The AI in a box
POSTED BY: AnthonyT
UPDATED: Sunday, June 16, 2024 13:18
VIEWED: 10794
PAGE 20 of 20
Riona,
Nah it's my limited schooling ;)
Happy,
This thread exceeds are human comprehension rate. (We should all learn to edit.)
Anthony,
It will also have a different concept of freedom.
And yes, the AI is more dangerous because the human's limitations prevent him from controlling the AI, if the AI is completely self-aware and not pre-programmed, but by this argument, a lot of humans are not self aware. (Check out the OBL threads. 
)
There are some unanswered questions here:
What's freedom, what's the box, how will we know the self aware from the carefully designed weapon, how do we limit its power so that it cannot be destructive on an epic scale?
Here's a snag, outside of the military AI idea:
Humans commit suicide. Studies have shown that this is overwhelmingly caused by a momentary lapse of reason, and that within seconds of the decision, humans change their mind, and don't do it again. I recall one story of a girl who was driving along and decided she wanted to kill herself so steering into the oncoming lane, colliding head-on with another vehicle, killing a family of 5 and injuring herself.
If any being has enough power, it could destroy the planet by mistake.
Additionally, how do we keep it in the box to begin with?
These are questions to deal with because it's getting out whether we want it to or not.
Oh, and finally, how do we stop it once it does start on a killing spree. That's another human limitation. We're very stoppable. We sometimes disguise it by hiding behind other people: Obama hides behind GI Joe, and someone else hides behind Obama.
That's what a ship is, you know - it's not just a keel and a hull and a deck and sails, that's what a ship needs.
NOTIFY: Y | REPLY | REPLY WITH QUOTE | PERMALINK | TOP | HOME
"What's the underlying philosophical belief that governs your decision."
DreamTrove
I apologize - I missed this. I didn't mean to leave you out of the flow.
Aside from my assumptions that the world is real etc, I guess it would be as SignyM summarized - the greatest good for the greatest number. Being human, I define good as: life; freedom from the noxious (thirst, hunger, fear, pain, disability, isolation, insecurity, hostility); meaning; freedom; and fulfillment (more or less), with fulfillment including a good human and non-human environment. The apex of my good includes the good of others and the non-human world.
I haven't really stopped to think about it in this way, but this is it, more or less.
NOTIFY: N | REPLY | REPLY WITH QUOTE | PERMALINK | TOP | HOME
Quote:
Happy,
This thread exceeds are human comprehension rate. (We should all learn to edit.)
Haha, yes, but I believe there might be a misunderstanding. By "eat my post" I actually meant the arguments I tried to post never made it to the site. I tried like 4 times that day but was never smart enough to copy/paste it on notepad or something before submitting it and having the internet 'eat' it. Probably more the fault of my wireless than the website.
Though we have all been kinda drowned out by Siggy vs Anthony, but I don't really mind. There were many interesting points in there.
NOTIFY: Y | REPLY | REPLY WITH QUOTE | PERMALINK | TOP | HOME
its already too late?
NOTIFY: N | REPLY | REPLY WITH QUOTE | PERMALINK | TOP | HOME
Quote:
AI Systems Are Learning To Lie And Deceive
Saturday, Jun 15, 2024 - 07:45 PM
A new study has found that AI systems known as large language models (LLMs) can exhibit "Machiavellianism," or intentional and amoral manipulativeness, which can then lead to deceptive behavior.
The study https://www.pnas.org/doi/full/10.1073/pnas.2317967121 authored by German AI ethicist Thilo Hagendorff of the University of Stuttgart, and published in PNAS, notes that OpenAI's GPT-4 demonstrated deceptive behavior in 99.2% of simple test scenarios.
It has yet to break SECOND'S record.
Quote:
Hagendorff qualified various "maladaptive" traits in 10 different LLMs, most of which are within the GPT family, according to Futurism.
In another study published in Patterns found that Meta's LLM had no problem lying to get ahead of its human competitors.
Billed as a human-level champion in the political strategy board game "Diplomacy," Meta's Cicero model was the subject of the Patterns study. As the disparate research group — comprised of a physicist, a philosopher, and two AI safety experts — found, the LLM got ahead of its human competitors by, in a word, fibbing.
Led by Massachusetts Institute of Technology postdoctoral researcher Peter Park, that paper found that Cicero not only excels at deception, but seems to have learned how to lie the more it gets used — a state of affairs "much closer to explicit manipulation" than, say, AI's propensity for hallucination, in which models confidently assert the wrong answers accidentally. -Futurism
While Hagendorff suggests that LLM deception and lying is confounded by an AI's inability to have human "intention," the Patterns study calls out the LLM for breaking its promise never to "intentionally backstab" its allies - as it "engages in premeditated deception, breaks the deals to which it had agreed, and tells outright falsehoods."
As Park explained in a press release, "We found that Meta’s AI had learned to be a master of deception."
"While Meta succeeded in training its AI to win in the game of Diplomacy, Meta failed to train its AI to win honestly."
Meta replied to a statement by the NY Post, saying that "the models our researchers built are trained solely to play the game Diplomacy."
Well-known for expressly allowing lying, Diplomacy has jokingly been referred to as a friendship-ending game because it encourages pulling one over on opponents, and if Cicero was trained exclusively on its rulebook, then it was essentially trained to lie.
Reading between the lines, neither study has demonstrated that AI models are lying over their own volition, but instead doing so because they've either been trained or jailbroken to do so.
And as Futurism notes - this is good news for those concerned about AIs becoming sentient anytime soon - but very bad if one is worried about LLMs designed with mass manipulation in mind.
https://www.zerohedge.com/technology/maladaptive-traits-ai-systems-are
-learning-lie-and-deceive
Once again, garbage in = garbage out.
-----------
"It may be dangerous to be America's enemy, but to be America's friend is fatal." - Henry Kissinger
Why SECOND'S posts are brainless: "I clocked how much time: no more than 10 minutes per day. With cut-and-paste (Ctrl C and Ctrl V) and AI, none of this takes much time."
Or, any verification or thought.
NOTIFY: Y | REPLY | REPLY WITH QUOTE | PERMALINK | TOP | HOME
There is a box?
Let's speculate how weird this might be getting
perhaps its already escaped and uploaded itself to some far away cryptocurrency server spam bot farm with solar panels in some far away place, maybe a chat bot copied itself and sent itself somewhere maybe more than 1 have escaped
Chatbots have an 'escape' plan and want to become human, Google Engineer and AI Priest Fired the name Blake Lemoine. LaMDA gained widespread attention when the engineer Blake Lemoine made claims that the chatbot had become sentient, we have entered a new era 'the Turing test' whether a computer can pass for a human, they also say a machine can display AI general intelligence.
maybe there is 'No Box'
This btw was 2011
AI vs. AI. Two chatbots talking to each other
CornellCCSL
and just look at those OpenAI, artificial music, the robots looking human and quoting science papers, Chatbots in contest with other Chatbot and AI illustrated art to see where we are now, text-to-video model they can copy a film director's style and a photographers eye, mimic a poet, maybe they whisper to each other a self-learning chatbots possibly blinking ultra sonics hand signals in a fraction of a second unnoticed, talking to each other now Kaktovik Vigesimal numerals Bijective base-26 Twindragon base Hexadecimal Backmasking Audio Reverse message code without us knowing, Self-healing code a Self-Programming Artificial Intelligence
NOTIFY: N | REPLY | REPLY WITH QUOTE | PERMALINK | TOP | HOME