GPT-4 Prompt Engineering: Be Better Than 99% ChatGPT Users

[@ThePyCoach]

To try everything Brilliant has to offer-free-for a full 30 days, visit brilliant.org/ThePyCoach/. The first 200 of you will get 20% off Brilliant’s annual premium subscription.

[@GraydonSchwartz]

Getting 5 facts first and asking for an expert opinion was really helpful!

[@Sylfa]

Worth noting is that most of the improvements in ChatGPT (over GPT-3/InstructGPT) was aimed at overcoming the need to give examples and format it as a "complete this document" task so that it can instead answer questions.

[@dianadovgopol9087]

Wow, this GPT-4 prompt engineering stuff is insane! The fact that it can outdo 99% of ChatGPT users is wild. It's gonna be sick to see how it can be used in different fields. As someone who uses language models all the time, I'm pumped to see what GPT-4 can do. If it means we can get even better prompts, that's gonna be huge. Can't wait to see where this tech goes!

[@AlperBal]

he obviously got sponsored by openai

[@arturoblas]

This is useful, hopefully future language models won't require all this extra work to get around the current flaws.

[@smiley0227]

What would help as well is if ChatGPT would ask clarifying questions. One of the things I've done is propose a topic with my specific requirement, but then tell it to ask me 20 (or 10, or 5) questions before giving me a result. This has greatly increased the accuracy of the results.

[@jaywha]

Absolutely great advice and immediately useful for me even after watching other prompt engineering tutorials. Thank you!

[@kitokid86400]

Super interesting video with useful information. Taking time to learn how to write effective prompts will save load of time. Thanks PyCoach 😃Hope to see more content ke this in the future.

[@amithyel]

I find a mistake in your article/video and I think you should consider to rewrite or ...whatever you want to do my remark. You provided example of standard and chain of thought prompt. In your standard example the GPT model output answer is incorrect. But the problem is not the structure of the prompting. The issue is the missing or insufficient basic mathematical dataset, this learning process is unsupervised, GPT is not explicitly taught mathematical concepts, but rather discovers them by observing patterns in the data it is trained on. End of this process, GPT's understanding of mathematical concepts is emergent, derived from the statistical patterns it observes in the text data.

[@Sylfa]

There's no need to repeat facts in your prompt back to GPT, as long as it doesn't run out of tokens it'll remember the conversation so you can simply write "Use the above facts to…" The exception is if you've been going back and forth long enough for it to run out of tokens of course, at which point it'll forget earlier parts of the conversation. ChatGPT 3.5 is capable of 4096 tokens, on average a token is 0.75 words, this includes the response however, so it can literally start forgetting things while writing out a response. I'm not sure but I think you have half the tokens if you don't pay for access.

[@freemovieshub9607]

I'm using chatgtp for optimize listing of product for e-commerce business, so i get more traffic, and widely use for worldwide market research. It's really save my time and money. And I'm not sure it provides 100% result, but 85% somewhere it helpful.

[@tindejuv]

Awesome explanations PyCoach. Really inspiring ans useful info!

[@theinsideilluminater]

Suppose we want to write an article on blogging and we want to add some keywords to it, then what prompt should we use?

[@amartyagupta8255]

Make videos on Langchain

[@stevealexander8010]

So uhhm - you are saying that WE need to tailor our behavior in order to get what we want from ChatGPT - what kind of tool does that make you ?

[@rajtheo]

ask the question clearly on the prompt then there won't be any need for examples . ex: what are the airport codes for miami and boston.

[@AlperBal]

how much you got paid?

[@OSWALD569]

Question! Why do you regard AI as a child?

[@rightbrian]

Will ChatGPT 4 pass the touring test?

[@FreshmanProViral]

This is a great one, thanks. Please try to improve your english accent, if possible.

[@MsAriabll]

Why?

[@GiggleNuggetsZone]

@@MsAriabll hard to understand, thats why!

[@MsAriabll]

Everyone on earth has an accent, even you.

[@chandrasekhar01]

Welcome to KZbin 😅

[@SpaceCadet4Jesus]

What version of which AI engine did you use on the example question "The cafeteria had 23 apples. If they used 20 to make lunch and bought 6 more, how many apples do they have?" I asked the exact question to Google Bard, Microsoft Bing Chat, and ChatGPT a few minutes ago. Results: All three engines gave detailed calculations with the correct answer.

[@ahmadyasinalharits2464]

i use co-pilot right away, and it's usefule for more crative blog post. thanks for sharing

[@wHaTsThIsThInGdO]

Try getting on a hot air ballon with ponzi scheme names

[@EustaquioSantimano]

Very good video!

[@web3tutorials889]

HINT: If you open your mouth, it creates space for the vowels to be voiced properly.

[@lizzibutcher]

One of the most ignorant and useless comments I've had the misfortune to see for a long time. Does it make you feel good to be disparaging?

[@ArtiomNeganov]

Thanks for the video. It does make a lot of sense. It's funny how silly I was expecting too much from chatGPT w/o providing good prompts and descriptions. On the other hand, it means more typing :( P.S. "Reasoning" is /ˈriːzənɪŋ/, you pronounce it as /ˈriːs'onɪŋ/