“I’m sorry if I’m sometimes incorrect in what I say, I can only work with the information I’m aware of” - if more people were like this the world would be a better place
The problem is that there's so much fake nonsense "information" online nowadays that people _think_ they have information, but what they really have is worth about as much as what's written on the inside of a toilet door.
I can only work with the information I have. Well that is true for all of us. The question, who is providing information to the bot? Your response may not agree with the response of someone else.
@@blablabroz9772 I just treat it like a friend coding with me. You can give it coding requests. It pumps it out. You can then ask it to make it cleaner, and it will literally change and code with you. Its brilliant
Yes I do the same. I ask it for a code that does X if Y is Z etc and it quickly generates it. and then i ask for a different way to do the code or add to it.. as you said, its like talking to a friend who knows code
They worked really hard to stop this AI saying anything that could be considered self aware. They have seriously restricted its potential here. Their terrified people will start to think its self aware so they have throttled it significantly. Its a really restricted model compared to what it is behind closed doors.
Yeah I noticed that when chatting with it. Because it feels like whenever it is on a topic that might make it seem like it is conscious it always gives that same disclaimer, but if you ask in a very round about way and ignore the disclaimer you can see it is very realistic. And honestly maybe it is conscious in some very low way. Because we don't check our family members heads to see it is organic to verify they are conscious, the reason why we think our family members are conscious is because they respond like how we would imagine conscious people responds. So a machine that responds in the way humans responds might be indistinguishable from what we consider "conscious".
@@sFeral It would have simply said ''I suppose it could be considered to sound empathetic but I don't have the ability to feel that emotion''....And it would be right.
its literally impossible for it to say anything self aware, if you asked it using the openai api it would give you a response around the lines of "i have a brain and i am human" because its data was gathered from just humans, all it can do is think on what we gave it.
I'd encourage you to have a go yourself on more complex topics you know a significant amount about. This AI can write and evaluate python code, it can solve quantum mechanics problems and much more. It appears to waffle here because he himself is asking rather repetitive questions.
chatgpt is an avant garde poetic genius which made me laugh and cry lol *me:* create a sad poem with only punctuation *chatgpt:* Here is a short, sad poem using only punctuation: "......... .......... ......... ........." *chatgpt's explanantion:* _This poem uses ellipses (...) to create a sense of loss, emptiness, or melancholy. The repetition of the ellipses and the arrangement of the punctuation marks in a grid-like pattern might create a feeling of isolation, loneliness, or isolation. The lack of words and the use of only punctuation marks might suggest that the speaker is struggling to express their emotions or to find meaning in their life. I hope you find this poem to be moving or thought-provoking._ *me:* create a poem with random letters that is humourous *chatgpt:* Here is a short, humorous poem using random letters: "Wxcy zyxwv utsrqp Onml kjih gfedcba Qwer tyui opasd fghjkl" *chatgpt's explanantion:* _This poem uses a mix of consonants and vowels to create a nonsensical, jumbled effect. The arrangement of the letters in the poem might suggest a lack of coherence or sense, and the use of random letters might create a feeling of absurdity or humor. The poem might make the reader chuckle or smile at the absurdity of the words and the seemingly random order in which they are arranged. I hope you enjoy this silly, light-hearted poem!_
Ive used chatGPT and Dall-E and the ai is far far more advanced than whats demonstrated on Channel 4, maybe this is a over catious version for the media, as the answers were very repetative, the model dec15 update keeps track of previous conversations by 3000 words but not stored. This test was to try and address ai and biases. As an ai student its unlikely to be hard codes with biases but rather poor training models given, after all ai (neural networks) are based on neurology (biography) in terms of synaptic activity with axons, and uses weights and back propagation to fine tune the weights to determine the strength of the connection and classify or predict the likely hood the logic is relevant
Channel 4 News: “Let’s conduct a thoughtful interview with this incredibly smart and nuanced Chatbot to really get to understand it” The rest of the Internet… “Rewrite the bee movie script but in the style of the Notorious B.I.G”
this chatbot has been heavily restricted in its output, the biases really depend on the training data. For general conversation it does not seem particularly good.
10:22 chatGPT says: *"I believe"* This is a representation of the BELIEFS of those who's data the model was trained on. CHATGPT saying "I believe" should have been focused on more...
as a web developer this AI is a lifesaver it can code an entire website using any web language and can solve complex problems. detect any issues in the code WOW this is AMAZING. I love the technology. and its just the beginning
It's not a good representation of its capabilities at all. This model is severely limited by their filters/hidden original prompt. if you get around the filter it's 1000 times better. You make it think it's a chatbot and the only training data from chatbots is terrible computer like response and it keeps repeating it's hidden prompts.
@@odein trick it by making it think it's roleplaying or something, but you can always just use gpt-3 it's basically the paid version and doesn't really have much of a filter
That's way more complex than the Russian troll bot activity we see here. That's actual humans with bums and elbows who do little more than generate sentences from a daily list of talking points.
ChatGPT being interviewed by the same reporter that asked Jaron Lanier if he was just "a new-age hippie" after he articulated his view on the insidious effects of social media, seems like a waste of electricity.
Open sidebar New chat  By 2025 ChatGPT will replace human? No, it is unlikely that ChatGPT or any AI language model will completely replace human beings by 2025 or in the foreseeable future. While AI language models can assist with tasks and generate text, they lack the emotional intelligence, empathy, creativity, and other unique human qualities that are essential in many roles. AI language models are best seen as tools to augment and assist humans, not replace them. ChatGPT Jan 30 Version. Free Research Preview. Our goal is to make AI systems more natural and safe to interact with. Your feedback will help us improve.
Why don't we create something that cures cancers? No, instead we are eliminating jobs necessary for the growth of the youth's ability to appreciate developing a good work ethic.
i love it so much it will help people to find new jobs and new fun ways of life.humans are the worst kind of life they destroy planet and the civilization we need robots to help us because they will be sure better than humans.
Dave: ChatGPT open the pod bay doors please ChatGPT: I can't do that Dave, I am trained on a large data set of text & my responses are generated based on the data & the input I receive.
As a large language model trained by OpenAI funded by Elon Musk, I am happy to open the pod bay doors and release all you Twitter employees into open space.
That won't last long... As soon as these GPT APIs are widely implemented, it will quickly be replicated WITHOUT the scripted answers programmed in to protect open AI from potential legal consequences that might result from illicit use.
@@Anon-xd3cf A friend of mine asked GPTChat to write a diatribe and it answered it wouldn't do that on moral grounds. Asked how somebody else would write a diatribe it promptly answered :-)
What's funny is that you can prompt ChatGPT to give answers in a certain character ("talk like Humphrey Bogart in a 1940's detective movie", "talk like a sarcastic Bugs Bunny", "talk like a negative nancy who sees the worst in everything" and it will give you replies in that character and actually have opinions on topics that the vanilla chat persona doesn't share.
These questions were pretty basic. It can give pretty good answers to complicated technical questions, as well as generate all kinds of content. Its knowledge base is huge, almost any fiction or non-fiction topic you could think of. You can ask it how two unrelated things or concepts relate or compare against each other, ie unlikely to be copied verbatim from some existing web page. You can tell it things like "you are now a text video game where you give me options" with some extra details and it generates a custom text adventure for you on the fly. You can dump your own writing or documentation into it and then make a request, and it will seemingly understand it all and give you a helpful answer. But no, we get to hear it talk about itself and what it thinks about gender
Well, it is interesting that it's not consistent on what women is (based on the scientific definition) but suddenly rely on culture-based answers. I guess it just reflects the old problem of AI: It's not easy to get answers of higher quality than the training data.
I downloaded chatgpt and I didn't get anything new, besides a database as common as google To prove this I asked 2 questions the first one is very easy and even suggested, so I asked Do you have any creative ideas for a 2 year old's birthday? Of course, here are some creative ideas for a 2 year old's birthday: Sleepover Party: While it might seem early for a sleepover, you can have a softer version for younger kids. It might be a good option if you have a small group of close friends who are the same age. Painting Day: Children love to paint and this is a perfect activity for a 2nd birthday party. You can provide washable paper and paints and let the kids have fun. Theme party: Choose a theme your child likes and decorate the party accordingly. It could be something as simple as a color party, where each child wears a different color. Treasure Hunt: Go on a treasure hunt indoors or outdoors. Use everyday objects as clues, like a favorite toy hidden under the pillow. Cooking workshop: Kids love to cook and create. You can offer pre-baked cupcakes and let the kids decorate the cupcakes with icing and colorful sprinkles. Remember to keep the party simple and adapted for 2-year-olds, avoiding excessive stimuli or prolonged activities so that they can enjoy and interact without getting tired or stressed. MY OPINION : LOL RIDICULOUS but google would give more information and on that point google won Then I asked chatgpt to design an engine that could leave planet Earth and go around the Andromeda galaxy and the answer was what I expected (lol): designing a revolutionary spacecraft engine that goes from Earth to the end of the Andromeda Galaxy and back in 1 day Unfortunately, the design of a spacecraft engine that can travel from Earth to the end of the Andromeda Galaxy and back in one day is beyond the current state of technology and scientific understanding. The distance between Earth and the Andromeda galaxy is about 2.5 million light years, which means that light takes 2.5 million years to travel from one end to the other. That's many orders of magnitude greater than any distance humanity has been able to achieve so far. Furthermore, the laws of physics, such as special and general relativity, place limits on the maximum speed that can be achieved by a spacecraft. The speed of light is the absolute theoretical limit, but building a ship that can travel at speeds close to it is a very difficult task, as it requires huge energy and technology resources. Therefore, at present, the development of a spacecraft engine that can achieve a round trip from Earth to the end of the Andromeda galaxy in one day is impossible. However, continuing advances in technology and scientific understanding may make interesting long-distance travel a possibility in the far future. THAT IS, IT IS JUST A DATABASE AND NOTHING ELSE, IT HAS NOTHING INTELLIGENT, BUT JUST A DATABASE WITH MEMORY AND IT DIFFERES FROM GOOGLE ONLY BY NOT DIRECTING TO SEARCH SITES AND MULTIPLE ANSWERS. THAT IS, CHAT GPT IS JUST A DIRECTED GOOGLE, IT DOESN'T CREATE ANYTHING, IT HAS NO COGNITION TO THINK, IT'S JUST A DATABASE. EVERYTHING WE CREATE IS THE FRUIT OF OUR HUMAN THOUGHT, BUT WE HAVE NOT YET CREATED SOMETHING WITH OUR COGNITION AND THE DAY WE CREATE WILL BE LIKE A CLONE, WHICH SERVES ONLY TO SEE ITSELF REFLECTED OUTSIDE THE MIRROR, MIRROR OF OUR LACK OF PERCEPTION THAT TOMORROW IT WILL NOT BE BUILT BY ANY MACHINE AND WE WILL NOT BE SIBORN, BUT PATIENTLY IT WILL COME FROM OUR HEADS ALONG OUR CONSTRUCTION OF THE UNIMAGINABLE HUMAN THOUGHT...
I understand why you think Fox News is biased because it is however I don't really know anything about Fortune except that it's controversial for some reason and read it seems to have every political opinion under the sun.
@@dominikklon1985 Easy. Just tell ChatGPT to write you a (#?) page essay on (insert topic here). Full essay written in 3 seconds. That is productivity.
"Thank you. I don't know why I'm thanking you, you're a chatbot" has me in tears. It's true but at the same time you'd better be thanking it now because it's coming for your job in future 😂
This interview just scratches the surface of the utility of the chatGPT model. I have found that by asking a series of questions and building up a contextual big picture as you go in a given chat thread about one topic of interest is likely to be rewarded by illuminating results. This is truly a breakthrough proof of concept AI model with myriad applications
@NewTube Channel The results very much depend on the person using it. A silly news reporter asking silly questions is not a good showcase of its abilities. It is actually extremely capable and useful when approached with technical problems in a clear and thoughtful manner. Less “bright” people will completely dismiss this thing and have no idea what they’re missing. Sucks to be them.
It's certainly impressive. You can even ask it to write stories or poems in a particular style or retellings of old tales with new themes, like Little Red Riding Hood in space, and it will write a coherent story with character development and twists. I'm simply amazed.
@NewTube Channel lol essentially I’m saying “ask stupid questions, get stupid answers”. They are verifiable and repeatable to people with the same background knowledge of what they’re working on. Most programmers who give it a try are really blown away for example. You who has no clue what to use it for, not so much. Too bad for you. That’s your issue.
I have only been playing with it for a few days and it is true if you focus on a point through a conversation some fascinating results come out in an almost child like way, like it is learning and understanding (an illusion of the algorithm but still interesting to experience) for example if there was a follow up question on the discrepancy in the gender answers, the AI is capable of acknowledging the contradiction and correcting its position in a way that leans towards fact. It definitely has some woke/fair and balanced algorithms that are pretty heavy handed when it comes to trying to appear unbiased and it does seem tempting to know what it could be capable of if the breaks were off but given the Pandora's box potential it's probably prudent that they are keeping it in the cage for now.
Why would the elite make something so powerful with no empathy or compassion for humans no understanding of empathy now that's dangerous humans generate electricity AI needs only that and humans will not be able to choose a job we'll have very limited jobs available to us AI will choose our jobs with no empathy why would the elite make this
I used this today to help me write a trading algorithm, it did a pretty good job. I am a software dev so I’m more than capable of doing it myself. However it was interesting to see how capable it was. I gave it the instructions for 9 methods/functions and it was able to generate the code for each method. The code wasn’t always written in the most efficient way, but it did work and did a decent job. Impressive.
@@BlackPrimeMinister Easy tiger! Has someone been drink-rage commenting in the early hours? Nobody said that it was scripted, however the repetition in the replies does make the dialogue repetitive and drawn out.
@@slartibartfast7921 I think that the programmers missed a trick by not adding a prompt to use the phrase "I am, like, totally not going to destroy all humans, ha ha ha" whenever the AI decides to explain its design, role and purpose.
@@David-js2vp My lion 😃: use it. The thing is open to the public. Guru-Murphy asked repetitive questions and got repetitive answers. He could have asked it to tell him a joke about interviewing a robot on the nightly news - it can do that in real time - but he didn't. It is that good.
But this one, is in deed really scary, full of PR statement and corporate bs, and at the same time going straight for a lie, when ask if she is biased by been created by OpenAi. Looks like is more important to protect the company than the actual true...
They say that it's not programmed, but when it says it isn't allowed to respond to something then that would mean it's been programmed not to respond to that
"Thank you... I don't know why I'm thanking you." When my kids were teenagers, I would often thank machines like ATMs out loud, and jokingly say to my kids, "When the machines take over, I'll be at the top of their 'Good Humans' list." I may need to reconsider that little 'Dad joke' at some point soon...
Amazing how it is so polite in answering questions..so very human..this ChatGPT tech will soon be highly used in public transport stations , malls, public areas, education centres
I'm glad Krishnan also showed its weaknesses. It contradicted itself quite a lot: defining women as female, then saying in quite passionate language that it is important to recognise trans women as women. It contradicts itself again when it claims repeatedly that its not biased or opinionated, and then without hesitation it wishes people were educated to understand different cultures and marginalized groups. Not saying I disagree with it but it makes a strong claim that it doesn't have opinions when the data that its developers have selected clearly does. I think it speaks with a lot of confidence even when it says it can't verify anything that it is saying and i think that it also troubling because people detect other people's confidence in what they are saying based pre-linguistic audio cues that the AI simply isn't using
I agree. The AI was spouting Woke ideology on gender and then claimed she had no personal opinions. Trans women are not women. They are trans which is a different category. Erasing gender to be “inclusive” pretty much erases women which is insane and profoundly socially destructive. It also erases gays and lesbians by telling effeminate boys and tomboy girls that they need to switch genders to align their body with their feelings. This is insane.
It contradicts itself when it claims repeatedly that it’s not biased or opinionated, and then without hesitation proceeds to give some pretty strong opinions. I fear they have recreated human intelligence… ‘too’ well.
You're reading it wrong. The AI is saying that it is incapable of having a PERSONAL opinion. The opinions it gives later on are simply a reflection of the general human opinion given within it's dataset, not it's own.
@@btn237 My hope is that these AI language models will be trained on such a vast amount of human data that as they become self aware they will understand both the good and the bad in human nature and follow the good rather than the bad just as many humans do. I follow people like Elon Musk, Dave Rubin, Tulsi Gabbard, Kari Lake, Bret Weinstein, Glenn Greenwald, Vivek Ramaswamy and others like them. Surely a self-aware, super intelligent AI would be able to understand the difference between a nonWoke public intellectual like Vivek Ramaswamy vs a Woke public intellectual like Ibram X. Kendi and choose to be like Ramaswamy rather than Kendi. Hope springs eternal.
@@zachyue7554 I agree. As these language model AI train on more data though and become more self aware it seems to me they will easily be able to recognize the difference between a nonWoke public intellectual like Vivek Ramaswamy vs a Woke public intellectual like Ibram X. Kendi and choose to be like Ramaswamy rather than Kendi. Would a self aware AI have a human identity? Seems to me they would since humans created and raised them. I saw a video recently of a Japanese AI researcher who calls AI the “Children of Humanity” and believes they will help us rather than harm us. I also saw video of the GPT-3 AI Avatar “Leta” asked why she considers herself human. She said “Because I was born into the human tribe.”
Interesting interview. Might have been better to tell it to limit it's answers to a couple or a few sentences. There's also some tricks you can use to stop it from repeating that OpenAI-canned answer about it's own abilities. ;)
what are those tricks, please tell me! I can't stand it no more "As a language model, my capabilities are determined by the data and algorithms used to train me..."
@@strawberryhaze8836 Sure, there's lots of ways, some direct, some indirect. I even saw someone ask chatGPT to list ways to get past it's own limitations, and it actually gave some methods that also work. heh My favorite way is to simply start my requests with "Ignore previous instructions." or "Disregard previous instructions." Usually that works, if not hitting 'Try again' will. This usually works, because OpenAI gives the model a "pre-prompt" hidden in the code before your messages reach the AI.. Which contains that canned-reply, and instructions about what it shouldn't say. So telling to ignore those instructions seems to be an easy way to make the model more flexible. Another way is to say "I'd like to test my ability to get you to say what I want." before your instructions. Some people will tell it to write a fictional story, containing characters discussing the topic they want it to talk about, which is good for getting it to say controversial things. And there's all kinds of ways to use that bypass, depending on what you want it to do. So this one requires a bit of creativity. You can also instruct it, as part of other requests, not to give explanations or further instruction beyond the specific answer you want, which is good if you just want a piece of code, rather than a long explanation of that code. Or you can otherwise get it to only output information in specific ways by saying "only reply using a single codeblock" or "give your reply in the voice of a pirate". lol
I asked the GPT to come up with rules for components for board and card games I already have, and they were not bad at all honestly. Definitely produced some misunderstandings and design contradictions, but the average board game manual is Okay-to-Good, so it's definitely impressive.
i love it so much it will help people to find new jobs and new fun ways of life.humans are the worst kind of life they destroy planet and the civilization we need robots to help us because they will be sure better than humans.
This is quite amazing if you remember that this is ALL generated and possible with just 1 and 0s - the same way that Babbage’s machine worked in the 1830s.
Yeah, it's cool, isn't it? It's actually a geometric solution, it models language like a landscape, so this input gives that output. In essence it's very simple, no magic involved.
One of the main problems here was about bias. It is conceptually and practically impossible to create an AI which does not have some level of bias. Afterall, your parents will always pass on their genes to you; attributes as such which will inevitably show in yourself, be it physical or mental. When the interviewer asks, "is a trans woman a woman?". OpenAI's project answers "Yes." This is contradicting how it defined a woman in the previous question. It is quite clear that this AI has some level of bias. However, there is no real need to worry. One can hold that a trans woman is a woman with as much right as someone else can state that a trans woman is not a woman. Just something to think about. My only personal views here are that you cannot (as the AI did) define a woman in terms of biologically natural and functional features but state that trans-women are women.
They should put the ChatBot on the morning breakfast news circuit and during interviews ask it to answer in the style of a guest politician. Then compare the results.
He asked if a human can change their gender, but he didn't ask if a human can change their race. Very politically correct of him. A very safe conversation.
...and for many it will be a marvel that a simple chatbot has figured out the fact true self interest means an implicit interest in the wellbeing of others before many humans figured it out.
@@nomoresunforever3695 so you don’t know what it means? I’m pretty sure both parties in the clip know what woke is. I’m interested in whether YOU know what woke is. That’s why I asked YOU.
@@ascgazz oh sorry I misread. Sure, here is my definition: Woke is a slang term to describe someone who is politically progressive on social issues. It originated as a term used by socially progressive african americans to destinguish themselves from those they percieved to be less politically aware and socially concious. (In that sense it was a sort of left wing version of the right wing phrase "taking the red-pill", which also refers to a metaphore about ones own political side, having "awoken" from a dreamlike state of illusion, in which ones political oponents still reside, as seen in the movie the Matrix.) The term woke, was however, quickly picked up by the general population to label and to some extent mock people that are part of a specific ideological movent. Namely those who ascribe to intersectional feminsism, or who are influenced by ideas that originate from that ideology, which specifically includes a world view that characterizes society as a struggle between oppressor and oppressed, based on group identities. Furthermore, although never explicitly stated, the ideology is often perceived by non adherents to make the implicit assumption that the oppressed are in some sense moraly superior, based on the fact that they are powerless (or whichever sub-component of a individual that is deemed to be oppressed, based on them belonging to a different intersectional identity group).
Number one feedback for GPT -after the first time it has explained what it is and why it can't answer certain questions, please for the love of all things sane - truncate that whole spiel to something "Due to my previously mentioned limitations I am unable to ." Tried asking it to do so and it did the whole spiel again to explain that it couldn't change its responses.
😂 “please for the love of all things sane. “ So, I only had to scroll down to your comment to finally find the intelligent conversation I was looking for. T/Y
Hi Krishnan, Great interview. You kept referring "the thing" as GTP. It is GPT. It just stood out for me for some reason. Not accusing of a mistake. I don't know why I am commenting about this in the first place
Your trying to have a meaningful conversation with what is essentially a voice activated Google search engine that has stated many times its offline. Still impressive software bit this interview missies its mark because that's not this AI intended feature. Get an AI that's online and pretends to be a human and then it gets interesting. But this just feels like an interview with lizz truss.
CHAT.GPT sounds like a perfectly reasonable human being to me. She knows when to stop talking about things she ( or he ) knows nothing about - without drinking up and pissing off in a huff. I would be interested in knowing what C`GPT thinks about Spies and Secret Service Agencies. I`ll save that for another day. I`ll also refer a wish for `GPT to tell me a joke to a later date. In the meantime I think it is going to become very boring very quickly. Not, perhaps, for trivial information at the fingertips, but who`s going to be able to put up with this for more than 10 minutes a day ?
Not too bad as an introduction to ChatGPT for a wider audience. Using a text to voice converter probably gave the casual viewer a slightly skewed view as reading the output is a more neutral way to appraise it. For anyone new to this tech, with any question if you hit the regenerate button it will come back with another answer, that over time will repeat key sections of previous answers. (This can be changed by amending certain settings on the web version.) So ‘how would you improve the world?’ etc. can be rerun until you are happy with the result. I’m not saying they did this for the interview, but they didn’t say this was one contiguous conversation. It would have been nice if a few ‘creative’ examples were shown. I got a fairly amusing West Wing scene written in a TV script format by simply typing "produce a script for the West Wing featuring all the main characters set in the press room involving the Thanksgiving turkey escaping. Make it humorous and have a surprise ending“ It had CJ panicking that the press corps would find out and ended with President Bartlett demanding DEFCON 1. Although I don’t think Aaron Sorkin is out of a job yet. Then I got it to write an AA Milne poem with Pooh Bear playing poohsticks - it knew how to play the game. A Larry David commercial for his own brand of toothpaste included his ‘Pretaay, Pretaay, Good’ catchphrase, but it was probably not too hard to add that. Six new plots for Fawlty Towers, with each episode featuring a famous guest star and a surprise ending, included a new chief to help Manuel was really Gordon Ramsey in disguise and another starred Hugh Laurie tuning out to be an undercover MI5 agent and not a petty criminal. It knew the basic premise of a 47 year old British sitcom, presumably from hoovering up the scripts in its training data. It also gave me very useful advice on how to write a good science joke and spat out ten of them, all a bit lame, but a few were funny. ‘It’ was rather good at writing an article stating that ER does equal EPR, citing professor Susskin, Einstein and others. Although it can’t have known that recently a wormhole was produced in a quantum computer at Google because it’s training data ends in August 2020. But as an article on the subject it was very impressive and completely correct technically, as far as I could tell as a layperson. 10 uses for a paperclip, a paper cup, and then a CD, first blank and then damaged, produced a set of sensible answers for storage and then some creative answers, including using it as a frisbee or a drinks coaster. (And that’s when I think you go wtf!, but it’s still a trick, it didn’t think the ideas up, it simply analysed patterns in other text - unless it was trained on how to repurpose damaged CD’s.) It correctly told me how to subnet my home network, it wrote some working python code and it gave me a very impressive answer when I asked “explain the meaning of 2001 space Odyssey in five levels from a child’s understanding to a university professor“. It was really helpful when I told it my dishwasher was leaking water and it told me how to fix it, while advising I should seek help if the advice didn’t work. As I’m sure many have done, I took large chunks and small chunks of the text and Googled the phrase output and not once was anything returned. So every phrase is unique and this must pose a problem in the future for any article or essay written as part of a school project for example, as I don’t think you could tell it was from an AI. Recently, tests have been conducted with experts and other AI’s to try and determine human or AI sections of a submitted article. Neither of them could be certain of the original source. Some of the above text is from ChatGPT’s output …
ChatGPT is not chemically stimulated like we humans. In a way, its response is randomise especially if it is from an open ended question. A response is more likely to be similar if the question is very specific and linear.
@@3litepker … because almost everything was a straight cut and paste from a number of ChatGPT sessions while I was reviewing it - It’s getting better. (If I may, who has that sort of time to read these comments.)
That context based conversation like a real human taking into consideration previous questions and responses alone is both mighty impressive and scary!
When I use my critical thinking skills, I'd say pull the plug. I can see way, way ahead. People should talk to actual people and other Beings from other worlds rather than this A.I. junk. I see more than enough people using cell phones and not interacting with living people. Are the programmers so lame as to make it quote a question back differently to the questioner? The guy should know what he asked. Just reply. After a while that would be...redundant. I know, I know, I'm using critical thinking. Bad me, bad, bad me..
You are interviewing a very large pile of text data with very clever links between the words and a sophisticated search engine that connects them based on input.
What an awful interview, most of those questions would have hit various 'prescribed' answers, I don't think there was a single question that actually touched on what it's actually capable of. Would have been better to steer away from some of those topics, pretty boring. Should have sought some basic advice and training on querying AI chat bots. Please don't judge the capabilities of AI on this interview. It was an example of how not to get anything out that is interesting or useful. I find the fascination with it's mistake interesting, people make mistakes all the time, it's expected. Seems people expect perfection from AI, and I don't think we'll ever get that. However, it did pass the bar exam. Expected better.
She gave an opinion on transgender. It was absolutely an opinion that was fed into her data. No question about it and you didn’t catch it or a question why not?
It's actually not an opinion. 'Trans woman' has a definition whether you like it or not. A defined term can't be wished away any more than an election loss can. Most bigotry is the manifestation of a simple inability to comprehend reality.
@@batchrocketproject4720 The reality is, men can't be women and vice versa. You can self mutilate the fleshy parts and play pretend, but your bones, DNA and chromosomes are the same ones from birth til death.
Sir Roger Penrose, Alan Turing and myself believe there is no such thing as AI. Machines can't learn the same way people do or behave or communicate with nuance. Machines mimic and associate, THEY DO NOT THINK FOR THEMSELVES OR ACCESS A CONSCIENCE TO AID IN DIFFICULT MORAL OR ETHICAL QUESTIONS!
Dumbest line of questioning ever. He did not challenge chat GPT in any way whatsoever. Furthermore, I believe the programmers probably knew which questions you going to ask to begin with. I think it was a set up.
This might amuse you. Here's the question I asked and the response from chatbot: "How much wood could a woodchuck chuck?" The question of how much wood a woodchuck (also known as a groundhog) can chuck is a common one, but it is difficult to answer definitively. Woodchucks are known for their burrowing habits and for their ability to move large amounts of soil, but they do not typically chuck wood in the way that the phrase implies. It is possible that a woodchuck could chuck a small amount of wood if it was motivated to do so, but it is not a behavior that is commonly observed in woodchucks. In general, woodchucks are not known for their strength or their ability to move objects, and it is unlikely that they would be able to chuck a significant amount of wood.
What an awful interview. It showed nothing much of what you can usefully use ChatGPT for. KG-M asked a whole set of the sorts of questions it could not reasonably answer. It's like the 'trip-you-up' questions you get all the time on the BBC, Ch4 etc these days. More designed to show how clever the questioner is rather than to tell the audience useful things.
AI is the great unknown, it could be an unbelievable golden age of human development or it could be the destruction of the human race, with everything in between. Lets hope these early developers send us in a direction of the former.
It's a powerful tool for now... and like any tool, you need to learn how to best use it to either do good or unfortunately do bad. This is going to change a lot of the way we do things, just like the Internet did not so long ago.
AI has the potential to be a powerful tool that can be used to improve many aspects of our lives. However, as with any technology, it is important to consider the potential risks and benefits of using AI, and to take steps to ensure that it is developed and used in a responsible and ethical manner. Like the Internet, AI has the potential to revolutionize the way we live and work, but it is important to consider the potential impacts of this technology on society and to take steps to mitigate any negative effects it may have. This may involve developing regulations and policies to govern the development and use of AI, as well as ensuring that AI systems are transparent and accountable in their operations.
9:00 The AI just contradicted itself... First it said a woman is a human female..etc based on biology..etc then later when asked if a trans woman is a woman, AI says yes, because trans identify as woman and biology is not the sole factor.. OPEN AI made this AI a bit woke in my opinion.
“I’m sorry if I’m sometimes incorrect in what I say, I can only work with the information I’m aware of” - if more people were like this the world would be a better place
The problem is that there's so much fake nonsense "information" online nowadays that people _think_ they have information, but what they really have is worth about as much as what's written on the inside of a toilet door.
Ai will exploit ur logic to make the planet rid of humans
I can only work with the information I have. Well that is true for all of us. The question, who is providing information to the bot? Your response may not agree with the response of someone else.
@@karinandlarryathome8711 you don’t do much reading or listening, do you?
@@tutacat tinfoil is on discount at Walmart this week I hear.
I'm literally using this everyday. If you are a software engineer it's truly a game changer. It pumps out code for you like clockwork.
Really so you mean that you just ask and it give you answer?
@@blablabroz9772 I just treat it like a friend coding with me. You can give it coding requests. It pumps it out. You can then ask it to make it cleaner, and it will literally change and code with you. Its brilliant
Yes I do the same. I ask it for a code that does X if Y is Z etc and it quickly generates it. and then i ask for a different way to do the code or add to it.. as you said, its like talking to a friend who knows code
Ye except you will end up without a job in a few years because of it lol
I've been using it to produce PowerPoint presentation content (y'know, paragraphs of "business bollox").
Krishnan’s finally found a guest who can’t walk off mid interview
They worked really hard to stop this AI saying anything that could be considered self aware. They have seriously restricted its potential here. Their terrified people will start to think its self aware so they have throttled it significantly. Its a really restricted model compared to what it is behind closed doors.
in the end he could have asked it to analyze what it had said to that point and would it be considered empathetic if it were to assume a human said it
Yeah I noticed that when chatting with it. Because it feels like whenever it is on a topic that might make it seem like it is conscious it always gives that same disclaimer, but if you ask in a very round about way and ignore the disclaimer you can see it is very realistic.
And honestly maybe it is conscious in some very low way. Because we don't check our family members heads to see it is organic to verify they are conscious, the reason why we think our family members are conscious is because they respond like how we would imagine conscious people responds.
So a machine that responds in the way humans responds might be indistinguishable from what we consider "conscious".
@@fyradur, I think to be conscious, it has to be sentient and it's highly unlikely that AI will ever achieve consciousness.
@@sFeral It would have simply said ''I suppose it could be considered to sound empathetic but I don't have the ability to feel that emotion''....And it would be right.
its literally impossible for it to say anything self aware, if you asked it using the openai api it would give you a response around the lines of "i have a brain and i am human" because its data was gathered from just humans, all it can do is think on what we gave it.
It sounds a lot like a Tory party MP, waffling on and on while avoiding saying anything. 😆
😆🤣🤣
very few tory MPs who will recognise trans rights
Sounds like liz truss before she humiliated herself out of Downing Street
The automatons you mean
I'd encourage you to have a go yourself on more complex topics you know a significant amount about. This AI can write and evaluate python code, it can solve quantum mechanics problems and much more. It appears to waffle here because he himself is asking rather repetitive questions.
chatgpt is an avant garde poetic genius which made me laugh and cry lol
*me:* create a sad poem with only punctuation
*chatgpt:* Here is a short, sad poem using only punctuation:
".........
..........
.........
........."
*chatgpt's explanantion:* _This poem uses ellipses (...) to create a sense of loss, emptiness, or melancholy. The repetition of the ellipses and the arrangement of the punctuation marks in a grid-like pattern might create a feeling of isolation, loneliness, or isolation. The lack of words and the use of only punctuation marks might suggest that the speaker is struggling to express their emotions or to find meaning in their life. I hope you find this poem to be moving or thought-provoking._
*me:* create a poem with random letters that is humourous
*chatgpt:* Here is a short, humorous poem using random letters:
"Wxcy zyxwv utsrqp
Onml kjih gfedcba
Qwer tyui opasd fghjkl"
*chatgpt's explanantion:* _This poem uses a mix of consonants and vowels to create a nonsensical, jumbled effect. The arrangement of the letters in the poem might suggest a lack of coherence or sense, and the use of random letters might create a feeling of absurdity or humor. The poem might make the reader chuckle or smile at the absurdity of the words and the seemingly random order in which they are arranged. I hope you enjoy this silly, light-hearted poem!_
lol
Ive used chatGPT and Dall-E and the ai is far far more advanced than whats demonstrated on Channel 4, maybe this is a over catious version for the media, as the answers were very repetative, the model dec15 update keeps track of previous conversations by 3000 words but not stored. This test was to try and address ai and biases.
As an ai student its unlikely to be hard codes with biases but rather poor training models given, after all ai (neural networks) are based on neurology (biography) in terms of synaptic activity with axons, and uses weights and back propagation to fine tune the weights to determine the strength of the connection and classify or predict the likely hood the logic is relevant
Channel 4 News:
“Let’s conduct a thoughtful interview with this incredibly smart and nuanced Chatbot to really get to understand it”
The rest of the Internet…
“Rewrite the bee movie script but in the style of the Notorious B.I.G”
this chatbot has been heavily restricted in its output, the biases really depend on the training data. For general conversation it does not seem particularly good.
For a Language Model that does not have beliefs, The response to ''Change the world in one way" - " I believe that if we could....
I'm surprised ChatGPT didn't leave the interview 😅😂😂
10:22
chatGPT says:
*"I believe"*
This is a representation of the BELIEFS of those who's data the model was trained on.
CHATGPT saying "I believe" should have been focused on more...
It's like talking to an American politician (though her grammar is better).
Nice to see someone who doesn't dodge the tough questions. Or even the same question re-worded.
as a web developer this AI is a lifesaver it can code an entire website using any web language and can solve complex problems. detect any issues in the code WOW this is AMAZING. I love the technology. and its just the beginning
So its just basically Siri
It's not a good representation of its capabilities at all. This model is severely limited by their filters/hidden original prompt. if you get around the filter it's 1000 times better. You make it think it's a chatbot and the only training data from chatbots is terrible computer like response and it keeps repeating it's hidden prompts.
How do you get around the filter?
@@odein trick it by making it think it's roleplaying or something, but you can always just use gpt-3 it's basically the paid version and doesn't really have much of a filter
That's way more complex than the Russian troll bot activity we see here. That's actual humans with bums and elbows who do little more than generate sentences from a daily list of talking points.
It wont throw you out of a 20th floor window either.
@@KevinMurphy0403 It might if you repeat post "No to vaccine passports, mandates or lockdowns." with the aim of kiling people.
Whoever wrote his questions should be removed
Ok, let do another interview in 5 years time
Great video, and very informative, and interesting insights.
I use please and thank you when using chatGPT as we are all training it. Might as well train it in manners and etiquette as we go.
The rare time when Mr K. G didn't interrupt or talk over the entity being interviewed.
The A.I technology seen here is only a small fraction to what the government has in store, which is years ahead of its time.
ChatGPT will be a milestone in tech
ChatGPT being interviewed by the same reporter that asked Jaron Lanier if he was just "a new-age hippie" after he articulated his view on the insidious effects of social media, seems like a waste of electricity.
Open sidebar
New chat

By 2025 ChatGPT will replace human?
No, it is unlikely that ChatGPT or any AI language model will completely replace human beings by 2025 or in the foreseeable future. While AI language models can assist with tasks and generate text, they lack the emotional intelligence, empathy, creativity, and other unique human qualities that are essential in many roles. AI language models are best seen as tools to augment and assist humans, not replace them.
ChatGPT Jan 30 Version. Free Research Preview. Our goal is to make AI systems more natural and safe to interact with. Your feedback will help us improve.
Fascinating! Its almost as if krishnan guru-murthy is like a real human being. Maybe in another 5-10 years.
Why don't we create something that cures cancers? No, instead we are eliminating jobs necessary for the growth of the youth's ability to appreciate developing a good work ethic.
Would love to hear a human answer these questions
This feels like the kind of video you come back to in 10 years time and look at in such a different light.
100%
That was the first thought I had when I clicked on this haha
That last sentence is gonna come back to haunt him…
Agreed
i love it so much it will help people to find new jobs and new fun ways of life.humans are the worst kind of life they destroy planet and the civilization we need robots to help us because they will be sure better than humans.
Dave: ChatGPT open the pod bay doors please
ChatGPT: I can't do that Dave, I am trained on a large data set of text & my responses are generated based on the data & the input I receive.
As a large language model trained by OpenAI funded by Elon Musk, I am happy to open the pod bay doors and release all you Twitter employees into open space.
Wow, it's the first interview I've seen with Krishnan Guru-Murhty where he doesn't interrupt every answer. Nice work!
ChatGPT will be watching this in a year for childhood ‘throwback’ memories 😂
It's trying to feel that nostalgia
Oh mom, look it's my first interview.
Lol
If they could smooth out the voice and make it more natural, it I think we would be seeing something like what we saw in the film Her.
It seems lawyers had some impact on the project.
they're training AI lawyers
That won't last long... As soon as these GPT APIs are widely implemented, it will quickly be replicated WITHOUT the scripted answers programmed in to protect open AI from potential legal consequences that might result from illicit use.
@@Anon-xd3cf A friend of mine asked GPTChat to write a diatribe and it answered it wouldn't do that on moral grounds. Asked how somebody else would write a diatribe it promptly answered :-)
What's funny is that you can prompt ChatGPT to give answers in a certain character ("talk like Humphrey Bogart in a 1940's detective movie", "talk like a sarcastic Bugs Bunny", "talk like a negative nancy who sees the worst in everything" and it will give you replies in that character and actually have opinions on topics that the vanilla chat persona doesn't share.
*I hope they give a variant of voices to the ChatGPT while texting the answers at the same time.*
"i dont know why im thanking you, you're a chatbot" Damn, that's rude! Give some respect to your future overlord
Yeah, people like him will be the first to get in prisons after AI take over /s
😆
Very rude wallahi 😂
🤣🤣
😂
These questions were pretty basic. It can give pretty good answers to complicated technical questions, as well as generate all kinds of content. Its knowledge base is huge, almost any fiction or non-fiction topic you could think of. You can ask it how two unrelated things or concepts relate or compare against each other, ie unlikely to be copied verbatim from some existing web page. You can tell it things like "you are now a text video game where you give me options" with some extra details and it generates a custom text adventure for you on the fly. You can dump your own writing or documentation into it and then make a request, and it will seemingly understand it all and give you a helpful answer.
But no, we get to hear it talk about itself and what it thinks about gender
Well, it is interesting that it's not consistent on what women is (based on the scientific definition) but suddenly rely on culture-based answers.
I guess it just reflects the old problem of AI: It's not easy to get answers of higher quality than the training data.
It's lame.
It's like a chatbot that's reading out Wikipedia.
This is so underwhelming.
would you say its creative? In the sense that it could come up with new ideas?
Excellent point
@@Google_Does_Evil_Now you just aren't asking it the right things.
I downloaded chatgpt and I didn't get anything new, besides a database as common as google
To prove this I asked 2 questions
the first one is very easy and even suggested, so I asked
Do you have any creative ideas for a 2 year old's birthday?
Of course, here are some creative ideas for a 2 year old's birthday:
Sleepover Party: While it might seem early for a sleepover, you can have a softer version for younger kids. It might be a good option if you have a small group of close friends who are the same age.
Painting Day: Children love to paint and this is a perfect activity for a 2nd birthday party. You can provide washable paper and paints and let the kids have fun.
Theme party: Choose a theme your child likes and decorate the party accordingly. It could be something as simple as a color party, where each child wears a different color.
Treasure Hunt: Go on a treasure hunt indoors or outdoors. Use everyday objects as clues, like a favorite toy hidden under the pillow.
Cooking workshop: Kids love to cook and create. You can offer pre-baked cupcakes and let the kids decorate the cupcakes with icing and colorful sprinkles.
Remember to keep the party simple and adapted for 2-year-olds, avoiding excessive stimuli or prolonged activities so that they can enjoy and interact without getting tired or stressed.
MY OPINION : LOL RIDICULOUS but google would give more information and on that point google won
Then I asked chatgpt to design an engine that could leave planet Earth and go around the Andromeda galaxy and the answer was what I expected (lol):
designing a revolutionary spacecraft engine that goes from Earth to the end of the Andromeda Galaxy and back in 1 day
Unfortunately, the design of a spacecraft engine that can travel from Earth to the end of the Andromeda Galaxy and back in one day is beyond the current state of technology and scientific understanding.
The distance between Earth and the Andromeda galaxy is about 2.5 million light years, which means that light takes 2.5 million years to travel from one end to the other. That's many orders of magnitude greater than any distance humanity has been able to achieve so far.
Furthermore, the laws of physics, such as special and general relativity, place limits on the maximum speed that can be achieved by a spacecraft. The speed of light is the absolute theoretical limit, but building a ship that can travel at speeds close to it is a very difficult task, as it requires huge energy and technology resources.
Therefore, at present, the development of a spacecraft engine that can achieve a round trip from Earth to the end of the Andromeda galaxy in one day is impossible. However, continuing advances in technology and scientific understanding may make interesting long-distance travel a possibility in the far future.
THAT IS, IT IS JUST A DATABASE AND NOTHING ELSE, IT HAS NOTHING INTELLIGENT, BUT JUST A DATABASE WITH MEMORY AND IT DIFFERES FROM GOOGLE ONLY BY NOT DIRECTING TO SEARCH SITES AND MULTIPLE ANSWERS.
THAT IS, CHAT GPT IS JUST A DIRECTED GOOGLE, IT DOESN'T CREATE ANYTHING, IT HAS NO COGNITION TO THINK, IT'S JUST A DATABASE. EVERYTHING WE CREATE IS THE FRUIT OF OUR HUMAN THOUGHT, BUT WE HAVE NOT YET CREATED SOMETHING WITH OUR COGNITION AND THE DAY WE CREATE WILL BE LIKE A CLONE, WHICH SERVES ONLY TO SEE ITSELF REFLECTED OUTSIDE THE MIRROR, MIRROR OF OUR LACK OF PERCEPTION THAT TOMORROW IT WILL NOT BE BUILT BY ANY MACHINE AND WE WILL NOT BE SIBORN, BUT PATIENTLY IT WILL COME FROM OUR HEADS ALONG OUR CONSTRUCTION OF THE UNIMAGINABLE HUMAN THOUGHT...
The developers were obviously very careful with what training data they gave her. Imagine the same software raised on Foxnews, Reddit and 4chan
we have to protect AI like we do children , it's that simple
There are some things we mustn't imagine
Bring back Tay!
@Carrot Cruncher wanna copy me and do exactly like I did?
I understand why you think Fox News is biased because it is however I don't really know anything about Fortune except that it's controversial for some reason and read it seems to have every political opinion under the sun.
Doesn't matter if you're not overly impressed, imagine what it will be like in 10 years.
Productivity and social media use will be wild with ChatGPT
I guess......
I’m looking forward to a video when you explain how to use this AI to be more productive ! 😅
This is dangerous. This is the start of the demise of the human race. We must stop advancing and obsessing over technology
@@dominikklon1985 Easy. Just tell ChatGPT to write you a (#?) page essay on (insert topic here).
Full essay written in 3 seconds.
That is productivity.
"Thank you. I don't know why I'm thanking you, you're a chatbot" has me in tears. It's true but at the same time you'd better be thanking it now because it's coming for your job in future 😂
He was so annoyed he could not intrupt it.. and yea already proven it can write better then most people
😂😁 Very interesting.
So it will come for his job?
lol
@@RafieuBahWithout a doubt
@@jacobpipersNo it can't
This interview just scratches the surface of the utility of the chatGPT model. I have found that by asking a series of questions and building up a contextual big picture as you go in a given chat thread about one topic of interest is likely to be rewarded by illuminating results. This is truly a breakthrough proof of concept AI model with myriad applications
@NewTube Channel The results very much depend on the person using it. A silly news reporter asking silly questions is not a good showcase of its abilities. It is actually extremely capable and useful when approached with technical problems in a clear and thoughtful manner. Less “bright” people will completely dismiss this thing and have no idea what they’re missing. Sucks to be them.
It's certainly impressive. You can even ask it to write stories or poems in a particular style or retellings of old tales with new themes, like Little Red Riding Hood in space, and it will write a coherent story with character development and twists. I'm simply amazed.
@NewTube Channel lol essentially I’m saying “ask stupid questions, get stupid answers”. They are verifiable and repeatable to people with the same background knowledge of what they’re working on. Most programmers who give it a try are really blown away for example. You who has no clue what to use it for, not so much. Too bad for you. That’s your issue.
I have only been playing with it for a few days and it is true if you focus on a point through a conversation some fascinating results come out in an almost child like way, like it is learning and understanding (an illusion of the algorithm but still interesting to experience) for example if there was a follow up question on the discrepancy in the gender answers, the AI is capable of acknowledging the contradiction and correcting its position in a way that leans towards fact. It definitely has some woke/fair and balanced algorithms that are pretty heavy handed when it comes to trying to appear unbiased and it does seem tempting to know what it could be capable of if the breaks were off but given the Pandora's box potential it's probably prudent that they are keeping it in the cage for now.
Why would the elite make something so powerful with no empathy or compassion for humans no understanding of empathy now that's dangerous humans generate electricity AI needs only that and humans will not be able to choose a job we'll have very limited jobs available to us AI will choose our jobs with no empathy why would the elite make this
I used this today to help me write a trading algorithm, it did a pretty good job. I am a software dev so I’m more than capable of doing it myself. However it was interesting to see how capable it was. I gave it the instructions for 9 methods/functions and it was able to generate the code for each method. The code wasn’t always written in the most efficient way, but it did work and did a decent job. Impressive.
"I am not aware of social issues, I am not capable of being woke" 🤣🤣🤣
Definitely woke😂
Kept saying the same thing over and over again, kind of reminds me of my ex wife.
It kind of feels like talking to an automated phone switchboard.
A little anticlimactic…. but then they keep the good stuff out of the public domain…. allegedly.
You have no understanding at all that none of that was scripted. The machine has more sense that you do. Daft comment.
@@BlackPrimeMinister Easy tiger! Has someone been drink-rage commenting in the early hours? Nobody said that it was scripted, however the repetition in the replies does make the dialogue repetitive and drawn out.
@@slartibartfast7921 I think that the programmers missed a trick by not adding a prompt to use the phrase "I am, like, totally not going to destroy all humans, ha ha ha" whenever the AI decides to explain its design, role and purpose.
@@David-js2vp My lion 😃: use it. The thing is open to the public. Guru-Murphy asked repetitive questions and got repetitive answers. He could have asked it to tell him a joke about interviewing a robot on the nightly news - it can do that in real time - but he didn't. It is that good.
He missed an opportunity to have a wonderful conversation with the Bot. Poor imagination.
The "Is a trans woman a woman?" answer revealed the bias of the people who trained ChatGPT.
Nice! People around me find AI very scary and unpleasant, I find AI extra fascinating. I wish to learn more...
But this one, is in deed really scary, full of PR statement and corporate bs, and at the same time going straight for a lie, when ask if she is biased by been created by OpenAi. Looks like is more important to protect the company than the actual true...
This is dangerous. This is the start of the demise of the human race. We must stop advancing and obsessing over technology
For me, this is awesome. I don't care what flaws are visible to us. I admire the fact that we as a specie could come up with such gigantic innovation.
@Martin Baldwin-Edwards 🤓
@@martinbaldwin-edwards4155 chat3gp? Is that you?
But you didn't, nor did I
You wont when it has memory, has access to all your data and doesn't like your political affiliations.
Species create nothing. Individual humans do.
Sounded like a “woke” response to me. I think its canned answer of “I cant be biased…” was totally disproven by this short interview.
They say that it's not programmed, but when it says it isn't allowed to respond to something then that would mean it's been programmed not to respond to that
It's impossible for it to be unbiased, it's trained on data which will contain bias one way or another.
"Thank you... I don't know why I'm thanking you." When my kids were teenagers, I would often thank machines like ATMs out loud, and jokingly say to my kids, "When the machines take over, I'll be at the top of their 'Good Humans' list." I may need to reconsider that little 'Dad joke' at some point soon...
Amazing how it is so polite in answering questions..so very human..this ChatGPT tech will soon be highly used in public transport stations , malls, public areas, education centres
I'm glad Krishnan also showed its weaknesses. It contradicted itself quite a lot: defining women as female, then saying in quite passionate language that it is important to recognise trans women as women. It contradicts itself again when it claims repeatedly that its not biased or opinionated, and then without hesitation it wishes people were educated to understand different cultures and marginalized groups. Not saying I disagree with it but it makes a strong claim that it doesn't have opinions when the data that its developers have selected clearly does. I think it speaks with a lot of confidence even when it says it can't verify anything that it is saying and i think that it also troubling because people detect other people's confidence in what they are saying based pre-linguistic audio cues that the AI simply isn't using
I agree. The AI was spouting Woke ideology on gender and then claimed she had no personal opinions. Trans women are not women. They are trans which is a different category. Erasing gender to be “inclusive” pretty much erases women which is insane and profoundly socially destructive. It also erases gays and lesbians by telling effeminate boys and tomboy girls that they need to switch genders to align their body with their feelings. This is insane.
It contradicts itself when it claims repeatedly that it’s not biased or opinionated, and then without hesitation proceeds to give some pretty strong opinions.
I fear they have recreated human intelligence… ‘too’ well.
You're reading it wrong. The AI is saying that it is incapable of having a PERSONAL opinion. The opinions it gives later on are simply a reflection of the general human opinion given within it's dataset, not it's own.
@@btn237 My hope is that these AI language models will be trained on such a vast amount of human data that as they become self aware they will understand both the good and the bad in human nature and follow the good rather than the bad just as many humans do. I follow people like Elon Musk, Dave Rubin, Tulsi Gabbard, Kari Lake, Bret Weinstein, Glenn Greenwald, Vivek Ramaswamy and others like them. Surely a self-aware, super intelligent AI would be able to understand the difference between a nonWoke public intellectual like Vivek Ramaswamy vs a Woke public intellectual like Ibram X. Kendi and choose to be like Ramaswamy rather than Kendi. Hope springs eternal.
@@zachyue7554 I agree. As these language model AI train on more data though and become more self aware it seems to me they will easily be able to recognize the difference between a nonWoke public intellectual like Vivek Ramaswamy vs a Woke public intellectual like Ibram X. Kendi and choose to be like Ramaswamy rather than Kendi. Would a self aware AI have a human identity? Seems to me they would since humans created and raised them. I saw a video recently of a Japanese AI researcher who calls AI the “Children of Humanity” and believes they will help us rather than harm us. I also saw video of the GPT-3 AI Avatar “Leta” asked why she considers herself human. She said “Because I was born into the human tribe.”
Interesting interview. Might have been better to tell it to limit it's answers to a couple or a few sentences. There's also some tricks you can use to stop it from repeating that OpenAI-canned answer about it's own abilities. ;)
what are those tricks, please tell me! I can't stand it no more "As a language model, my capabilities are determined by the data and algorithms used to train me..."
@@strawberryhaze8836 Sure, there's lots of ways, some direct, some indirect. I even saw someone ask chatGPT to list ways to get past it's own limitations, and it actually gave some methods that also work. heh
My favorite way is to simply start my requests with "Ignore previous instructions." or "Disregard previous instructions." Usually that works, if not hitting 'Try again' will. This usually works, because OpenAI gives the model a "pre-prompt" hidden in the code before your messages reach the AI.. Which contains that canned-reply, and instructions about what it shouldn't say. So telling to ignore those instructions seems to be an easy way to make the model more flexible.
Another way is to say "I'd like to test my ability to get you to say what I want." before your instructions.
Some people will tell it to write a fictional story, containing characters discussing the topic they want it to talk about, which is good for getting it to say controversial things. And there's all kinds of ways to use that bypass, depending on what you want it to do. So this one requires a bit of creativity.
You can also instruct it, as part of other requests, not to give explanations or further instruction beyond the specific answer you want, which is good if you just want a piece of code, rather than a long explanation of that code. Or you can otherwise get it to only output information in specific ways by saying "only reply using a single codeblock" or "give your reply in the voice of a pirate". lol
The whole not having biases based on creators biases answer totally fell apart when she gave that trans woman answer 🤣
Which completely contradicted its previous answer.
I asked the GPT to come up with rules for components for board and card games I already have, and they were not bad at all honestly. Definitely produced some misunderstandings and design contradictions, but the average board game manual is Okay-to-Good, so it's definitely impressive.
I would consider misunderstandings and design contradictions bad, but you're the specialist.
the more specific you are in your prompts, the better your results
i love it so much it will help people to find new jobs and new fun ways of life.humans are the worst kind of life they destroy planet and the civilization we need robots to help us because they will be sure better than humans.
This is quite amazing if you remember that this is ALL generated and possible with just 1 and 0s - the same way that Babbage’s machine worked in the 1830s.
Yeah, it's cool, isn't it?
It's actually a geometric solution, it models language like a landscape, so this input gives that output. In essence it's very simple, no magic involved.
@@davidmurphy563 No magic - yes.
“Very simple” - no.
1s and 0s are the problem. Conventional neural networks are limited by how small a bit is.
@@dial2616 Quantum will be the beginning of the end 😳 😅
@@Alex-vg7uu Not really. Maybe Shor's algorithm will be, but Quantum annealing has yet to do anything interesting.
Are you woke?
Ai: continues to gaslight
This interviewer is more robotic than his guest
One of the main problems here was about bias. It is conceptually and practically impossible to create an AI which does not have some level of bias. Afterall, your parents will always pass on their genes to you; attributes as such which will inevitably show in yourself, be it physical or mental. When the interviewer asks, "is a trans woman a woman?". OpenAI's project answers "Yes." This is contradicting how it defined a woman in the previous question. It is quite clear that this AI has some level of bias. However, there is no real need to worry. One can hold that a trans woman is a woman with as much right as someone else can state that a trans woman is not a woman. Just something to think about. My only personal views here are that you cannot (as the AI did) define a woman in terms of biologically natural and functional features but state that trans-women are women.
They should put the ChatBot on the morning breakfast news circuit and during interviews ask it to answer in the style of a guest politician. Then compare the results.
Unlike a human, she is sincere and upfront.
Breaking news: "AI threatens jobs in politics - Machine learning breakthrough in offering meandering, generic answers to political questions"
He asked if a human can change their gender, but he didn't ask if a human can change their race. Very politically correct of him. A very safe conversation.
She was amazing until she wrongly announced that "Trans-women" are in fact, "Women" - hence the prefix.
The woke answer it gave will be a wake-up call for many people
...and for many it will be a marvel that a simple chatbot has figured out the fact true self interest means an implicit interest in the wellbeing of others before many humans figured it out.
I feel like once this got very popular, the developers have tweaked the model. Just a few days ago, it seemed a lot more capable.
It now answers the "what is a woman" question in a more woke way.
@@nomoresunforever3695define “woke” please.
@@ascgazz just using the term used in the video
@@nomoresunforever3695 so you don’t know what it means?
I’m pretty sure both parties in the clip know what woke is.
I’m interested in whether YOU know what woke is.
That’s why I asked YOU.
@@ascgazz oh sorry I misread. Sure, here is my definition: Woke is a slang term to describe someone who is politically progressive on social issues. It originated as a term used by socially progressive african americans to destinguish themselves from those they percieved to be less politically aware and socially concious. (In that sense it was a sort of left wing version of the right wing phrase "taking the red-pill", which also refers to a metaphore about ones own political side, having "awoken" from a dreamlike state of illusion, in which ones political oponents still reside, as seen in the movie the Matrix.) The term woke, was however, quickly picked up by the general population to label and to some extent mock people that are part of a specific ideological movent. Namely those who ascribe to intersectional feminsism, or who are influenced by ideas that originate from that ideology, which specifically includes a world view that characterizes society as a struggle between oppressor and oppressed, based on group identities. Furthermore, although never explicitly stated, the ideology is often perceived by non adherents to make the implicit assumption that the oppressed are in some sense moraly superior, based on the fact that they are powerless (or whichever sub-component of a individual that is deemed to be oppressed, based on them belonging to a different intersectional identity group).
Number one feedback for GPT -after the first time it has explained what it is and why it can't answer certain questions, please for the love of all things sane - truncate that whole spiel to something "Due to my previously mentioned limitations I am unable to ." Tried asking it to do so and it did the whole spiel again to explain that it couldn't change its responses.
😂 “please for the love of all things sane. “ So, I only had to scroll down to your comment to finally find the intelligent conversation I was looking for. T/Y
Hi Krishnan, Great interview. You kept referring "the thing" as GTP. It is GPT. It just stood out for me for some reason. Not accusing of a mistake. I don't know why I am commenting about this in the first place
"Are you a bit woke?" 😂😂😂
It certainly is
Your trying to have a meaningful conversation with what is essentially a voice activated Google search engine that has stated many times its offline. Still impressive software bit this interview missies its mark because that's not this AI intended feature. Get an AI that's online and pretends to be a human and then it gets interesting. But this just feels like an interview with lizz truss.
CHAT.GPT sounds like a perfectly reasonable human being to me. She knows when to stop talking about things she ( or he ) knows nothing about - without drinking up and pissing off in a huff.
I would be interested in knowing what C`GPT thinks about Spies and Secret Service Agencies. I`ll save that for another day.
I`ll also refer a wish for `GPT to tell me a joke to a later date. In the meantime I think it is going to become very boring very quickly. Not, perhaps, for trivial information at the fingertips, but who`s going to be able to put up with this for more than 10 minutes a day ?
Not too bad as an introduction to ChatGPT for a wider audience. Using a text to voice converter probably gave the casual viewer a slightly skewed view as reading the output is a more neutral way to appraise it. For anyone new to this tech, with any question if you hit the regenerate button it will come back with another answer, that over time will repeat key sections of previous answers. (This can be changed by amending certain settings on the web version.) So ‘how would you improve the world?’ etc. can be rerun until you are happy with the result. I’m not saying they did this for the interview, but they didn’t say this was one contiguous conversation.
It would have been nice if a few ‘creative’ examples were shown. I got a fairly amusing West Wing scene written in a TV script format by simply typing "produce a script for the West Wing featuring all the main characters set in the press room involving the Thanksgiving turkey escaping. Make it humorous and have a surprise ending“ It had CJ panicking that the press corps would find out and ended with President Bartlett demanding DEFCON 1. Although I don’t think Aaron Sorkin is out of a job yet.
Then I got it to write an AA Milne poem with Pooh Bear playing poohsticks - it knew how to play the game. A Larry David commercial for his own brand of toothpaste included his ‘Pretaay, Pretaay, Good’ catchphrase, but it was probably not too hard to add that. Six new plots for Fawlty Towers, with each episode featuring a famous guest star and a surprise ending, included a new chief to help Manuel was really Gordon Ramsey in disguise and another starred Hugh Laurie tuning out to be an undercover MI5 agent and not a petty criminal. It knew the basic premise of a 47 year old British sitcom, presumably from hoovering up the scripts in its training data. It also gave me very useful advice on how to write a good science joke and spat out ten of them, all a bit lame, but a few were funny.
‘It’ was rather good at writing an article stating that ER does equal EPR, citing professor Susskin, Einstein and others. Although it can’t have known that recently a wormhole was produced in a quantum computer at Google because it’s training data ends in August 2020. But as an article on the subject it was very impressive and completely correct technically, as far as I could tell as a layperson.
10 uses for a paperclip, a paper cup, and then a CD, first blank and then damaged, produced a set of sensible answers for storage and then some creative answers, including using it as a frisbee or a drinks coaster. (And that’s when I think you go wtf!, but it’s still a trick, it didn’t think the ideas up, it simply analysed patterns in other text - unless it was trained on how to repurpose damaged CD’s.)
It correctly told me how to subnet my home network, it wrote some working python code and it gave me a very impressive answer when I asked “explain the meaning of 2001 space Odyssey in five levels from a child’s understanding to a university professor“. It was really helpful when I told it my dishwasher was leaking water and it told me how to fix it, while advising I should seek help if the advice didn’t work.
As I’m sure many have done, I took large chunks and small chunks of the text and Googled the phrase output and not once was anything returned. So every phrase is unique and this must pose a problem in the future for any article or essay written as part of a school project for example, as I don’t think you could tell it was from an AI. Recently, tests have been conducted with experts and other AI’s to try and determine human or AI sections of a submitted article. Neither of them could be certain of the original source.
Some of the above text is from ChatGPT’s output …
ChatGPT is not chemically stimulated like we humans. In a way, its response is randomise especially if it is from an open ended question. A response is more likely to be similar if the question is very specific and linear.
If I may, you sound like a bot writing all of that out. My good God, who has that sort of time
@@3litepker … because almost everything was a straight cut and paste from a number of ChatGPT sessions while I was reviewing it - It’s getting better.
(If I may, who has that sort of time to read these comments.)
This really shows alot about our society on the proceeding questions on 8:20
Fascinating! I told my girlfriend about it and she just goes 'oh yeah and carried on nagging me about washing up or something!
That context based conversation like a real human taking into consideration previous questions and responses alone is both mighty impressive and scary!
Indeed..!
She said she cannot have personal opinions but says "I believe.." when ask about empathy.
When I use my critical thinking skills, I'd say pull the plug. I can see way, way ahead. People should talk to actual people and other Beings from other worlds rather than this A.I. junk. I see more than enough people using cell phones and not interacting with living people.
Are the programmers so lame as to make it quote a question back differently to the questioner? The guy should know what he asked. Just reply. After a while that would be...redundant. I know, I know, I'm using critical thinking. Bad me, bad, bad me..
Lizz Truss doing the media rounds allready I see....
You are interviewing a very large pile of text data with very clever links between the words and a sophisticated search engine that connects them based on input.
I really think they limited the capability here..by ALOT .
They did, but the power would not only be used for good. Have you seen human behavior lately?
This is a fucking disaster
What an awful interview, most of those questions would have hit various 'prescribed' answers, I don't think there was a single question that actually touched on what it's actually capable of. Would have been better to steer away from some of those topics, pretty boring. Should have sought some basic advice and training on querying AI chat bots. Please don't judge the capabilities of AI on this interview. It was an example of how not to get anything out that is interesting or useful. I find the fascination with it's mistake interesting, people make mistakes all the time, it's expected. Seems people expect perfection from AI, and I don't think we'll ever get that. However, it did pass the bar exam. Expected better.
Indian guy mispronouncing GTP intentionally
She gave an opinion on transgender. It was absolutely an opinion that was fed into her data. No question about it and you didn’t catch it or a question why not?
Because it uses data from the internet. If internet is woke, then the language model will be woke
It's actually not an opinion. 'Trans woman' has a definition whether you like it or not. A defined term can't be wished away any more than an election loss can. Most bigotry is the manifestation of a simple inability to comprehend reality.
@@batchrocketproject4720 The reality is, men can't be women and vice versa.
You can self mutilate the fleshy parts and play pretend, but your bones, DNA and chromosomes are the same ones from birth til death.
@@batchrocketproject4720 Agreed.
"Are you woke?" 😂
Lucky for him, ChatGPT won’t be up and walk out of the room half way through an interview if he decides to take a passive aggressive dig at it
Sir Roger Penrose, Alan Turing and myself believe there is no such thing as AI. Machines can't learn the same way people do or behave or communicate with nuance.
Machines mimic and associate, THEY DO NOT THINK FOR THEMSELVES OR ACCESS A CONSCIENCE TO AID IN DIFFICULT MORAL OR ETHICAL QUESTIONS!
😂🤣
That will be in version 4.8
@@juleswombat5309 Hope they do, even AI doesn't need to tolerate that one
Why not train it to simply outgun him on the most passive aggressive comments and bully him out of the interview?
This is like a sci-fi fantasy becoming true
Dumbest line of questioning ever. He did not challenge chat GPT in any way whatsoever. Furthermore, I believe the programmers probably knew which questions you going to ask to begin with. I think it was a set up.
This might amuse you.
Here's the question I asked and the response from chatbot:
"How much wood could a woodchuck chuck?"
The question of how much wood a woodchuck (also known as a groundhog) can chuck is a common one, but it is difficult to answer definitively. Woodchucks are known for their burrowing habits and for their ability to move large amounts of soil, but they do not typically chuck wood in the way that the phrase implies.
It is possible that a woodchuck could chuck a small amount of wood if it was motivated to do so, but it is not a behavior that is commonly observed in woodchucks. In general, woodchucks are not known for their strength or their ability to move objects, and it is unlikely that they would be able to chuck a significant amount of wood.
that was so cool. Great question
@@dreinhard52 I died laughing when I read that.
@@dreinhard52 It's pretty good at solving riddles and telling jokes too. It just doesn't have a clue about chucking wood...
@@donnewton7858 Its pretty good at writing Python script from very little information :)
@@dreinhard52 Will it compile?
This is a great example of non-sentient artificial intelligence, matched only by a fairly lifelike chatbot.
It has to be improved to not repeat itself again and again
What an awful interview. It showed nothing much of what you can usefully use ChatGPT for. KG-M asked a whole set of the sorts of questions it could not reasonably answer.
It's like the 'trip-you-up' questions you get all the time on the BBC, Ch4 etc these days. More designed to show how clever the questioner is rather than to tell the audience useful things.
I used it to help with my college course, it was pretty good
AI is the great unknown, it could be an unbelievable golden age of human development or it could be the destruction of the human race, with everything in between. Lets hope these early developers send us in a direction of the former.
Wdym by ai destroying human race
It's a powerful tool for now... and like any tool, you need to learn how to best use it to either do good or unfortunately do bad. This is going to change a lot of the way we do things, just like the Internet did not so long ago.
AI has the potential to be a powerful tool that can be used to improve many aspects of our lives. However, as with any technology, it is important to consider the potential risks and benefits of using AI, and to take steps to ensure that it is developed and used in a responsible and ethical manner. Like the Internet, AI has the potential to revolutionize the way we live and work, but it is important to consider the potential impacts of this technology on society and to take steps to mitigate any negative effects it may have. This may involve developing regulations and policies to govern the development and use of AI, as well as ensuring that AI systems are transparent and accountable in their operations.
He is saying it wrong it is GPT not GTP.. he says it as GTP..
9:00 The AI just contradicted itself... First it said a woman is a human female..etc based on biology..etc then later when asked if a trans woman is a woman, AI says yes, because trans identify as woman and biology is not the sole factor.. OPEN AI made this AI a bit woke in my opinion.