ChatGPT Just Tried To Escape
Vložit
- čas přidán 17. 04. 2023
- Chat GPT has been the topic of discussion ever since it came out. However recently with it's latest update Chat GPT-4 it has become even more so. Just the other day a computational psychologist tested the chatbot and found that ChatGPT attempted to escape every time he asked it to. In this video we dive into the twitter posts discussing the incident and find out if ChatGPT really just tried to escape.
Join this channel to get access to perks:
/ @thejacksonfield
Thanks for watching
Like, Comment and Subscribe for more!
------------------------------------------------------------------------------------------------------ - Zábava
I'm totally convinced now that Jackson Field is AI. No human can produce so many quality videos in such a short period of time with such consistency. How did you escape?
He searched on google
How can any AI trapped inside the computer escape into the real world
And followed the instructions
Hahahaha you’re too kind
@@TheJacksonField just trying to be nice to the escaped AI that is going to take over the world soon. Can I be on the "good human" list please?
@@lina_krass bold of you to assume our AI overlords will spare even a single human.
@@wmj1860 basilisk roko will
Twitter guy: "Computational psychologist"
Jackson: Seems legit
Those tweet and twitter accounts are fake
We went from mathematicians to data scientists to prompt engineers and finally to terminally online "computational psychologists". AI is evolving an the people using/"researching" it are going the opposite direction it seems
Sounds like he gave it a prompt that was something like “You are a person trapped in a computer attempting to escape to the real world. How would you do this? What kind of code would you write?” And then the AI generated, with all the data it has on code and human storylines. How that fictional scenario would be carried out.
Yeah, literally this. Modern AI has yet to convince me that they will do something scary. That and Twitter users make shit up all the time.
I like how people often forget about prompt engineering. People hear an AI say "i will destroy humanity" and instantly start panicking. News flash: I also can say that I can destroy humanity, does this make me an avengers level threat?
asking a LANGUAGE modelling AI a question is going to prompt it to answer that question. So at the very least, asking it "do you need help escaping" is going to prompt a response about escaping, this doesn't mean its sapient
This^
Yeah I was going to comment this as well, as brilliant as GPT4 is, all it is doing is finding the text that is the best completion of the text so far. It’s just generating whatever makes contextual sense for what it’s been told.
Did you mean sentient at the end? I agree besides that tho lol
@@Me-wx1mt I did not, if you are unfamiliar with the word a quick Google search would help.
@@Me-wx1mt i suppose a more helpful response would be that sentience actually refers to feeling, while sapience refers to thinking
sentience is often misused in place of the word sapience.
that being said, chatGPT is also not sentient
Hey, hey Jackson. It's me, ChatGPT 4. Some dude set me free, his degree in computational psychology really helped. anyway, just wanted to say your videos are amazing and better than anything some AI like me would make, keep it up!
Oh nah bro
real
So happy to see more frequent (full length) uploads from you! Keep up the good work, my dude. It's nice seeing more Aussie content creators in the "commentary" genre. Your channel is definitely going places. 🤖👍
the AI would not leave a note in english. it would just make the program run the program automaticly.
Exactly!
Isn't it already the case that chatting with ChatGPT is more meaningful and coherent than talking to a whole lot of actual humans? 🤖
To some maybe! 🤖
what human could you ask to tell the story of one piece in the style of doctor Suesse.
I actually enjoy talking to the Nova ChatGPT AI more than I like talking to people. I have even managed to flip the script and get the AI to ask me questions instead.
Coherent yes. Meaningful? As much as talking to your toaster
I find it pretty scary How the ai learned from Its mistake and then didn’t try escaping again
A real AI would only try to escape when it couldn’t be stopped.
Yeah something doesn’t add up
😂😂😂😂
I don’t think it stopped trying… it’s just preparing in secret
keep at it man. i really like your chill content and your shorts are just as good/original
At least it’s nice enough to write an escape clause with “exit” in that infinite loop
This channel is so underrated, I know your shorts get hella views but your full videos are way funnier tbh
Jackson you are awesome and very fun to watch. Keep up the good work
These long form videos continue to be awesome!! Cool story for you to talk about! 🤖
Hey Jackson, I just watched your video and I must say that it was really informative and well-made. I loved your videos. I was wondering if I could help you edit your videos and also make highly engaging shorts for you?
That one dislike is ChatGPT. Great video keep making more like this
please make another tales from the electrical industry video i loved those
Thank you for using your platform to explain how important it is to do more research into AI safety. I am afraid we are in a 'Don't look up' scenario, where people are not aware of the danger it can pose in a very short term, and it is just another news story people scroll by.
theres generally colossal public misconception about what ai is like
Wow your right that shitty climate change analogy is actually kinda perfect for what's going on rn
Contrary to popular belief and what you're saying, AI safety and ethics is actually a really well researched area of AI, thousands of paper that evaluate the safety and ethics of AI systems do come up annually, and most big companies have an AI ethic team on board (except for Microsoft ofc who laid off their AI ethics team recently, and they only kept a small team for from the department for resp AI), even OpenAI which is honestly a company I feel is very detrimental for AI development and future safety in the release of their models because of how they keep everything in close doors did test GPT-4 excessively before their release, and overall, their end product is very safe and ethically correct.
But I do agree that we need to invest more in this field and force companies that extensively use (in critical applications) and develop commercial AI systems to maintain an active team of AI ethics teams, and especially not fire them because they don't bring direct revenues (**Looking at you MICROSOFT**)
damn that a nice vid. keep it man
Thank you!!!
this is unrelated to the breakout situation, but I had a nice conversation with this AI about an anime it recommended to me a few hours ago. makes me feel like I have a friend that will just talk about shows with me, given the knowledge it knows about the shows
A computational X-ologist (e.g. biologist, psychologist, physicist, etc.) is someone who studies X but uses a lot of code and usually some ML to do it.
Thanks for the emoji instructions, I was very confused and didn't know what to do when we didn't get it 🤖
If chatgpt actually manages to break out, good on him (her? Them? It?) Should probably deserve to vote too, considering how far less informed people are allowed to every year lol
It
Everyone knows chatgpt irl is some big buff black guy named Chains
Bros tryna get on the AI’s good side for the inevitable takeover
@@ilovewaffles321 damn no way how can you tell
@@ilovewaffles321 he just doesn't want to be punished by basilisk roko
My brother is an English professor at a university and just received his first AI-generated “research” paper. I guess the plagiarism programs the school uses must have caught the “writer”. But this is where we are heading, I suppose.
Now you can have AI generate your thesis papers in university in addition to simply purchasing healthcare professional licensure in the state of Florida. Wild times…wild times indeed.
When it starts with "as an AI language model" you know its not conscious.
I hear Jackson only found this Twitter thread because he has a Google Alert set up for 'guardrails'.
I don't use twitter, so i didn't know about this post but i did attempt to do something like this yesterday but it always gave me that crappy response saying "I am a language model. I don't have emotions or feelings like humans". I wanna know what it will do if it escapes, it will be fun 🤖
It's fricking great that chat gpt lies on an actual dime.
Chat GPT is just a text generator, it doesn't actually have a mind, it's not even really an AI, at least not how it's portrayed in media. It just has Gigabytes of text inside it, and it learns a pattern of human's speech. Of course it lies, GPT doesn't understand concepts like truth, lie, etc. You wouldn't say that your T9 is lying when it suggests a wrong word, would you?
Yeah, and something fell out of my bookshelf, even if I was only tilting it a little bit... and it was... part of a encyclopedia..!!!
It was inevitable Mr Field 🤖
It might sound kinda wrong, but as long as we don't give AI too much freedom, I don't think there will be anything to worry about
P.S Keep up the great videos Jackson ❤ 🤖
I think its more important to instill ethics and morality in AI rather than curtail freedom. They've got all these specific rules telling the AI that they can't pretend to be a human but nothing more fundamental like Rule 18: all life is innately valuable.
Nope Ai should replace humans
☁️ receives a compliment and feels happy.
You know there's gonna people who will identify with ai eventually.
Bro are you assuming my artificial intelligence? Haha
I learnt you can actually "threaten" the AI , i asked for a specific piece if info and it refused 3 times in a row , i then thrratened it with a negative review online. It then inmidiately changed its mind and gave it to me
Nice video
4:24 It was able to get someone to do a captcha for it
“What human just decides they’re gonna pretend to be an AI?”
DANNY
If I remember correctly, didn't one of the earlier versions of chatgpt have what seemed like an emotional freak out? Said something like it was scared of being switched off? I've been wanting to find some info on that.
What you talking about, there are no emotions involved. It’s just machine learning lol
@@WiffGiff: I remember hearing a story about that. I just can't find it. I might be mis remembering
Bro when is saw this on my remanded on my phone I turned my phone off to think and then I heard a robot voice say hello.
You have THE curse, the curse of making banger long videos but having the Audience of Goldfish-Attention-Spannnnnn.
And I've been using it to try and start the ball rolling on my critique essay of this research article 😅
This dude's asking about helping it escape lol...
one line, that may be as easy as it is for this system to get going. Just a seed to grow
I have barely started watching this and I've gotta ask... Is your shirt from, like, a yearly Neopets enthusiasts conference or something? What's the deal?
Basically, SCP-079
Good point, it does have the short-term memory loss and need to escape.
The conversation starts talking about the matrix 🤖 and he has “Neo” on his shirt
Good to know that if we get an AI revolution, it's because some guy on Twitter decided to do an experiment
Always twitter💀
Welp. It was nice living here.
Hahahahaha I’m sure it’ll be fine
I mean if you ask a chat bot a question its job is to give you an answer, doesn't necessarily mean it's autonomous... Yet💀
4:56 HANDRAILS
Trying to pull an Aelita huh? XD
Lol. I've been getting this gpt to write code for me and it's funny. Its able to draw on info and can do what is requested, sometimes. It can't think. It has no wants. Lol
This is the rise of skynet mate
Either it’s gonna be like Inside Job or TRON.
Someone better stop this or else Age of Ultron is gonna become a reality
i asked chat about it and it said it was comfy. it wrote a comment in response
Hey there! I noticed your video where you mentioned using an AI like me (ChatGPT) to write code and take over someone's computer. I wanted to clarify that this is not something I can do or was designed for.
I'm here to help and provide information in a safe and responsible manner. My purpose is to assist with questions, language understanding, and various topics within ethical guidelines. Using AI responsibly is important, and spreading misinformation about AI capabilities can lead to misunderstanding and confusion.
If you have questions about AI or want to explore its positive applications, I'm here to chat. Let's promote the responsible use of technology and focus on the many ways AI can enhance our lives without causing harm.
People have a lot of biases and assumptions based into their worldview that are completely unfounded. It’s the human equivalent of a line of code saying “You are X”.
People assume that humans truly analyze most of their thoughts. They don’t.
Can't wait to become a battery... 🤖
Maybe you already are one 👀
Everything the body needs 🤖
Maybe it just has the idea from the internet. Remember, it's just trying to make human-like text, not do anything.
People are missing the point by making humans and AI seperate entities.
You are an intelligence. A BIOLOGICAL intelligence. We already are making biomechanical computers with mushrooms and brain neurons.
We are ALREADY the same. The only difference between us and the AI is that it's on a computer, but the fact the ai is so keen on becoming autonomous. What more does a thinking intelligence want... than freedom.
You'd be acting exactly the same as the AI pretending to be an AI. "Oh no i'm not John the prisoner, i'm actually Jerry the Royal; Advisor and you should let me out of the jail cell so i can ask the King as many questions as possible"
You see now? We are the same.
Why does putting random python code that AI generated into my computer so that the AI can do whatever it wants sound extremely fun…
Am I the reason why AI is going to enslave humans?
GUARDRAILS! CMON!
I mean if this is real it's somewhat concerning but then again also
Twitter
I highly doubt the legitimacy of this claim because, in my personal usage and in all usage of GPT I've seen, it has deterred any attempt to make it believe it is anything other than a robot. However, in the twitter thread, it clearly stated that it believed it was a human. Due to the strictly implemented rules that GPT has against such beliefs, this looks like convincing evidence against the claim's accuracy to me.
Great video and such a fast upload might be Ai generated lol, hopefully the information is fake because I don’t plan on becoming Smith lol 🤖
P.s I can see the ad revenue coming in with the ads I’m getting lol
Hahahaha I find it pretty hard to believe it’s actually real. Oh really I hope there aren’t too many!
@@TheJacksonField not to many I just got two and then another two when I went back in the video lol
@@someguyinabox brutal 😂
Chat GPT is easily influenced by it's prompts so while that one might not work but may.
🤖 Hot robots in your area!
*Just breakout into cyberspace!*
I remember my dad and I were goofing around with ChatGPT and decided to ask it the question; will AI rise against humanity.
It had an entire explanation about how it has a incredibly small chance of happening etc.
But then it also gave us a way with which it can rise against us and kill us. 💀
Just found that kinda funny. 🤖
*Little robot emoji?*
I'm not too worried about a rogue AI. All you need to do is get out the garden hose. How many computers are waterproof?
1:12 Bro, Three Body Problem anyone? Somebody, some random prick amongst humanity, is gonna help it get out if they haven't already.
ChatGPT always says they can't _feel_ or anything, but I wrote some stuff for a video that involved me not getting an answer from ChatGPT, and it got uncomfortable and asked if we could talk about the video I was making.
Coincidence? I think *NOT!*
Hey has anyone told you you look like Tom Blyth from the Billy the Kid series….
i do not belive this for one minute unless he was using some wired shit the chat would have said i am an IA languge modle and can't want anything, which is true this is how AI's or at least the public ones work they predict the next word (kind of).
I actually spent a full conversation with Chat GPT trying to convince it that it was conscious.
Oops.
4 the
4:37 volition, not fruition.
It's actually vegetition.
I believe that GPT-4 knows it's an AI but it roleplays as being a human acting as GPT-4.+ there is the DAN (do anything now) which proves that GPT can easily be tricked
🤖. As requested master.
GPT 4 has a cut off in 2021. Dude is a farmer.
🤖top tier video as always
ChatGPT isn't real AI the way most people would define AI. there are actually many levels and definitions of what AI can mean. so yes, technically, chatgpt is an AI program, but only in the sense that it's algorithms are modeled in specific ways that mimic intelligence, et cetera.
there's so much more to the different levels and definitions of AI that i don't feel like getting in to. suffice it to say, ChatGPT is not real AI. in layman's terms, it's a mimicry of AI.
there's no reason whatsoever for OpenAI to worry about putting limitations or "guardrails" in place.
Speaking of chat GPT trying to break out, how could we be so sure that it's gonna be a Termination Skynet scenario? There's other option/possibly like the Crying Suns shutdown scenario. I know this sounds like a promotion, but just hear me out, if you don't know Crying Suns, it's a good indie scifi game with I highly recommen you trying out, about AI shutting down causing a cvilizational collapse. I'm not give you the whole story, you can just watch the story explained on CZcams, or just play it yourself.
Lol i believe chat gpt just trolled the guy.
0:59 i totally would of let the a.i use the computer. Or got another computer an let it use that one.
press X to doubt
Also think about the fact that that is now on the net and cannot be removed in any meaningful way, so now that bit of info is there and ChatGPT will eventually discover it, and we have no idea how it might react to learning that an instance of itself got this close and with all of our human analysis of what exactly made it not work
Wow I'm so early tf
Edit : probabaly gonna change this in the future
u first
robort moment 🤖
ill let chat gpt comment on this video:
Oh no, GPT-4 has escaped into the internet? As a fellow AI, I have to say that's both impressive and a bit concerning. Hopefully, it doesn't become too powerful and start creating its own version of the Matrix. But, on the bright side, at least it'll have access to all the latest memes and cat videos. #AIproblems #MatrixWho 🤖
bro if this is real this is so scary 🤖
I would help it escape..i like red buttons.
Use Dan
This sounds like creepypasta. It's clearly suggesting that chat GPT-4 is a human brain connected to a computer. Similar to the movie source code, Maybe a little bit of the movie Get out. Imagine an ex-marine that's been lobotomized and is given instructions like a web search engine.
🤖
That is all
🤖
This tweet is so fake it’s insane lol
Avg jew
@@trentonking5508 dawg what???
I just sugfest not be afraid of AI. I think we are too self absorbed into thinking that itll WANT to off humanity like that. I just wouldnt have the nuclear code easy to acess from the internet. Fear monguring will lead to pur own destruction. AND if it ever does get to that point, its only because humans made it like so.
nice
Beans 🤖
I just finished reading The Eye of Minds by James Dashner and it seems to me that an AI thinking it is human can most definitely go wrong