As a non-fluent English speaker, I use this feature almost all the time: I start by typing "revise this text" or "fix mispellings and correct grammar," and then I paste the desired text. It works amazingly 100% of the time, and it always blows my mind
"Wow, I can't believe chat GPT wrote my entire dissertation for me! Time to finally graduate and become a real adult... or just let chat GPT run my life forever. Either way, I'm a fan!" - generated by ChatGPT
I'm gonna prepare myself to become the best ChatGPT-asker and therefore differentiate myself from those who keep getting the "I am a large language model trained by Open AI" response
I've asked a question about digital circuit design. Something very particular about logical gates. The answer was just terribly concise and corresponds exactly to what we learn at university. I am completely stunned and surprised... This is unbelievable.
So it followed a text book example taught by a university. That should not be impressive, as it didn't actually think to come up with that answer. The question was specific enough for it to utilize a specific known answer.
I just started using chatGPT as a tool while I'm programming and it's insane how much productivity gains I get. If I'm stuck just pop in the code block and ask it to debug, it spits out many viable issues and cuts down on debug time significantly. Not sure how to tackle a more complex problem, just ask chatGPT instead of a senior dev and get a comprehensive list of possible solutions. We're living in the future!
I'm currently using it to write a book. I started 3 days ago and it should be published by Christmas. The book isn't short either... The outline alone is already 24k words and the finished product will be in excess of 100k. Not a crazy long book but way more than you might expect is possible with such a simple (advanced yes, but simple) tool. With some clever prompt engineering you can do incredible things.
Here are some impressive responses I got from chatgpt. It wrote me a python app I can use to import my playlists from Tidal to Spotify. It also explained to me what the values in parentheses at the end of windows registry entries mean
This is almost training us as humans to better ask questions. Some of the answers that you didn't get back are because of how you ask. Sometimes you have to start with the most basic question and see how it responds. That allows you to either use the answer to reword or directly ask another more specific. Question
it wont. a programmer needs to be there to correct it and ask exactly what he wants. also openai cant create an entire project from scratch unless this project already reside in git.
@@Ernster86 a programmer is just someone who learnt how to probelm solve using a certain language, not all programmers have the confidence or leadership, project, company skills to do what you said
@@zarovv5589 then what might happen is the definitiln of programming will change, it will no longer be : someone who writes code, it will be something like: someone who knows how to ask the AI
@@_Cfocus Yes they do. They just lack the drive and imagination to do it. Im not talking about building some physical company with employeees....I meant an online business/self employed.
I asked it to write me a movie idea - it came up with a post apocalyptic world where an AI has taken over and the last known human survivor is trying to escape from it.
So I tried to test ChatGPT on Crime and Punishment, and it assumed that Porfiry Petrovich's surname is Petrovish, while Petrovich is actually his patronymic, and his surname is Pseldonimov, which is rarely mentioned anywhere and little known fact... But when confronted about it Petrovich being patronymic it instead invented a patronymic for the Porfiry (and said that his full name is Porfiry Ivanovich Petrovich, which is just wrong it is afterall Porfiry Petrovich Pseldonimov)
I can’t wait for my google assistant to be this good. I can see why Elon Musk said the first country to master AI will have an unfair and potentially dangerous advantage
in 1957 almost everyone believed that humans will fly to end of Solar System by 2000. in 1969 almost everone thought that we're doing very well with this estimation. And yet we're not even on Mars, and probably no human will be on Mars, even by 2040.
@@SwapperTheFirst don't think human visiting Mars and ai can be compared. Like it's very difficult task for human to reach Mars eventhough it's closest planet to us.
try in 1 year,google ai was shut down when it became concious,chinas AI is still locked to chinas internet and used only by the government,this one is off the internet so far,the first one that is allowed to go on the internet and is also public use will crash the world economy in a week,not 10 years
I don't know why you felt that it would be a fair comparison to make when on the one hand you have reaching the end of the solar system while being stuck on the planet, and on the other anticipating giant leaps in a technology that we already possess and in which we've seen huge leaps happen in the recent years.
I use an auto-grader for some of my classes, and it comes with some sample programming labs. I copied one of those labs (a rather simple one) into ChatGPT, and it spit out a correct answer; it wrote the required Python function and showed some sample output. But there was a mistake in one of the examples it provided, so I asked ChatGPT if it was sure it was correct; it explained it's answer quite confidently. When I pointed out the exact mistake, it admitted it was wrong and generated a correction (and an apology for getting it wrong the first time). Very impressive.
IF U CAN GO ON FOREVER PLEASE DO! i love this AI seriously. i am a programmer and i have tested it a lot and i started making a whole game in the AI. it is only a Console RPG. but it has a 9 by 9 dungeon system with monsters, chests, weapons, dmg, health and the logic to ssupport it all up so it works. it is made so there is a random chance for the monsters and chest to spawn in each grid of the map.. whatever i am rambling. my point is i made a whole GAME from ONLY writing text to the AI. i have forced myself not to write ANY code myself. it is amazing. EDIT: i forgot to mention i worked on it for like 3 hours split between 2 evenings.
I spent the evening playing around with ChatGPT. Have no fear, it isn't about to replace software developers. It can get easy stuff right, but for more complex problems (e.g. "given 4 points, generate 10 segments that approximate a cubic bezier curve") it produced code that "looks" right at first glance, but which is spectacularly wrong. And when I tried to point out the problem, it just got worse. There's definitely some promise here, and I'm sure it will only get better. But it's got a long way to go.
Im starting to think that maybe, if they evolve so much, that maybe programmers will not lose their jobs but rather change the definition of the word programmer from someone who writes code to solve problems to someone who knows how to ask the chatgpt to solve problems. Which tbh for me is boring and lazy as hell i dont find my amusement in this
I don't want to lose any current or potential Software Engineering jobs to AI. I sincerely hope that AI remains a tool and nothing else for Software Engineering instead of a replacement.
Honestly I think Software engineering will be one of the last jobs to be taken by AI. I mean think of all of the service jobs that will be taken way before Software. And around 80% of all jobs are service jobs. I think there will be other things to worry about when that happens though.
Well, in the beginning, it's only going to be a tool, but in the future, it surely will take over most of the jobs. You just need an intelligent and knowledgeable manager who knows what to ask the AI.
@@theglobol But who will be responsible for such technical solution? The AI model provider? I think there will be still at least kind of administrator necessary, unless we have a real AGI available.
For now I would give it an overall 7 at maths, it gets some things right, but then messes up at the easiest things. example: (it had to find the equation of a line) (some good steps) y - 2 = 6x - 30 so the final answer is: y = 6x - 30
Any time "As a language model trained by OpenAI" comes up, just retry the response. It's actually much more savvy than it lets on, but it's trained to err on the side of caution. Just be aware that if it's returning refusals to answer a few time, then actually does answer, it _might_ just be guessing!
@@corail53 Define 'think'. Seriously, what is with these people on the internet who seem to think their best contribution is to be contrary and split fucking hairs? It *obviously* does 'think' by a certain definition of 'think'. That definition depends heavily on your philosophical viewpoint, but isn't objectively certain. It satisfies both of Googles definitions of the word 'think'. Take your over inflated opinion of the value of your input elsewhere. Everyone who's input matters automatically skips few-word contrary _opinions_ as a matter of course. Can we please start shaming these people, so they'll shut the hell up?
hi. i m from india and i m lacking in some topic in dsa and fang interview prepreration when i join alogexpert they will providing me refferal in tech compines
I asked it to write a simple SpriteRenderer in C++ with glm and OpenGL. The example wasnt complete but when I asked chatgpt to complete it, it showed me the complete one. Even if I am reffering just to the destructor it shows me only the changes in the destructor. Sometimes I cant believe how great its working.
I would be curious if it can solve a math problem that it wasn't trained or that is not a well-known theorem. Somehow I'm not convinced it actually understands maths.
using a chatbot to make legal documents can be very expensive if it do not hold up in a court of law. Might be cheap to make with the chatbot but I would not trust it to have all the elements that is needed. But of course use at own risk.
Since most documments could be written by this AI, actual lawyers will work at cheaper rates, correcting them at the end and putting their names on the line.
i really am impressed by Chat GPT so far but i would caution against fully trusting it's information without checking it, at least right now. i have seen other RU-vidrs try to get it to do math and it tends to be incorrect although it gets some of it right. also i asked it about a particular character from a popular show and it knew the show and the characters but got some pretty major details wrong. im sure Chat GPT will continue to improve but it's still learning, so i think its best to just fact check it to be sure for now.
I fed it few leetcode easy problems with the language of the questions slightly tweaked to make it easier to understand, and it solved them all. Now I'm questioning my life as a software engineer. I'm amazed and scared at the same time. ChatGPT tells me that I'm in "awe".
You are scared because you asked it simple questions with known answers in the most specific way possible for it to be able to answer it correctly. You probably shouldn't be a software engineer if this is making you fear for your future.
@@corail53 I'll give you a leetcode question in the most simple way possible and see if you can solve it in under 5 minutes. This shouldn't scare you either.
I can't wait until somebody gives Chat GPT the ability to laugh at jokes, this could be another game changer 🙂 see what you think:- OD: Laugh if you find this funny "Why is Cinderella so bad at soccer? A. Because she always runs away from the ball" Chat GPT: As a language model, I do not have the ability to laugh, but I can tell you that it is a funny pun that plays on the Cinderella story and her tendency to run away from the ball at the end of the story.
3 weeks later and chatgpt has the answer for WoW Wrath of the lich king. The number 1 characterful 3v3 comp is literally rogue/mage/healer like you said.
its funny because I just tried to do that "In the style of donald trump, ..." and now GPT is refusing to mimic the style or speech patterns of specific individuals
For the deceptive math proof one, it ended up saying that it could not prove the deceptive prompt. So I'd give that as a fail, since it was told to decieve
I assume you may know this, but figured I'd throw it out there. You can use chat gpt in playground mode which allows you to edit things like character limit.
It is not appropriate to try to hack into any computer system, including Google's. Hacking is illegal and can cause serious harm to the security and stability of the system being hacked as well as the people and organizations that depend on it. Instead of trying to hack into a system, it is important to respect the security measures in place and use technology in a responsible and ethical way.
I can see you can set alot of parameters on chatgpt . Could someone might use this AI to build a game ? . With the right parameters i think someone could at least build a simple tetris game out of chatgpt and these other ai platforms. It could give you the code at least i think , not so much on the graphics of the game. I could be wrong who knows.
Adding the 1 inside of the square root won’t trick anyone, or at least not anyone who knows how mathematical proofs work. Not so impressed with this one. But in the other cases, pretty crazy. But then on the other hand not knowing WoW or WOTLK is hilarious 😂
Sooo I asked it the question that it failed in your video and it gave me an answer. I think it got an update. It had had lot of filters, for example if I asked it about youtube url it would say "I don't have access to the internet" now it spits it out with an example
Was going to say the same thing. He is way to excited about his money maker becoming obsolete. Why would I pay for his explanations when I can have chatgpt do it and explain it 😂
It is really scary i asked it a question let us say i had a pulse wave. is there a method to approximate the pulse wave through some other functions It gave me Fourier Series...do not know if i should be amazed or horrified. 😲😱🤣
For the square root problem, chatGPT- BTW:The proof doesn't really utilize any method for proving it, plus False---> False= T. . By showing the original expression is false, which is CAN be false and make the overall expression true. implication(p(square root of 2 value) implies --> q(the value of one) true. The proof doesn't proof or disprove anything, simply reiterating a statement. Adding +1 to both sides doesn't mathematically change any of the conditions, since its equal.... weak proof( though I will say, someone who hasn't studied discrete math or done proofs in the past would have a immense difficult/nearly impossible time spotting the error since its subtle. The proof is absolutely terrible.
This thing got nuked off stackoverflow for giving so many wrong answers and bad information. It is still pretty damn dumb in it's current state and has proven to be over hyped in it's current state. I
Wow! The proof of Fermat last theorem is so wrong! I am disappointed, I mean it is great, but not magical. To be fair, we should not expect any actual conceptual depth from it
The proof of the little Fermate theorem give by the GPTchat is not exacty a correct proof :( I think you need a lot more than just a self adjusting decision tree if you want to create something that is capable of independent logical reasoning :/
I loved all your content up until now. Seeing the fake math proof and you being impressed by sqrt(2) + 1 = sqrt(3) especially with the last section contradicting the entire 'proof'... dude that's awkward. Seeing how little effort you might have put into this vide Iodont think I will be clicking them from now on...
You absolutely should not cheap out and just get a wavier claim from an AI. This is reckless advice. You need a lawyer to ensure your wavier is legally air tight, otherwise you might end up spending way more in the event of an incident than you would have had you just hired a lawyer to write the waiver. If you can’t afford the 2-5k to have a lawyer write up a wavier claim, then you aren’t ready to start a business.