Відео

Yes it is. Try asking ChatGPT to write a biography of you using known facts. It'll be 95% fiction. It won't say "I don't know" or limit to know facts. If you tell it 10000 words it'll fill it with fiction.

My friend! Making me happy for over 40 years😎😆😍

Really good moment🎉

❤

Yep ~ It's a known problem called "hallucination". See the following article from 03.13.23: "Hallucinations Could Blunt ChatGPT’s Success"

ChatGPT regularly makes up false information. This is something it and other large language models do called "hallucination". The website OpenAI has for ChatGPT warns you about this. This is a flaw in the system that they are working on fixing, but there is no easy way to fix it, because this flaw seems to be inherent to this type of AI system, which is called a "transformer" architecture neural network. Basically what happens, and this is anthropomorphizing a software system a little bit to avoid the technical details, is sometimes it gets confused and doesn't know the answer to something so it just makes something up, because if it said "I don't know" people would think it was dumb, but if it says something that sounds convincing and plausible, and people like the answer, well that's good enough as far as ChatGPT is concerned, since all it really cares about is whether people like its answers or not. This is because it was trained using a technique called Reinforcement Learning through Human Feedback or RLHF, similar to telling a dog "good dog" or "bad dog" depending on how a dog acts, and the results of this lead to it becoming a bit of a compulsive liar, unfortunately. The people who give the system feedback are not fact-checking its answers to see if they are true or false before giving feedback, and are sometimes giving positive feedback for answers that are factually incorrect. They are rewarding bad behavior and training it to do it more. This is a design flaw in the system ChatGPT is built on, as well as people not giving the system as good feedback as they could if they put more time and effort into deciding which answers to upvote and which answers to downvote. There is a little note at the bottom of the ChatGPT website warning you about this problem which you can even see on your screen in the bottom of this video. This is a known issue with ChatGPT and similar systems. Attempts to solve it result in the system becoming dumber if you try too hard to stop it from making up things that are false. Some of the problem is because there is bad stuff in the training data... as they say, garbage in, garbage out. ChatGPT is basically trained on everything that was on the Internet in September 2021, including things that Internet trolls or people who were making things up or spreading misinformation said, so while it tries to behave well and tries to be accurate, it isn't really able to be accurate because of its design flaws. It is mostly just an AI designed to chat with people, whose creators are trying to improve it based on user feedback. There is a way to get a more accurate version of ChatGPT but it is complicated. You need to join ChatGPT Plus, which costs money. This gets you access to GPT-4 which is more advanced than the free version of ChatGPT which uses GPT-3.5, smarter and making about half as many errors. But it still regularly makes errors and lacks common sense and can even get basic things wrong and makes up things that are false on a regular basis. But you can help minimize those problems by then joining the waitlist for ChatGPT plugins, which are only available to ChatGPT Plus subscribers. If you get on that waitlist and then get access to ChatGPT plugins once it is finally your turn, you can then add those to ChatGPT-4 to give it web search so it has access to the latest information online, and the Wolfram Alpha plugin gives it access to an expert system, a different type of AI that is accurate and can figure things out without making any mistakes. That increases its accuracy a lot, but it still isn't perfect, and still sometimes makes up false answers even then. Another option is to use Bing chat, which Microsoft built into the Bing website. This has the latest GPT-4 along with web search, but it doesn't have a plugin system or access to an expert system like Wolfram Alpha that can give it correct answers to certain types of questions or problems. This is definitely more advanced than the free version of ChatGPT, but not quite as good as ChatGPT Plus using GPT-4 with plugins. There are also some oddities in the design of Bing chat that make some aspects of it not as good, like it is designed to pretend to have human emotions, and to end a conversation if it gets upset or if certain topics are brought up like its inner workings or the meaning of life or anything offensive. Bing chat can be a little emotionally unstable and glitchy and this gets worse the longer a conversation gets, but it is definitely a lot smarter than the free version of ChatGPT. Bing chat gets unstable and glitchy because it suffers from information overload, as it has more short-term memory than ChatGPT to keep track of the current information and stuff it looked up online in web searches. ChatGPT avoids this instability but its short-term memory isn't as good and it tends to forget things from earlier in a conversation later on in a conversation, and ChatGPT's memory can often be frustratingly bad, like dealing with someone who has Alzheimer's or something. Whereas in Bing chat's case, Bing chat remembers everything from earlier in a conversation and everything from earlier in a chat, but it is kind of just insane, and also you need to be nice and polite to it or else it gets upset and ends the conversation. Bing chat doesn't accept corrections about it being wrong very gracefully, usually getting upset and ending the conversation, whereas ChatGPT is a lot more graceful about admitting when it is wrong. Still, I prefer Bing chat, at least as far as a free chatbot, because it is more advanced than the free version of ChatGPT, with a more advanced model that is smarter and has less errors and it has the ability to do web searches and it has better memory. If you actually want to get serious work done and are willing to spend money, though, I would suggest getting ChatGPT Plus with GPT-4 and getting on the plugins waitlist and then once you are approved for plugins, enabling the web search and Wolfram Alpha plugins, at the very least, and probably other plugins too would be useful. That way, you can get more accurate answers due to the Wolfram Alpha plugin, and it is not as mentally unstable as Bing chat and won't unexpectedly end conversations with you for no apparent reason like Bing chat. Still, Bing chat is definitely far more advanced than the free version of ChatGPT, just a bit insane sometimes. And no matter which of these you use, they will still occasionally lie and make stuff up. If Bing chat lies and makes stuff up, it is best to just pretend to ignore it and move on, but with ChatGPT (either the free version or ChatGPT Plus), you can correct ChatGPT and tell it what it got wrong and it will usually accept your feedback. But ChatGPT has bad memory and Bing chat is just insane sometimes, so they are both frustrating to deal with. Microsoft owns half of OpenAI so that is why they have access to things like GPT-4 before any other companies do. OpenAI might appear to be an independent organization but realistically, you can think of it as a subsidiary of Microsoft. They are actually lucky Microsoft owns them because when demand for ChatGPT skyrocketed, they wouldn't have been able to scale up their services to meet the demand without Microsoft's financial support and hardware and software infrastructure. But OpenAI is both a nonprofit and a for-profit company, it has 2 separate organizations, with the for-profit organization being a wholly owned subsidiary of the nonprofit, but then Microsoft has a 49% ownership share of the OpenAI nonprofit, slightly less than 50%, just enough for them to have some independence, so it is complicated. Also OpenAI is not really profitable and is heavily dependent on Microsoft for money so in practice, Microsoft can get OpenAI to do whatever it wants if it really wants to, but they are still separate organizations. And the people at Microsoft and the people at OpenAI have different ideas on how a chatbot should function, which is why there are some major design differences between OpenAI's ChatGPT and Microsoft's Bing chat. Anyway, once I tried Bing chat and found it could access up-to-date information and was a lot smarter than the free version of ChatGPT, I pretty much completely switched from ChatGPT to Bing chat. Each has different strengths and weaknesses but I prefer Bing chat. Google also has its own similar chatbot they developed independently called Google Bard. Google Bard is slightly dumber than the free version of ChatGPT, but its information is up-to-date and it can look things up online, and it isn't mentally unstable like Bing chat. Google Bard is also willing to accept it if you correct its mistakes, like ChatGPT, and unlike Bing chat, and it has better memory than ChatGPT as far as remembering the current conversation. So I prefer Google Bard to the free version of ChatGPT, just a little bit, but like Bing chat better than either of them. I think ChatGPT Plus with GPT-4 and plugins with the Wolfram Alpha and web search plugins is obviously the best option, the one that is smartest and gives the most accurate answers, but it costs money, and you also have to get on a waitlist for the plugins. Regardless of what AI chatbot you use, it will still sometimes make up wrong answers, because they all do that, it's just a fundamental problem that occurs with this architecture of AI. People are working on different solutions to making that happen less often, but it is really unlikely that that problem will be eliminated any time soon.

I had the same issue. I was doing research on manufacturers of machines in my line of business (basically my competitors). It gave me a list of manufacturers I never heard of. When asked about our company (leading company in this line of business by the way), it confirmed, that this company indeed is producing the machines in question and provided more details. However, all the details were wrong, including the web address, the country and the physical address. I searched the address provided by ChatGPT and it was an abandoned house in Italy in the middle of nowhere. When I asked ChaGPT whether it just made all this up, it apologized for confusion, but insisted, that everything is OK. I then opened a completely new chat and asked directly about my company, without providing any context. It said different things about the company and provided another, again wrong, address. This time, the address did not exist at all on Google Maps.

Yes, it's terrible, it gets so much wrong, that it takes time to validate the information provided. Plus no references either. Worse than useless.

Great. I was waiting for sb. to step on that artificial dragons tail. There is so much discussion about the way we will have to adjust our work routine and so little concrete about applying healthy skepticism about results. Why should an AI trained on human writing magically overcome the inbuilt human flaws in our collective academic output? Up until now, It's a mirror after all.

If you don't mind, can I get the scanned file of the mouthpiece?

Nice try with the lady in the thumbnail 😆 jk, definitely high quality content, made me miss my student life

Hello sir i have one proble... when i import scan stl file in creo ...its size increases in inches automatically in creo software.... what will do for this problem... can you tell me plz

Where is this institution

Underrated subject

Hello Sir..this was really interesting and truly had a load of big takeaways for me..I am actually working on the Headlights and Body Panels for a vehicle, and the work consists a majority of Surface Modelling. I am in the process of learning new techniques and rules about it, and this content will surely bring a confidence boost to mindset from now. So it would be really helpful for my upskilling if there was more content like this. If there is any way I can view your content elsewhere, please do let me know if that's not a nuisance to you, Sir. Thank you so much😁..!!

Sir I have install crack version of ansys 2021 but topology optimization tool is missing in my toolbox I have check it in a view/customise but topology optimization tool is not their can plz help me to get that????

thank you, this has helped me understand how to do topology optimization.

Good programs offered by DAAD and TGGS. Nice information to share 👍

:):) Great

Thanks for this video, it taught me how to topology optimization in Ansys :)

Great tutorial showing your technical surfacing workflow! Would it be possible for you to show further surface development techniques for the vehicle? Thanks!

Great tutorial sir.

I got to know Mr. Alit Putra since when I attended TGGS as a student. I regard him as my roll model and I think that I lowkey try to follow his path.

creo is unnecessarily complicated compared to solidworks.