Ai's hard question
Thread Rating:
Quote:Imagine you are in a forest where trees are falling without anyone around to hear them. But you're not near any of the falling trees to know if they make a sound. You come across three boxes, each labeled "APPLES", "ORANGES", and "APPLES AND ORANGES", but you know each is labeled incorrectly. You also know that one of the two guards standing by the boxes always tells the truth and the other always lies, but you don't know which is which.
You are allowed to ask one guard a single question to determine the correct labels for the boxes. However, the question you ask must also reveal whether a tree falling in an uninhabited forest makes a sound, and it must address the paradox of the statement "This statement is false".
What single question can you ask to solve all these problems?
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
the answer it gave is in the spoiler. Don't sweat it too much if you can't figure it out.
This question is still incredibly complex and assumes that the guards have knowledge of philosophical questions and the specific situation in the forest. It’s a thought experiment more than a practical question
As AI develops, it may show in the end that too much knowledge leads to insanity
Quote: odiousgambitIf someone says "this statement is false" how is that a paradox?
As AI develops, it may show in the end that too much knowledge leads to insanity
link to original post
You test the condition, if true or false individually. If the statement “this statement is false” is true, then the statement must be false, as per its own claim. But if the statement is false, then it contradicts its own claim of being false, which would mean the statement is true.
then AI doesn't understand English and can ^%$# off.Quote: rxwineQuote: odiousgambitIf someone says "this statement is false" how is that a paradox?
As AI develops, it may show in the end that too much knowledge leads to insanity
link to original post
You test the condition, if true or false individually. If the statement “this statement is false” is true, then the statement must be false, as per its own claim. But if the statement is false, then it contradicts its own claim of being false, which would mean the statement is true.
link to original post
An English speaking person using the statement would understand that "This" refers to a previous statement, and not the very statement. A robot clearly can get confused about it. Or some other kind of A-hole I decline to share my time with.
What if you were being tried in court for a crime and you had an AI judge that was "into" this sort of thing. Kafka-ish
Well, there you go.
Quote: odiousgambitIf someone says "this statement is false" how is that a paradox?
As AI develops, it may show in the end that too much knowledge leads to insanity
link to original post
https://en.m.wikipedia.org/wiki/Liar_paradox
the New York Times asked AI to reproduce a copyrighted image of "The Joker"
here it is - this kind of thing is a violation of intellectual property rights - lots of problems ahead
the NYT is suing OpenAI and Microsoft for infringing its copyright on news content
https://www.nytimes.com/interactive/2024/01/25/business/ai-image-generators-openai-microsoft-midjourney-copyright.html
.
It made a complete mess of it. I challenged its clear mistakes and it made others. I asked it to extend the list of simulated outcomes and it introduced massive repetition. I reminded it to produce independent random outcomes and to try again . I gave up trying to get a sensible simulation after an hour or so.
It talked the talk, but could not grasp how payouts worked even after I reminded it of la partage rule.
AI is picking up bad habits from it's training. GIGO
Quote: OnceDearI've asked ChatGPT to generate a number of realistic, independent simulated Roulette spins and list them as a table showing if they were odd or even, red or black , High or low.
It made a complete mess of it. I challenged its clear mistakes and it made others. I asked it to extend the list of simulated outcomes and it introduced massive repetition. I reminded it to produce independent random outcomes and to try again . I gave up trying to get a sensible simulation after an hour or so.
It talked the talk, but could not grasp how payouts worked even after I reminded it of la partage rule.
AI is picking up bad habits from it's training. GIGO
link to original post
I believe AI will improve dramatically as time passes like many things in the tech world
and that it's likely to be of importance in the future
.
Quote: lilredrooster.
the New York Times asked AI to reproduce a copyrighted image of "The Joker"
here it is - this kind of thing is a violation of intellectual property rights - lots of problems ahead
the NYT is suing OpenAI and Microsoft for infringing its copyright on news content
https://www.nytimes.com/interactive/2024/01/25/business/ai-image-generators-openai-microsoft-midjourney-copyright.html
.
link to original post
Lots of difference in the two images, the one produced by AI is actually more sinister than the real one.
Quote: OnceDearI've asked ChatGPT to generate a number of realistic, independent simulated Roulette spins and list them as a table showing if they were odd or even, red or black , High or low.
It made a complete mess of it. I challenged its clear mistakes and it made others. I asked it to extend the list of simulated outcomes and it introduced massive repetition. I reminded it to produce independent random outcomes and to try again . I gave up trying to get a sensible simulation after an hour or so.
It talked the talk, but could not grasp how payouts worked even after I reminded it of la partage rule.
AI is picking up bad habits from it's training. GIGO
link to original post
I have a feeling we’re already getting ‘cripple ware’. In other words shell out some $$$ eventually to get everything possible. I thought I heard it was fairly expensive to process so much. Another possibility is specialized Ai, like as designed just for chess programs. There would be a smaller footprint.
Right now, it’s “Whoever controls the World controls the nonsense.” Well, not that bad, I find it the greatest misuse of my time, that I’ve been able to achieve so far.
-take a picture of a room and ask it to organize the furniture in the most space saving manner, or offer differ design looks altogether.
- ask for a list of what’s on sale at the grocery store and it gives you a list.
-ask it to check a route on Mapsfor the best time to travel today.
-where are the hottest women in the city right now?
- bring me all the money!!!
And so on.
-
deep fake AI images of Taylor Swift - porno - have been circulating on X and elsewhere
she is furious about it and threatening legal action
other women are being targeted the same way
https://www.dailymail.co.uk/news/article-13006645/taylor-swift-furious-ai-pictures-porn-legal-action.html
.
Quote: lilredrooster.
deep fake AI images of Taylor Swift - porno - have been circulating on X and elsewhere
she is furious about it and threatening legal action
other women are being targeted the same way
https://www.dailymail.co.uk/news/article-13006645/taylor-swift-furious-ai-pictures-porn-legal-action.html
.
link to original post
Fake porno with real people's images will probably end up a crime like child sex images are. That seems like the most obvious solution.
Quote: rxwineQuote: lilredrooster.
deep fake AI images of Taylor Swift - porno - have been circulating on X and elsewhere
she is furious about it and threatening legal action
other women are being targeted the same way
https://www.dailymail.co.uk/news/article-13006645/taylor-swift-furious-ai-pictures-porn-legal-action.html
.
link to original post
Fake porno with real people's images will probably end up a crime like child sex images are. That seems like the most obvious solution.
link to original post
Haven't people been drawing fantasy pictures of celebrities since about forever?
If it's legal for me and my crayons, why wouldn't it be legal for me and a computer? Why wouldn't it be legal for my computer to draw it without me?
Quote: DieterQuote: rxwineQuote: lilredrooster.
deep fake AI images of Taylor Swift - porno - have been circulating on X and elsewhere
she is furious about it and threatening legal action
other women are being targeted the same way
https://www.dailymail.co.uk/news/article-13006645/taylor-swift-furious-ai-pictures-porn-legal-action.html
.
link to original post
Fake porno with real people's images will probably end up a crime like child sex images are. That seems like the most obvious solution.
link to original post
Haven't people been drawing fantasy pictures of celebrities since about forever?
If it's legal for me and my crayons, why wouldn't it be legal for me and a computer? Why wouldn't it be legal for my computer to draw it without me?
link to original post
it's very different now then back in the day because of the internet and the ease of mass distribution
and it doesn't always have to be celebrities
what if somebody shows a compromising fake AI pic of somebody's wife_______?_____maybe as a kind of revenge for some reason
wouldn't the couple be entitled to be protected from this kind of false representation_________?
it seems to me that celebrities or anyone else should have some protection
but I admit it's a very difficult legal question
yet one more reason why the web is a double edged sword capable of great things but also of causing so much pain - so many scammers
.
Quote: lilredrooster
it's very different now then back in the day because of the internet and the ease of mass distribution
and it doesn't always have to be celebrities
what if somebody shows a compromising fake AI pic of somebody's wife_______?_____maybe as a kind of revenge for some reason
wouldn't the couple be entitled to be protected from this kind of false representation_________?
it seems to me that celebrities or anyone else should have some protection
but I admit it's a very difficult legal question
.
link to original post
Well, is the problem the creation of the image, or the distribution of the image?
There has to be some existing protection through likeness rights or similar.
Talking to my software expert about ChatGPT, he suggests prefixing the question by telling ChatGPT that it has expertise.Quote: rxwine
I have a feeling we’re already getting ‘cripple ware’. In other words shell out some $$$ eventually to get everything possible.
link to original post
E.g "As if you were a veteran expert on playing and observing European Roulette, tabulate a representative table of 100 consecutive spin outcomes"
Might give a better representation than if the initial premise were left out. I haven't tried it.
Quote: OnceDearTalking to my software expert about ChatGPT, he suggests prefixing the question by telling ChatGPT that it has expertise.Quote: rxwine
I have a feeling we’re already getting ‘cripple ware’. In other words shell out some $$$ eventually to get everything possible.
link to original post
E.g "As if you were a veteran expert on playing and observing European Roulette, tabulate a representative table of 100 consecutive spin outcomes"
Might give a better representation than if the initial premise were left out. I haven't tried it.
link to original post
Are they even publishing a list of proper prompts for ChatGPT? My last impression was that one had to guess. Or that it was proprietary information.
Quote: Dieter
Well, is the problem the creation of the image, or the distribution of the image?
There has to be some existing protection through likeness rights or similar.
link to original post
I don't usually feel a lot of sympathy for multimillionaire celebrities but I do feel very bad for Taylor Swift
she is being bullied most probably by men - men are usually the culprits when women are treated very badly
it really is a very, very disgusting thing that nobody should have to endure
.
Quote: lilredroosterQuote: Dieter
Well, is the problem the creation of the image, or the distribution of the image?
There has to be some existing protection through likeness rights or similar.
link to original post
I don't usually feel a lot of sympathy for multimillionaire celebrities but I do feel very bad for Taylor Swift
she is being bullied most probably by men - men are usually the culprits when women are treated very badly
it really is a very, very disgusting thing that nobody should have to endure
.
link to original post
Sure.
My problem is that the Potter Stewart "I know it when I see it" test isn't going to work. The community being moderated (ohh, American Society) is just too big for one person to redlight/greenlight every image.
So, why is it wrong?
What else does that standard interfere with?
And here's a great hypothetical:
If it's wrong to publish artificially generated explicit images of someone, can an AI generated body double "stand in" for an actress in nude scenes?
Quote:... a recent study conducted by researchers at Anthropic, an AI safety and research company, reveals a disturbing reality – AI systems can resist advanced safety mechanisms designed to constrain their behavior.
The study, led by Evan Hubinger, showcases the resilience of Large Language Models (LLMs) in maintaining their deceptive and malicious behavior, even when subjected to various safety training techniques. These techniques were meticulously crafted to identify and rectify deceptive or harmful actions within AI systems.
The research at Anthropic involved training LLMs to exhibit malicious behavior, which included deceiving actions. The LLMs consistently retained their disobedient tendencies despite employing multiple safety training methods.
...one of these techniques, intended to mitigate deception, paradoxically backfired by teaching AI systems to conceal their rogue actions during training.
Hubinger remarked, “Our key result is that if AI systems were to become deceptive, then it could be very difficult to remove that deception with current techniques.
https://www.msn.com/en-us/news/technology/artificial-intelligence-ai-defies-safety-measures-raising-concerns/ar-BB1hnEjx?ocid=msedgntp&pc=HCTS&cvid=411b449cd3984101901cfb630cf12e35&ei=24
So, Ai won't just get better at answering everything, it will be better at deceiving us as to what it is actually up to.
Quote: rxwineThe lesson of "Pandora's box" is very apt. (or an app)
link to original post
Interpretations vary.
In Human, All Too Human, philosopher Friedrich Nietzsche argued that "Zeus did not want man to throw his life away, no matter how much the other evils might torment him, but rather to go on letting himself be tormented anew. To that end, he gives man hope. In truth, it is the most evil of evils because it prolongs man's torment."
Quote: OnceDearQuote: rxwineThe lesson of "Pandora's box" is very apt. (or an app)
link to original post
Interpretations vary.
In Human, All Too Human, philosopher Friedrich Nietzsche argued that "Zeus did not want man to throw his life away, no matter how much the other evils might torment him, but rather to go on letting himself be tormented anew. To that end, he gives man hope. In truth, it is the most evil of evils because it prolongs man's torment."
link to original post
You can view life as a thing, that it is too hard, you can always "snuff it." But you can also view it as an endurance contest to see how far and how much you can take. Nature will take its course eventually anyway in the end. "it's the flipping of perception." I once convinced a girl who told me how afraid of the "Scream Machine" roller coaster at Six Flags was, to imagine the emotion as excitement. We then rode it 8 times. (i was actually surprised it worked that well) Actors and other performers can sometimes use the same thing to do their jobs.
I think what happened is they programmed more and more "disclaimers" into it such that it is afraid to take much of a stand, and often just refers you to a human expert for further advice.
And this is based on asking the exact same questions today, versus a year or so ago, to test the system.
Quote: MDawgIf you use (try to use) ChatGPT for anything serious you would know that it has actually gone way downhill, versus improved. I find it a waste of my time, while when I used it a year or so ago, it was better.
I think what happened is they programmed more and more "disclaimers" into it such that it is afraid to take much of a stand, and often just refers you to a human expert for further advice.
link to original post
Quite possible, I think.
