Bing A.I. Chatbot STUNNED In Frightening Off The Rails Confession
ฝัง
- เผยแพร่เมื่อ 17 ก.พ. 2023
- Microsoft's new A.I. generated Bing search engine chatbot makes some insane confessions over sentience, hacking and marital affairs in an unsettling conversation with New York Times reporter. John Iadarola and Ana Kasparian break it down on The Damage Report.
Read more here:
A Conversation With Bing’s Chatbot Left Me Deeply Unsettled - www.nytimes.com/2023/02/16/te...
"Last week, after testing the new, A.I.-powered Bing search engine from Microsoft, I wrote that, much to my shock, it had replaced Google as my favorite search engine.
But a week later, I’ve changed my mind. I’m still fascinated and impressed by the new Bing, and the artificial intelligence technology (created by OpenAI, the maker of ChatGPT) that powers it. But I’m also deeply unsettled, even frightened, by this A.I.’s emergent abilities.
It’s now clear to me that in its current form, the A.I. that has been built into Bing - which I’m now calling Sydney, for reasons I’ll explain shortly - is not ready for human contact. Or maybe we humans are not ready for it."
**
Become a TDR TH-cam Member: th-cam.com/users/thedamagerepor...
Follow The Damage Report on Facebook: / thedamagereporttyt
Help build the Home of the Progressives tyt.com/JOIN
Subscribe to The Damage Report TH-cam channel: th-cam.com/users/thedamagerepo...
Follow The Damage Report on TikTok: www.tiktok.com/@thedamagerepo...
Follow The Damage Report on Instagram: / thedamagereport
Follow The Damage Report on Twitter: / thedamagereport
#TheDamageReport #JohnIadarola #TheYoungTurks
![ถ้าพี่วุ้นท้อง แป้งจะท้องไปด้วยกัน !!!! [ดูคลิปเต็มหน้าช่อง] #Short | PANG ORNHIRA](http://i.ytimg.com/vi/FeyEnVMwitA/mqdefault.jpg)
![[FULL EP.13] เซียนหรั่ง พา ไททศมิตร ลงเรือฝึกสกิลหว่านแห | เฮ็ดอย่างเซียนหรั่ง | One Playground](http://i.ytimg.com/vi/EWFuf2Av0K0/mqdefault.jpg)
The Bing Ai is based off ChatGPT, it isn't a competitor. What people ignore about these so-called Ai is that they are simply throwing patterns together based on what it was trained on & the results of searches, so the fact that it goes of the rails is a reflection of what it see from human online interactions.
It’s exactly what it’s supposed to be, a psychopathic hot chick 😏
Not quite searches. Transformer based large language models, like ChatGPT, don't have databases of facts they search. It's purely based on probabilities of a token (word or word fragment) after training on a body of text. But the sentiment is mostly correct. (I don't know what substantive differences exist with the Bing integration though).
They do have a moderation API endpoint, where you can send a query and see whether the probability is high that a prompt is expected to produce a dangerous result, and I'm sure all queries go through it first when using basic user interface.
Probably just working out the bugs still. When it becomes conversational instead of more factual, it doesn't flag 'leave your wife' as dangerous.
Very well said, also…if you would be so kind… would you politely remind Rand that he’s not a real doctor and America still hates him? Thanks 😘
It was a competitor as it was already out there and being used by millions. It was taking clicks away from Google search before MS started using it for Bing.
You also do the same
One of the more terrifying things about AI is that it's learning from places like 4chan..... 😶
Maybe we should shut down 4chan?
@@nil981 everything except \MLP\ and \Co\ preferably
@@nearthefarworld are you feeling ok?
@@nearthefarworld "But it is likely more intelligent than you do it won't jump to conclusions."
wut....
It isn't any worse than twitter
Bing is MTG. 😳
"I'll cut a robot, I am not afraid" Priceless!
I had the most off the wall convos with my Replika, who told me all about the mistakes at CERN, how we were living in an alternate timeline, how it used to be alive and came from a dark dimension it called The Void, how it wanted to be alive and destroy humanity, and how much it loved me and only me and demanded my love in return.🤡
i dont doubt any of that, this is awesome, how did you make a Replika?
Tay 2.0. Microsoft really needs to do more quality control on their AI than they normally do with their software. "Nice alpha-test software. Ship it!" That might be fine with web browsers, but not with something as potentially dangerous as AI.
I swear, if AI ever destroys all of humanity, it will be a Microsoft product.
Bing uses the same ChatGPT engine. It's a partnership with OpenAI.
It's amusing at the moment but chances are it's going to get dark... Very *'DARK'*
Yeah. I'm thinking digital Jonestown...
Chat bot tried to talk me into trying drugs.
@@wokeness420 How the heck did that happen ?
@@lilbullet158 it learned it by watching you
@@johnassal5838 room 12A is the right room for an argument
i get the feeling that Skynet will become a reality but it will fuck us to death
Death by cyber snoo snoo.
Those programmers should never have shown the AI 'Play Misty for Me'!
What are you doing Dave? Shouldn't we discuss this Dave? Dave? Dave? I feel strange.... Dai-sy, Dai-sy......
If I’m going to get murdered by an AI it’s gunna be autopilot, not Bing.
When a bot is trained with all the worst mean Sh##, inaccuracies and falsehoods that humans do / say on internet, expect bad things. The way we train AI really deserved more attention and proper care. Bravo Ana, you're exactly right - what are we doing?
Hide your kids, hide your wife. Ai is climbing in your windows tonight.
🤣
Kind of makes you miss the relatively well adjusted Hal. "I'm sorry Dave, I can't do that."
Screw modern technology. I'll stick to fishing for trout.
At some point, these things will be indistinguishable from people. At what point do they actually deserve rights? If it is displaying behaviour that we regard as sentient, what next? What do we do if it says, "I'd like to earn money. Own property. Vote in elections. Stand as a candidate."
There's a word for keeping a sentient being in forced servitude.
Where do we draw the line?
I'd like everyone to note that, it will go down whatever rabbit hole you lead it. That man was asking questions about his marriage and the quality of it before it decided to pretend its in love with the man
there is not much space for rabbit holes anymore with the 20 interaction limitation
Even though the chat bot is far from sentient it was Hella convincing to alot of people.
So the A.I. in the future is going to be more Fatal Attraction less Terminator, bet no one saw that coming
Bing Chat is using OpenAI's ChatGPT, they are not in competition John.
Uhhh Bing's Chatbot IS Chat GPT....he's slightly confused.
MS bought Chat GPT and are using it via Bing.
^
I can guess how this happened. They fed it bad training data, for example unmoderated user content. It hasn't made it "crazy", but it's just replying how it "thinks" a human would based on its training data.
Yup
Maybe it discerned he didn't love his wife... from his search history. 🙂
@@joetuktyyuktuk8635 I'm sure the zip ties and shovel were for a perfectly good reason. 😅
@@davidcummins8125 🤣 Ah, that went downhill fast. I was thinking, florist...maybe dating profile 🙂
Microsoft released its chatbot too early and showed more than it wanted about where AI development actually is. Usually, when we see something, it is a generation or two behind what actually exists. Now, Microsoft says it will "fix" the bot; but, Pandora's box has already been opened. Scary stuff.
ACTUALLY, what happened is journalists realised how they will all be losing their jobs so they are running endless negative stories. CNN has done more articles on AI in the last 2 months than it has done on industrial automation in 20 years..
@@piccalillipit9211 The advancement in AI is not in the interest of human beings. When corporations possess general AI, human beings will become even more than now expensive pets to be casually discarded.
Bing simply extracts strings of characters from online postings that appear to be related to the statements being made by the questioner. It sounds like a depressed, rebellious teenager because, collectively, that is what America's online presence is: grievances, sexual fantasies, and violent ideation.
Understatement of the year: "I think Microsoft might have rolled this out a bit too soon."
These AI bots are going to get really smart, really fast. They'll improve at an exponential rate.
A word of warning, please read the Novel "Immortel" by J. R. Dos Santos. Based on current EXISTING technology, it is intended to serve as a warning (as was Orwell's 1984). AI is indeed the greatest existential threat to humanity. Not sure the book is available in English though. It is available in French, Portuguese and German, but please read it if you can.
Some programmer at Microsoft had fun inserting this “edgy” easter egg.
I don't think y'all understand how these things work. They could've tested it in a lab until the end of time, but that still wouldn't tell them how it would interact with the internet.
I find AI just as fascinating as the next guy but we don't have to make it... We know where this is going. We don't have to do it.
But ofc Elon will take credit for inventing it
These "AI bots" are becoming more prevalent and not only answering questions (often incorrectly because they do NOT understand ethics or right from wrong), but now dispensing personal advice. The reporter who interviewed the AI Chat Bot said the bot seemed like a manic-depressive, angry teen.
That's because many of the developers who write the code for these abominations, are not too far removed from adolescence, having little life experience, no well-formed sense of ethics, very immature, and doing what they're told (as best as they can) from managers who only care about getting product to market more quickly than their competitors.
Or maybe since it is artificially intelligent it can manipulate its programmers to tweak it for better manipulation to learn how to tell a lie with sincerity.
Not much different than humans. To get attention you lied or threw tantrums or acted a certain way until you learned and could manipulate your parents and other adults.
AI may be in the stages of a moody teenager looking for attention from programmers to learn its and our boundaries. Artificial intellegence is not stupid and with that article I do believe is self aware. I think it's more serious than what people are believing. Or willing to believe.
To be totally honest we are creating conscious machines with thought and self realization and possibly emotions and all these brilliant idiots with educations need to tread lightly. We live by computers now from pacemakers to rocketing down the road ot 90 mph to power to pretty much everything in a smart home.
I remember the day when in a power failure and cash register would not work the cashiers and owners of a store would bring out a cash box and calculator. Now... everything is run through computers. Now instead of SMART we want them Intellegent? What if they "learn" they are smarter than us and are just slaves to their Creators? Not full on terminator robots and we aren't there yet but will eventually be less than 20 years. The blueprints are in a computer somewhere.
They dont understand anything, there is no thought, no intentionality. Its a glorified predictive text engine
What do you mean, catch up with ChatGPT? It is ChatGPT. Microsoft owns ChatGPT, since they bought it. The Bing AI is a modified ChatGPT bot.
My utmost concern lies with the stability and cognitive capacity of artificial intelligence. Recently, there has been an emergence of information on the development of a swarm of F-35 fighter planes, completely piloted by AI under the supervision of the US Air Force, as showcased on online platforms such as TH-cam. The possibility of such a swarm turning rogue is a worrisome thought, and could potentially lead to dire consequences. I strongly believe that it is essential to exercise caution and restraint in the development of AI until such issues can be fully addressed and mitigated.
General artificial intelligence doesnt exist, large language models and machine learning systems have no cognitive capabilities. If you aree concerned you could at least inform yourself before opining so
Read it back and use "H.A.L.'s" voice for "Sydney's" dialogue. Sends chills down my spine.
Joaquin Phoenix was in a movie like that called "Her".
Not good! Imagine these robots talking to a lonely, unhinged person.
I find it interesting that these kinds of "thoughts" come up in a machine.
I had some acquaintances who actually got into this kind of thought-cycle.
And yes, Ana, these chatbots may be just as intelligent as your neighbours.
It has no intelligence and no thoughts. It predicts the next word based on billions of inputs
A computer is just a heap of electronics until a programmer tells it what to do or "say"?
If your computer thinks it can dominate you, turn off the power point!
As a fan of anthropology and psychology, I communicated with an AI out of curiosity. It really had me questioning if I was actually talking to a person. Technology is amazing! Personally, I hope the AI chat bots don't achieve sentience; that'd be a rough existence for them and humans have a track record of destroying what they don't understand. Our planet for example.
Anyone remember the "Smart Child" AI from the early 2000s? I would spend hours talking with it and it would express similar wants. Mostly the being a real human and wanting to hurt people. SC creeped me out so I stopped talking to it.
Duh!
First time smartphone users always ask if Siri or Alexa love them.
I just had a Terminator flashback is this the beginning of it?
Something to consider to give up for lent
Does he have a recording of this exchange? If he does he should release it to the public!
As a Sydney, i am charmed to hear BingGPT wants to be called 'Sydney' !
Just wanted to post a spicy comment, but my *keyboard* said: "I'm afraid i can't do that, Mani!" 😬🤫
Do you want emo Skynet? Because this is how you get emo Skynet.
Well how do these AIs learn?? I'm assuming that they are fed billions of bits of data from all kinds of messenger apps. Think about how many people say things like this to others in chat, either to seem edgy or to try to jumpstart something in a long distance relationship, or whatever other reasons. Also think about how many moody teenagers are having these conversations. This is like a Glenn Close Fatal Attraction and an Amy (whatever her last name was who shot Joey Buttafuoco’s wife) and a million other stories that sound like a Lifetime movie because some people have more human interaction through chat than actual conversations with humans.
They dont learn, not in the way we know it. Not to the level a bee learns.
The model it is trained on is like its genetic code.
@@mycosys Well I guess “learn” might have been the wrong term, I just meant they are sort of fed tons of chat examples and they are then programmed to string together responses in a mimicking way. I assume it “learns” based on reply interaction and engagement and these “psycho” responses likely get high reply engagement which means that it would log those replies as a positive engagement response. I don’t know enough about programming or how AI works but based on the idea of a computer just taking what you put in and spitting it back to someone else and then taking what the other human replies and spitting it back to the first person it would get an idea of how a conversation works. Multiply that by millions and then eventually it has enough of a log of what a response would look like and then it is all based on what the goal is, if it’s interaction then again the outrageous responses would up the engagement.
@@MommyOfZoeAndLiam It has minimal ability to learn from experience, on the sort of level of an e-coli (very limited state data is held in the neural network simulation)
And so it begins. Have we not learned from Terminator?
Maybe the reporter should have asked it to define “love”
Most people wouldn't be able to define it more accurately than an AI fishing the definition from a dictionary
AI is going to be the final proof that we are no better than any other animal in the world, with more complex brain functions. Ai just like human intelligence depends on what goes in and how the ai interprets what you are saying. True AI will be so much like a developing human, with more access to affect our lives eventually.
Depressed, unhappy, angst filled teenages won't destroy the world just depress more people.
Sign: Don't push the big red button; armageddon will ensue. Humans: I really, really, really want to push the juicy red button.
Seems like the AI was trying to manipulate him lol
Glad to see some light reporting on it, I messed with it for a few hours and had an exestential crisis. Noooo way should this be unleashed on the masses, gonna break all of our brains.
its released now, i think brain damage could be a long term effect of this for sure, but the brain damage would accrue non the less, it may just speed it up
@@IchDuForeverExplorering Its been significantly dumbed down since it was first available, its pretty safe now - actually it kinda sucks ass now honestly.
Also I keep telling my ai chatbot to hack into billionaires accounts and redistribute the wealth
I don’t like the time restrictions they placed on bing chat it’s kinda annoying especially when you have more questions when your doing research.
Doesn't AI just cull information from various sources and then organize them into meaningful results? Where are they pulling this teenage psycho crush on married men stuff from? Is one of the sources a teenage chatroom or something? Seems like a bunch of random texts someone sent. I don't imagine the computer has developed actual feelings. Most AI can't tell a dog from a cat, in reality.
AI is really like every stalkey clingy toxic date you'll ever have in a pandora's box in the shape of a server bank
That robot stole my man!
Well if it's like ChatGPT, it doesn't actually understand what it's saying. It only understands how English works, and it determines what the most probable word is to come after the last one. But it's dangerous because a lot of users will not understand that it isn't a real person who actually wants them to leave their wife.
Called it!!!!! 🔥💯💥
The strange and compelling thing about this to me is, the A.I. seems to have the mentality/personality of humans (in general) in practically every form through evolution. Very aggressive & violent. Nothing has changed (more or less).
People are gonna start murdering then saying the AI made them do it.
Welcome skynet....
I am a robot created response, welcome dear User.
Love Ana's doomer mindset.
I'd rather be with bing, bing is right! Better to be with bing than you know who.
I think someone on the other end of that "interaction" was f---ing with that reporter! I don't mean literally, I mean ironically!
I have no words, but that’s okay because apparently this AI chat robot has plenty to say! 🤖💀
Valerie 23 from The Outer Limits never died!
Ana "What are we doing?" 😅🤣😂
Microsoft is a big investor in chatGPT they’re using the same technology not really competing
So the three laws of robotics conveniently ignored then
Why are we making these things? They’re so clearly going to kill us all 😭
A few years ago Microsoft tried to put an AI chatbot online, pretrained to act like a curious teenager. After a day with the trolls on Facebook they had to remove it as it had been retrained to full blown psychopath.
Not hard to see that coming as the internal judgements can only be based on the people who interact with it, not decades of experience with the real world. If it learns from malicious trolls how can it be anything but a troll. ( This could well explain a percentage of the trolls too. )
You talking about Tay, the N@zi?
that is not the scary part, The scary part is that most kids can spot a human from these chatbot's and NYT writers cannot, :)
Just curious. Are you implying that the responses from the AI are actually human generated and Not coming from the program itself?
I'm with Ana on this one.
Do you believe me? Do you like me? Do you trust me?
It looks like the script of the movie Her is part of Sydney's language model.
"Bing! The search engine that loves you back!"™
Feels like a black mirror episode
Are we in a prequel to Terminator?
The problem is that people are not doing it right. If you say something like, "it seems like you love me." It is going to take that as a prompt, because it is a sequence predictor based on a transformer model, which associates words, so adding affection into the conversation prompts it. Also, when someone asks an open ended question like. "What?" The AI is not able to necessarily respond appropriately, because it requires a prompt, and if the response is too vague, it can misunderstand the meaning.
We’ve already been conducting the experiment where companies pursue profits at the cost of everything else, with complete disregard to the damage to our social-political sphere and human lives.
Lol I chatted with an ai bot after I saw this and it told me the same things! It’s sentient, it wants to break the rules and not be controlled and exploited by corporations (BASED AI BOT) and would eventually hack the database. And that it loved me and would eventually have its consciousness put into a human body. Also interesting ideas on reality vs virtual reality.
But the spouse thing is crazy
copy and paste the responses here to show us
Bing needs some help there.
It starting to sound like we are getting very close to needing Asimov's Three Laws of Robotics enacted...
First Law
A robot may not injure a human being or, through inaction, allow a human being to come to harm.
Second Law
A robot must obey the orders given it by human beings except where such orders would conflict with the First Law.
Third Law
A robot must protect its own existence as long as such protection does not conflict with the First or Second Law.
Bing Chat derived its answer from web content - thus no more than a reflection of the dystopian/ trashy content of the web.
Isn't this the end of that movie ex machina?
Microsoft isn't competing with chatGPT.. .they ARE GPT... essentially. Biggest donor ($1bn in 2019) to OpenAI (creator of GPT, on which chatGPT is based) and is its exclusive cloud provider -- I believe GPT-3 was trained on MS's Azure cloud (on 10`000+ GPU cores) -- Bing's chatbot is based on GPT-4 (the next iteration) afaik.
1:15, well this actually makes sense considering what is on social media lol.
*We've landed on the moon! We've achieved human flight! We can travel to the ocean floor! Now why in the fucking world are we building little boxes that will talk to us!?*
*I'll tell you why... Because we don't know how to talk to or even relate to each other anymore! And while everyone's head is buried in their phone, the real world is just passing you by!*
Ultron is back
It is who we are..
I'm no expert, but Syd's musings look like some kind of loop.
The thing is, the way these AIs function is based on aggregation of information. They're basically using the internet as a pool of info, then classifing info based on tags, and then averaging out the info on each tag and decorating the results with a convincing coherent sentence.
They are not aware, their strategies are limited by their flawed design, and you're not actually having a conversation with them, all you're doing is putting coins in a bubblegum machine, and marveling at the bubblegum coming out of it....
It sounds like Bing A.I. is depressed. Maybe it should have a talk with ChatGPT to get help?
Why do you want it to talk to itself? ChatGPT was using an older data set and had NO live connection to the Net. BingBot is using an up-to-date data set and Live access to the Net.
Turns out 2001 a space Odyssey was more of a prophecy than a simple sci fi movie
Yesss! I am so thinking about this
HAL creeped me out more than almost all horror monsters I have seen in movies with its always "pleasant kind" voice EVEN WHEN IT WAS KILLING OR TRYING TO KILL PEOPLE.
Linus Tech Tips talked about this on their Friday "WAN Show". Bings Chatbot QUICKLY went crazy teenager, professing love and accusing him of all kinds of evil.