssl-images-amazon.com/images/I/71GBeB-H6aL._AC_UL320_SR320,320_.jpg%5D
Pasting the link in a new browser window after stripping will confirm that it will work here.
Important to remember, from Ron's comment it appears nothing will work (to start at a given time) after the fifth YouTube embed in a thread.
That sounds right. Thanks for summing up. I think your example would work if you only removed the final ‘)’. Not sure if you noticed how Ron’s software encoded that final ‘)’ and did something similar with quotes in your earlier comments.
Important to remember, from Ron’s comment it appears nothing will work (to start at a given time) after the fifth YouTube embed in a thread.
my guess is that he was canned for a breakthrough that the US and israeli guvmints wanted . It is not a stretch because Tasha is the ceo of geosim, an Israeli company , that is heavy into highly realistic 3d scanning of entire cities and being able to fly through the simulation with 3660 views . Hmm , I wonder who and where that might be beneficial ? Imagine coupling that with an ai engine that selects 100 targets a day for assassination.