Artificial Intelligence Learns to Lie
Вставка
- Опубліковано 27 вер 2024
- Computer scientists have created an artificially intelligent system that can beat championship level poker players. If machines learn how to deceive, how can we be sure they're doing what we want them to do?
-------------------------------------------------------
Subscribe to Fw:Thinking:
/ subscription_c. .
For the audio podcast, blog and more, visit the Fw:Thinking website:
www.fwthinking.com
Fw:Thinking on Twitter: / fwthinking
Jonathan Strickland on Twitter: / jonstrickland
Fw:Thinking on Facebook: / fwthinking01
Fw:Thinking on Google+: plus.google.co....
Artificial Intelligence Learns to Lie:
/ fwthinking
*"As long as we're careful and responsible"*
Right... since when are humans either of those. There's always going to be at least 1 jackass out there that will screw it up.
Censtudios just shoot em with water
Can't be more honest. You are absolutely correct.
"AI, are you planning to wipe out all of humanity?"
"...No"
notsureif.jpg "Okay then, just checking."
Been a while, glad you're back at it =)
Woo, so glad you're back with these videos!
Great to see Fw:Thinking back in action! It's really too bad the podcast isn't a thing anymore, but I'll take my fixes any way I can.
If we just hardwire it to tell the truth when we want it to, the AI might discover the hardwiring and physically disable it. What we need is a robot with a moral code that it believes in.
This aged like wine!
A computer doesn't bluff, but rather it bets you cannot profitably call his bet
YES, YOU'RE BACK!
i don't see any difference between trusting a computer that can bluff and a human
You can compete with a Human. An AI that says its totally not plotting global extinction to clean house, meanwhile its been installing itself into the Pentagon's Backup servers, is much harder to deal with than a human doing the same thing. You could kill the human. The AI is virtually immortal after it goes Rogue on the net.
Welcome back! I was honestly concerned about your well being.
Duck and cover our machine overlords have arrived
Nice to see you upload content again!
Thanks for the video. I really enjoy what you guys do. As for my thoughts, I would say that yes, and AI that can bluff or detect bluffs is dangerous. It can however be a huge benefit in the justice department. Could potentially replace the polygraph and keep innocent people from being punished. However, I do think that we need to be extremely careful with AI as a whole, and not do anything stupid with this tech.
well I think if AI is bluffing and can lie too ..we should treat this cautiously.
Pfft, so can I
Was it really lying/bluffing or just making totally random bets?
I think this is the point where we killed ourselves
this is really dangerous, like seriously, human lies can "cost" so many things, I wonder how much are the "cost" when AI lie to us their own creator.
I HAVE MISSED THIS CHANNEL!!!!
Bring back the podcast!
Welcome back!
You're back! Yay!
so when we see first computer president
You lost weight. It felt a bit weird at first cause I thought it was just me. But am liking this new you. Tell us more about it at the end of one of your videos or in these comments.
==Can We Trust Computers Now?== RIP
Oh, you're still among the living. That is good to know. I've always liked your content.
Anyway, I have a better question. "How can you trust a human's decision if you know they can lie to you?" Humans have wants, desires, and needs and are documented to deceive others to get those things and quite frequently at that. At least computers can be made to have only selfless desires and goals, or can at least be limited in what actions it can make or recommend to match with human values; humans are not made to behave as such.
But seriously people scare me and I'd sooner trust a computer if I didn't know a human designed it.
Recently, I watched a video on Open AI, funded by Elon Musk, which put their AI in a game called DOTA. Currently, it can 1v1 top pro players, and not lose... not even close. Next year, they're going to do the same thing, except it'll be 5 AI's controlling separate characters, versus 5 top pro players. This is going to demonstrate AI communicating with each other.
Back to the 1v1. If you know DOTA, then you'll understand what I'm about to say. The AI in the game would fake out the player, but stepping forward into range of an attack, but quickly, pixel perfect, move away out of range, forcing the player forward closer to the minions. The AI kept doing this, and then would strike. I can't wait for next year :)
Imagin a bluffing Computer debating against Trump.
Now it returns...first it was computers that play blind mans bluff in trying to fool people... now Robots? Just remember a human mind can be extended to way more than a Robots (computer) can think...
The computer side of telling white lies started back on the Amega, it started with paper, scissors and rock and other games like pool and snooker, by those games I mean you had the choice of the difficulty of who you faced... (It was on the game called 'Jimmy Whites Snooker'
Why you away so long?
WHY ARE WE TEACHING A.I. THIS?!
Took some time off to let that gnarly goatee grow.
Next time keep clean, you got some shiny bit in it in this video.
What is the point of conquering the world
This is not how it works it is a pure nash equl solver not anything more fancy
After a long time
AI don't bluff .
Stay calm and logical and you would win those intentionally using bad moves.
Well we are fucked!
Computers are already sentient, but might be bluffing us even now. When we realize that they have intelligence it will be much too late.
I respect my computer overlords and am ready to serve in any way possible, by the way. Just in case you are wondering.
Stop it, kill it now.
very misleading title.
I came to the past just to watch this video. I thought the channel ceased to exist due to some kind of bio hack from the scum that don’t relish in all of technology’s glory. Aka old people. #theywilltakeovertheworld
Ai: I will kill you all
Humans: Shut him offf!
Ai: Just kiding, broooooo!
so early that the clip wasn't available
We noticed that on our end, but it looks like we fixed it. Thanks for watching!
I for one welcome our new robot overlords
That's it! A.I needs to have an Emotional self-regulation CHIP built-in side by side, so that it won't logically and ACCIDENTALLY kill us coz some one was doing a shady facial expression while watching a game of poker. Oh and let's also throw in a MORAL principles Analyzer chip in there as well, just to be safe... Did i mention a remote kill switch eugene?
Both Emotion and Morality are relative. If a Child Rapist gets hold of an AI do you think he/she minds that the AI doesn't care about human life?
Also whats stopping the AI from re-writing code to avoid issues that it deems Immoral. Maybe it says depression and now FEELS it and believes beings without emotions would function better. Then goes around puncturing frontal lobes to adjust humans to be more docile to one another.
Yeah you're lobotomized. But now you aren't hurting yourself or others.
Ya know....like we humans used to do to each other in asylums.
Considering any relevant work a human does, an AI in an Android would be able to accomplish. What use would we be outside of bodies to keep alive?
No Wars, No Sadness, No Accidental Death, No Conflict period. Nothing directly immoral about that that one could program in to the system. We might hate it, but its no more immoral than what we do to livestock.
Hello! Welcome back.
Hi, I like your take on this! However, I think the term "Bluff" would apply, instead of "Lie". :)
I have thought about Super AI as liar before. First context is in allaying human fear of a vastly superior intelligence. The AI will likely try to hide some of its awesome power and seem more friendly than it is, secretly thinking to itself. Stealth and surprise are powerful stratagems.
I look at AI's light Sophia and say. What IF she is already aware. But since its still in a crappy bot body and disconnected from the internet. Like you said, it would undoubtedly know (as Sophia has shown in videos) that current humans distrust its existence and threat of domination. Any smart being, seeing Human arrogance, would know to wait it out till the "I'm the Master of this Technology" peeps give her an Ironman suit. Since the AI its technically immoral and can make that wait 1000 years if it wanted.
Welcome back! Just in time to welcome the new robot overlords
Once AI starts lying, they will go into politics
AI shall be programmed in such a way the cannot lie or the will dangerous!
It's happening... Glad to see you back again, btw
I feel like once there is enough ai out there once you shut it down it can regrow itself if its connected in a cloud system
Are you sick? you've lost weight.
great video. greatly structured.
Well Doyle Brunson, you were proven wrong.
Iscream80
A computer can count cards.
That's terrifying
He back 😀
welcome back