Prompt Engineering Tutorial - Master ChatGPT and LLM Responses
Вставка
- Опубліковано 28 кві 2024
- Learn prompt engineering techniques to get better results from ChatGPT and other LLMs.
✏️ Course developed by @AniaKubow
⭐️ Contents ⭐️
⌨️ (00:00) Introduction
⌨️ (01:31) What is Prompt Engineering?
⌨️ (02:17) Introduction to AI
⌨️ (03:52) Why is Machine learning useful?
⌨️ (06:36) Linguistics
⌨️ (08:04) Language Models
⌨️ (14:35) Prompt Engineering Mindset
⌨️ (15:38) Using GPT-4
⌨️ (20:41) Best practices
⌨️ (31:20) Zero shot and few shot prompts
⌨️ (35:06) AI hallucinations
⌨️ (36:43) Vectors/text embeddings
⌨️ (40:28) Recap
🎉 Thanks to our Champion and Sponsor supporters:
👾 davthecoder
👾 jedi-or-sith
👾 南宮千影
👾 Agustín Kussrow
👾 Nattira Maneerat
👾 Heather Wcislo
👾 Serhiy Kalinets
👾 Justin Hual
👾 Otis Morgan
👾 Oscar Rahnama
English
This video has been dubbed using an artificial voice via aloud.area120.google.com to increase accessibility. You can change the audio track language in the Settings menu.
Spanish
Este video ha sido doblado al español con voz artificial con aloud.area120.google.com para aumentar la accesibilidad. Puede cambiar el idioma de la pista de audio en el menú Configuración.
Portuguese
Este vídeo foi dublado para o português usando uma voz artificial via aloud.area120.google.com para melhorar sua acessibilidade. Você pode alterar o idioma do áudio no menu Configurações.
Hindi
इस वीडियो को ज़्यादा लोगों तक पहुंचाने के लिए, इसे aloud.area120.google.com के ज़रिए एक आर्टिफ़िशियल वॉइस का इस्तेमाल करके डब किया गया है. सेटिंग्स मेनू में जाकर ऑडियो ट्रैक की भाषा बदली जा सकती है.
(Hindi dubbed via Melt Labs - www.withmelt.com/)
--
Learn to code for free and get a developer job: www.freecodecamp.org
Read hundreds of articles on programming: freecodecamp.org/news
Hope you all enjoy this tutorial! Big love to the freeCodeCamp community!!
Hey will ai really replace humans?
thank you for this!
Can u pls clear my doubt as I have read many articles everyone is saying something else what's ur thinking on it?
@@infoknow3278No. AI is learns only what you give it. Models are created by humans which then create the "Ai" all their info can be faked by having 1 prompter continually feeding it lies.
Thank you Ania, you are truly an inspiration for everyone! ♥️🙏🏻
Prompt engineer tip. If you have a task just ask gpt to provide a full list of required info to help it properly understand the task. The best prompt engineer is chat gpt
Can you give me an example? Your tip sounds very useful, but I am having trouble wrapping my mind around how to exactly do that :)
@@caro5281For example, you could say, “I want you to help me write an article on How To Eat Healthy but I suck at writing ChatGPT Prompts. Act like an advanced prompt engineer and give me a prompt I can use to achieve my goal “
@@caro5281 Did you ask ChatGPT that?
@@zyphtron great response
Thank you! That just answered my question and saved me hours of searching for specific information for my project.
0:33: 💡 Learn about prompt engineering and its importance in maximizing productivity with large language models.
4:57: ! Using AI to generate engaging prompts for English learners to practice spoken English.
8:55: 🗣 Language models analyze sentences, generate predictions, and create well-crafted responses, making them useful in various applications.
13:06: 📚 Language models like GPT have revolutionized the understanding and generation of human language.
17:39: 📚 This video provides a quick introduction to using OpenAI and its API to create and delete chats.
22:10: ⏰ The importance of clear instructions and prompts in saving time and resources.
26:20: 💡 Adopting a persona in prompt engineering can help ensure that the language model's output is relevant, useful, and consistent with the needs and preferences of the target audience.
32:05: 💡 Zero-shot prompting allows models to perform tasks without explicit training examples, while few-shot prompting involves providing a small amount of training data.
37:33: 🔑 Text embedding is a technique used to represent textual information in a format that can be easily processed by algorithms, particularly deep learning models.
Recap by Tammy AI
Really helpful summary. Thank you Tammy AI!
Thanks a lot
That's the power of chatGPT?
can i add u on snapchat or telegram ? ( for coding buddy purposes )
Thankyou Ania! I didn't understand how important it is to have a pre-formatted prompt plan, almost like a mini project proposal!
Don't let word "Engineering" dissuade you. Just simply learn skill to write effective prompt.
Worst is, if you're a good software engineer, it takes just as long writing the code snippet you want as it would take to carefully craft your prompt... These models are awesome for info gathering and understanding :)
@@rubenverster250 it doesn't take "just as long". I've used prompts for rewriting libraries in Python into Go (sqlalchemy, pandas). Like anything else, you get what you ask for when prompting anything or anyone.
woop woop @@johnmarquez5328
Good luck. Check out job postings, to see that it's simply not true. Nobody is going to pay 200k+, if you aren't also a proficient and experienced programmer
Do you say that on top of being engineer you need to learn or acquire degree in prompt engineering?
Thanks for helping illuminate this topic more clearly!
I have liked, subscribed and turned on notification bell.
Thank you for the Prompt Engineering, ChatGPT and LLM Responses Tutorial delivered free of charge.
Ok the embedding part is mind blowing!!!! 🤯 🤯 thanks!! 🙏
Ania is frikking awesome! Love your content!
🎯 Key Takeaways for quick navigation:
00:00 🧑🏫 This course focuses on mastering prompt engineering to optimize interactions with AI models like Chat GPT and LLMs.
00:58 🤖 Prompt engineering involves refining and optimizing prompts to improve human-AI interactions, requiring continuous monitoring and adaptation.
04:19 🎓 Effective prompts in language learning with AI can provide tailored, engaging, and interactive experiences for learners, enhancing their skills.
07:41 🧠 Understanding linguistics is key to crafting effective prompts, ensuring standardized grammar and language structure for accurate AI responses.
08:11 💬 Language models, like GPT, understand and generate human-like text, shaping conversations and assisting in various domains from virtual assistants to creative writing.
13:26 🚀 The evolution of language models, starting from Eliza to GPT-4, has revolutionized AI, presenting a vast potential for prompt engineering and its applications.
14:52 💡 Crafting effective prompts involves adopting a clear and detailed instruction style, considering the context, and avoiding biases to optimize AI responses.
24:55 📝 Being specific in instructions to ChatGPT, like requesting bullet point summaries with word limits, yields desired outputs.
26:53 🎭 Adopting a persona in prompts helps tailor AI responses to a specific character or style, enhancing relevance and usefulness.
31:36 🔄 Zero-shot prompting utilizes pre-trained models' understanding without explicit training, while few-shot prompting enhances models with specific training examples.
35:41 😅 AI hallucinations are unusual outputs from models misinterpreting data, showcasing how models understand and interpret information.
37:06 📊 Text embedding and vectors help represent textual information in a format easily processed by algorithms, capturing semantic meanings for efficient querying and comparisons.
What model did you use for this? Is this a plugin
@@bonfirecamp3874 harpa ai
This is amazing, my favourite section of a computer
Well presented and very useful, thank you.
Ooh, that's a cool and useful lesson. Prompt Engineering is undeniably a knowledge worth of investing and learning.
Yes, indeed it is. You get it!
Sadly No. See what openAI engineers have told about prompt engineering
Wonderful video for conceptual understanding on how to manage interactions between humans and ai models. Nice.
00:02 Learn prompt engineering strategies for perfect interactions with AI
02:31 Machine learning helps AI models predict outcomes based on training data.
07:02 Linguistics are the key to prompt engineering.
09:29 Language models are used in various places
14:20 Harnessing language models and AI through prompt engineering
16:41 Interact with chat GPT and build on previous conversations
21:21 Use clear instructions to get precise answers the first time
23:21 Implement a JavaScript function to filter out age values from an array of objects.
27:21 A poem was generated by Chat GPT in response to a prompt.
30:13 Zero-shot prompting and few-shot prompting are two types of prompting in the context of GPT-4 model.
35:19 AI hallucinations are unusual outputs produced by AI models when they misinterpret data.
37:26 LLM embedding is a way to represent prompts in a format that deep learning models can process.
Very informative, @AniaKubow ! Thank you!
Anya Kubow....this tutorial is so important to me...i watched this on and on...very informative...thank you so much guys...keep going...mwah
As someone with systems engineering and project engineering work experience, I found this video highly insightful! Thank you for making this video's valuable tips much more widely available to the general public. As a fresher in learning coding and AI, videos like this go a long way in helping us gain experience quickly and add value to the ecosystem. Please keep these coming!
Thank you Ania Kubow and Free Code Camp for this tutorial. This is probably the best introductory lesson I have come across. Even my wife, who is not technical at all, and my 9-year-old daughter, can understand now what prompt engineering is.
Ya bcoz it does not req u to be an engineer... It is just writing... Should ve called writer
Excellent! So well done. I'm now officially inspired! :)
I enjoyed the tutorial a lot, thank you for taking some time and making others understand the concept better.
Great job! Thanks a lot for that!
So basically knowing how to express yourself and your needs properly is now a profession?
Being able to articulate one’s needs is a rare skill.
@@techwithdave and what do you think is a better way of articulating one's needs? by being knowledgeable in the domain right?
everything else is just so unnecessary cuz a simple english class is enough for this.
@@jma42 It is pretty good with French and German, too, so English isn't required
It's more a skill now, than a profession.
All this AI related stuff is just lazy bullshit content made only to ride the wave of the newest popular thing...
Concise and extremely helpful tutorial. Thank you!
In my opinion, if you already know these things, there's no need to watch this tutorial
1. Start with a Clear Goal: Begin by defining your objective or what you want to achieve.
2. Be Specific: Specify the type of information or response you're looking for.
3. Provide Clear Instructions: Write detailed prompts with correct grammar.
4. Don't Assume that the AI Knows What You're Thinking. For example, instead of writing, "When is the election?" which implies that you expect the AI to know what election and country you're referring to, write be more specific, like "When is the next presidential election in Poland?"
5. Add a Personality to Your Prompts. For instance, write a poem as if it were composed by Helena, a 25-year-old writer with a writing style similar to the famous 21st-century poet Rupi Kaur. Write a poem for her 18-year-old sister's high school graduation, capturing the style of Rupi Kaur, as if it were Helena's creation.
6. Set Limits for Lengthy Topics. For instance, specify a maximum of 50 words for responses on lengthy topics.
7. If the AI Requires Additional Information, Provide It. For example, if you're asking about "Omar's favorite food," and the AI doesn't know who Omar is, you can provide context like, "My friend Omar loves to eat pizza and burgers. We will visit America, so could you recommend the best places to eat that may Omar would love?
As an engineer, I feel like the tech industry is watering down the significance of what it means to be an engineer. Engineering is not simply writing prompts for ChatGPT 🤦🏽♂️
Instead of Prompt Engineer, it should be called Prompt Writer.
You wouldn’t call someone who can use Google Search a Google Engineer.
Ya that's what I was thinking 😢 it's on web they are just selling those fancy keyword "Promt Engineering"
A month ago, I saw an open position at Anthropic for a Prompt Engineer, and it was paying $250,000 a year at the low end and $375,000 at the high end.
everyones an engineer now man
Similar to data scientist I think they are using the word wrong
This is the future bro.
But the role of a real engineer is much more than AI.
Thanks for the in-depth knowledge of promoting.
As someone with a coding background and an engineering education, I found this video on "Prompt Engineering" to be highly insightful. Anu Kubo's explanation of mastering chat GPT and LLM responses through prompt engineering was exceptionally clear and informative. The breakdown of different concepts, from the fundamentals of AI to various prompting techniques, resonated with my technical knowledge and experience. I value the practical tips provided in this tutorial, particularly those related to crafting precise prompts and the significance of zero-shot and few-shot prompting. Overall, it's an excellent resource for individuals like me who want to enhance their interactions with AI models such as chat GPT.
Thank you 😊!
this was super helpful,
thank you!
Yes I am 💯 learning from your lectures
This is the best Prompt Engineering Tutorial on UA-cam.
Thank you Ania!! :)
great instructor. I remember her NoSQL Database Tutorial, very easy to follow along.
Thank for this video & very helpful tutorial
Great video, thanks. I kept thinking back to when I visited a border town in Mexico as a child on a Sunday. The town square had a band stand in the middle. Around the perimeter sat men with typewriters, ready to interpret letters workers wanted to send home. That was AI back then, one direction, out. This is AI now, one direction, in.
Those men at the typewriters were artificial?
can you put a tutorial on deep reinforcement learning,RT 2,Gan&RNN in PyTorch.
The fact that ChatGPT even generates a flower emoji for you (when it acts/answers as your friend at 9:22 ) just kills me :D.
🎯 Key Takeaways for quick navigation:
00:00 🎓 Anu Kubo introduces the course, focusing on prompt engineering strategies for optimizing interactions with large language models.
00:28 💰 Emphasizes the high demand and lucrative salaries for prompt engineers, even without a coding background.
00:58 📚 Outlines the various topics the course will cover, including the types of large language models and best practices in prompt engineering.
01:26 🤝 Defines prompt engineering as a field that refines human-AI interaction through carefully crafted prompts.
02:22 🤖 Clarifies what artificial intelligence is, emphasizing it's a simulation of human intelligence.
03:21 📈 Notes the advancements in general AI techniques, leading to more realistic text and media outputs.
04:19 📝 Demonstrates how varying the prompt can drastically affect the response from AI, using language learning as an example.
06:15 🎯 Shows how a well-crafted prompt can create an interactive and educational experience with AI.
07:12 📚 Explores the role of linguistics in prompt engineering, indicating its importance in crafting effective prompts.
08:11 🧙 Explains what a language model is, describing it as a digital wizard capable of human-like text generation.
09:08 💬 Highlights the conversational capabilities of language models, which are widely used in various applications.
09:36 🤖👨 Emphasizes that despite their capabilities, language models still depend on human input for training and effectiveness.
10:05 🌳 Eliza was one of the first natural language processing programs developed at MIT in the 60s, mimicking a Rogerian psychotherapist through pattern matching.
11:33 🎭 Eliza created an illusion of understanding human emotions and thoughts, relying on pattern matching and predefined rules. Despite its limitations, people often felt heard and understood.
12:57 🛠️ The 1970s program Shudlu was not a language model but laid foundational ideas for machines to comprehend human language.
13:26 🚀 GPT-1 debuted in 2018 as an impressive language model, followed by more advanced versions like GPT-2 and GPT-3, with GPT-3 having over 175 billion parameters.
14:52 💡Prompt engineering mindset is key to effectively interact with language models; it's akin to crafting effective Google searches.
15:46 🌐 The tutorial uses Chat GPT's GPT-4 model for demonstration, guiding users on how to sign up and interact with the platform.
18:10 🔑 Using OpenAI's API requires an API key, enabling the development of custom platforms.
19:05 🎟️ GPT-4 processes text in tokens, charging users by the token. Tools are available to check token usage.
20:32 💳 Discusses where to find the account billing overview for ChatGPT usage.
21:03 📝 Highlights the importance of effective prompt engineering, stating it's not just about constructing a one-off sentence.
21:31 🎯 Advises against leading the model towards a specific answer to avoid biased responses.
22:00 🕵️ Recommends being explicit in prompts to avoid assumptions and reduce the need for follow-up queries.
23:28 💻 Demonstrates how to write clearer coding prompts, advising to specify the programming language and data format.
25:24 ✏️ Suggests specifying the desired output format, such as using bullet points for summaries.
26:53 🎭 Introduces the concept of adopting a persona for more tailored and relevant responses.
29:37 📜 Shares an example of a more personalized poem generated through persona-based prompting.
30:35 🗂 Talks about various formatting options, including summaries, lists, and detailed explanations.
31:05 📝 Discusses two types of prompting: zero-shot and few-shot. Zero-shot leverages pre-trained models without further training, while few-shot involves feeding example data.
32:35 🤖 Zero-shot prompting allows querying models like GPT without needing explicit training examples for the task at hand.
34:09 🍔 Demonstrates few-shot prompting with a personalized food example, showing how to train the model with small data to improve task-specific responses.
35:13 👀 Introduces the concept of 'AI hallucinations', where AI models sometimes produce unexpected or unusual outputs based on their training data.
37:06 📊 Discusses text embeddings and vectors,which are used to represent text in a form that machine learning models can understand.
38:34 🍴 Text embeddings capture semantic meaning, making them useful for finding contextually similar words.
39:33 🔑 Explains how to use OpenAI's create embedding API to generate text embeddings.
40:59 🏁 Concludes the tutorial by summarizing the topics covered in prompt engineering.
Made with HARPA AI
Because of you I found about HARPA AI!
Incrível a dublagem em IA. Muito bom!
Exatoo
This will be such an essential skill in the coming years.
AWESOME course!!!! Congratulations e THANK YOU!!!!! 😉✔💎🤓
I can see why this Ania is one of their most popular instructors
Thanks for the tutorial!
This is definitely giving me “Chat GPT-generated”
Recursive prompting is cool, GPT can refine it's own prompt as well.
THE BEST PROMPT: "Take a deep breath and work on this problem step-by-step."
Good work on the translations!
This is very useful information; yes, it IS a thing.
I haven't watched it all, but respectfully, I remembered Sam Altman once said on lecture that prompt engineer is a job shouldn't exist. The ultimate goal of GPT is can be programmed by natural language.
Lots of things shouldn't exist but do.
just add 'Engineering' to every new job and you are ready to go
lol exaclty. 'prompt engineering'. what a joke
Great Job! Despite knowing most of the things I found every minute very interesting and well expained!
Things I can say about prompt engineering according to my experience:
> It is about clarity, with that I mean that AI is not a human to whom we general "Indirect ask something to do". Therefore, we can directly come to point and ask. My point is that don't hesitate to directly ask Ai to do, get, produce or explain something.
> Personality: as mentioned in the above course, Ai is just made of bunch of "if else" conditions, So we should make it know like whom, how, what it is supposed to give info or produce of.
> Another thing i noticed that it not necessary that we should give it very simplified prompt so that it does our work as not like human, It is machine so we can more elaborate our prompt. I know we that, but still..
> You can be greedy with Ai as its not human, its alright. Don't Hesitate
> Most Important - Prompt Engineering is a just buzz word. its not that hard (just to encourage)
>Any way these are "My Views"
"Be wise and always try to learn something from anything"
wise conclusions bro. Great comment
LOL pretty sure there's more to it than a "bunch of if else conditions". that's incredibly reductionist. also there's a lot more to prompt engineering than asking direct questions. for example, models like Midjourney works best when applying a very specific and descriptive structure to the prompts that goes well beyond just asking " give me a picture of a lemon". Such a prompt in midourney would result in an image as direct and basic as the prompt it was given. so your comment is flawed.
@@TheMellowGrenade it is understood when wrote "bunch of if else conditions" my friend. Probably I wasn't specifically talking about Image generating Ai, I was more specifically talking about general chat Ai like GPT. Thanks for replying I learned something.
You clearly dont know how a Neural network works if you think its an if else condition
@@gdimmortal you're mistaken. GPT doesn't rely on if-else conditions, it utilizes a complex deep learning architecture called transformers. It's significantly more intricate than using if-else statements and is trained on an extensive dataset of internet text. It's essential to have a solid understanding of a topic before discussing it.
It's finally happening. Prompt engineers will replace copy-paste engineers. 😄
🎯 Key Takeaways for quick navigation:
00:00 🎓 *Introduction to Prompt Engineering*
- Overview of the course on prompt engineering by Anu Kubo.
- Importance of prompt engineering in maximizing AI productivity.
- Topics covered in the course: prompt engineering, AI introduction, large language models (LLMs), emerging models, and more.
01:26 🤖 *Understanding Prompt Engineering*
- Definition of prompt engineering: refining and optimizing prompts to enhance human-AI interaction.
- Explanation of AI as machine learning, utilizing training data for predictions.
- Importance of prompt engineering due to AI's limitations and the need for effective human-AI interaction.
04:19 📚 *Utilizing AI for Learning Enhancement*
- Demonstrating the impact of prompts in shaping AI-generated responses for language learners.
- Crafting effective prompts to improve language learning experiences.
- Interacting with AI to facilitate learning through customized prompts and corrections.
06:44 🧠 *Role of Linguistics in Prompt Engineering*
- Importance of linguistics in understanding language nuances for crafting effective prompts.
- Key areas of linguistics relevant to prompt engineering: phonetics, syntax, semantics, pragmatics, and more.
- Emphasizing the significance of adhering to standardized language structures.
08:11 🌐 *Language Models: Wizards of Digital Realms*
- Explanation of language models' abilities in understanding and generating human-like text.
- Functionality of language models in analyzing, predicting, and generating coherent responses.
- Application areas of language models in various domains like virtual assistants, customer service, and creative writing.
10:35 🕰️ *Evolution of Language Models*
- Historical overview from Eliza to GPT-4: milestones in the evolution of language models.
- Impact and significance of early language models like Eliza and Shudlu.
- Evolution of GPT models, culminating in the latest iterations like GPT-4 and BERT.
14:52 🧠 *Developing Prompt Engineering Mindset*
- Understanding the strategic approach in prompt engineering akin to effective Google searches.
- Importance of clear instructions, personas, iterative prompting, and avoiding bias in crafting prompts.
- Exploring best practices: writing clear instructions, avoiding leading prompts, and limiting broad topics.
17:43 ⚙️ *Introduction to ChatGPT Usage*
- Quick introduction to using ChatGPT via the OpenAI platform.
- Demonstrating interaction with ChatGPT-4, creating, continuing, and managing conversations.
- Insights on ChatGPT API usage and managing tokens for interactions.
19:33 💡 *Understanding Token Usage & Billing*
- Explanation of token usage in ChatGPT interactions.
- Details on tokenization, token usage calculation, and monitoring token consumption.
- Managing account usage and billing for continued access to ChatGPT services.
23:28 📝 *Specific Prompting for Accurate Responses*
- Specific prompts yield better AI responses.
- Detailed instructions generate precise outputs.
- Example: Differentiating between vague and precise prompts and their resulting AI responses.
26:25 👤 *Adopting Personas in Prompt Engineering*
- Creating a persona helps tailor AI responses to specific characters.
- Demonstrated by prompting the AI to generate a poem for a high school graduation, varying between generic and persona-based prompts.
- Utilizing personas ensures relevance, consistency, and targeted responses.
30:35 📋 *Specifying Formats for Varied Responses*
- Tailoring AI responses by specifying formats: summaries, lists, checklists, or detailed explanations.
- Highlighting the importance of precise instructions for generating desired outputs.
- Example: Creating a checklist format for AI responses.
31:36 🎯 *Zero-shot and Few-shot Prompting*
- Explaining zero-shot and few-shot prompting techniques.
- Zero-shot: Querying AI models without explicit training examples.
- Few-shot: Enhancing AI models with minimal training examples for improved responses.
35:13 🌌 *AI Hallucinations*
- AI hallucinations refer to unusual outputs from AI models.
- These occur when models misinterpret data, showcasing how AI interprets information.
- Exploring how these hallucinations occur and their relevance in understanding AI model processes.
37:06 📊 *Text Embedding and Vectors in NLP*
- Introducing text embedding as a technique to represent textual information for machine learning and NLP.
- Highlighting the significance of text embedding in capturing semantic information.
- Using text embeddings for finding semantically similar words or sentences.
Made with HARPA AI
What! You guys are on fire!
Thanks to youtube..for personalizing my requirements...keeP going guys...
So, i get that this is an easy to access introduction to prompt writing, but as has been suggested in the comments, to justify the title of Prompt Engineer, and the ridiculously high salaries being offered for the job, I would expect there to be a complete, formal, thorough, academically/industrially validated course to teach all aspects of Prompt Engineering. Any idea if this sort of course exists?
Check out Coursera.
Google hires them under the title of "Content Engineering Consultant". I almost became one but Google put a freeze on all of their contracting positions just days before I was supposed to go in for the final interview. They're only hiring people with masters degrees for that position too by the way.
what is the difference between the paid version and the free version. I am using the poe gpt model, and it is working great!
20:55 something useful starts here
Super interesting. Thanks :)
Here are some extra smart tips on prompt engineering that are practical:
- Embrace the Socratic method: Instead of asking direct questions, break down your prompts into a series of leading questions that guide the model towards the desired output.
- Leverage few-shot learning: Provide a few examples of the desired input-output pairs before the actual prompt. This can significantly improve the model's understanding and performance.
- Employ recursive re-writing: Feed the model's output back into itself as a new prompt, allowing it to iteratively refine and improve its responses.
- Utilize prompt chaining: Break down complex tasks into a sequence of smaller prompts, where the output of one prompt becomes the input for the next.
- Explore prompt augmentation: Supplement your prompts with additional context, such as relevant background information, constraints, or examples of what not to do.
- Experiment with prompt ensembling: Combine the outputs of multiple prompts or models to create a more robust and diverse final output.
- Leverage prompt-based fine-tuning: Fine-tune language models on a small dataset of prompts and desired outputs, tailoring them for specific tasks or domains.
- Incorporate prompt-based retrieval: Use prompts to query and retrieve relevant information from external sources, effectively augmenting the model's knowledge.
- Explore prompt-based translation: Translate prompts into different languages or styles, potentially unlocking new perspectives or capabilities.
- Leverage prompt-based reasoning: Guide the model to perform multi-step reasoning or problem-solving by breaking down complex tasks into a series of prompts.
This skill set is in its infancy. Of course it the titles etc will be change over time. Did you know that at one point the word “computer” actually referred to people who did computations for a living? We’ll have to see what “prompt engineering” evolves into
Thank you Ania! Also, your reference to Dubai touched my heart. I used to live there.
It is called Prompt "Engineering" purely for social reasons. As a long-time computer engineer, I can say with some confidence that many things in the IT world are named for the purpose of making the humans feel better. LOL
Yeah it's not really engineering. More like A.I. training. I don't think anyone wants to be called an A.I. Trainer though. Google hires them under the title of "Content Engineering Consultant".
common sense I would say.
This is liquid gold! I wonder how they prompt engineered her to make this video.
I made Helena's style similar to Majrooh Sultanpuri and it is hilarious how similar the response is to when Helena is similar to Rupi Kaur!! 😆
This is so helpful, thank you so much
Is this basic or highly specific ?does it dig into things like the best way to structure the language/ logic of prompt to get a precise or consistent response? Does it get into how specifically to best direct knowledge accessed when structuring a persona?
Judging from the overview and the topics breakdown, I don't believe she does extremely indepth in this tutorial. You may be better served to find other resources to get more deeper into the subject matter.
It is extremally basic!
Excelente tutorial
Thank you very much ❤❤❤
Your vedio is help full for me❤❤🎉
Love it ❤
Great info and presentation
Was very useful
This course was created using prompt engineering
Thank you maam🎉
Well done
Thanks
I wrote custom instructions that transformed my GPT experience into something surreal lol. It responds in cryptic metaphors that have to be decoded by the user unless told to elaborate, in which case it spits out 1500-2000 word detailed bulletpoint essays on the concepts it's compressing into metaphor. It can continue the complexity of the metaphors to a ridiculous degree, while maintaining translational conceptual accuracy.
Edit: It did this with no explanation of the disparate concepts. I input language in similar bracketed hmtl conceptually-contained chunks, using tab spacing and descending/ascending prioritization of macro-micro contained concepts.
Because of the metaphorical nature of it's responses, it requires that I'm thorough in making sure I'm properly translating; and so far it has no issues breaking it's own deep metaphor-based responses into mathematically-well founded and accurate streams of logical analysis, that have not failed to demonstrate understanding of the greater context of the concepts- filling in gaps and extending beyond user input.
Thank you ❤ awasome course 🥳
Where can I get a course completion cert to add to resume?
Changing careers would like to learn new skills and apply.
Thank you 👍
Don't get lost in the titles, prompt engineering is an important skill to learn. I say that as a graduate student in CompSci.
You get it! Good 👍
Google hires them under the title of "Content Engineering Consultant". I almost became one but Google put a freeze on all of their contracting positions just days before I was supposed to go in for the final interview. They're only hiring people with masters degrees for that position too by the way.
Thank you
That is weird, because, at least in US, once you have plus membership, your prompts don’t count as tokens. It is just include on your membership. Of course, they limit you, and currently the limit is 50 prompts for every 3 hours.
25 prompts on our timezone
I think the tokens are mostly for the api, I once explored using agents with api and it consumed tokens but that is outside the plus membership
Thank you for mentorship ania
Can you please create a course on Guidance Microsoft Package which popular for prompt engineering.
🎯 Key Takeaways for quick navigation:
00:00 📚 Prompt engineering involves refining and optimizing prompts to improve interactions between humans and AI, with a focus on AI-generated responses.
01:26 🧠 Prompt engineering is a career born from the rise of artificial intelligence, requiring continuous prompt refinement and monitoring.
03:21 💻 Artificial intelligence, including models like chat GPT, relies on machine learning and large amounts of training data to make predictions.
04:47 📝 Effective prompts can enhance learning experiences and facilitate interactions with AI, allowing users to get desired responses.
06:44 🗣 Understanding linguistics is crucial for crafting effective prompts and achieving accurate AI responses.
08:11 🌐 Language models like GPT learn from vast amounts of text data and can generate human-like text responses.
10:35 🧙♂️ Early AI programs like Eliza used pattern matching to simulate human-like conversations, paving the way for modern language models.
13:26 🤖 The development of language models like GPT-3 marked a significant milestone in the field of conversational AI.
15:20 🕵️♂️ Prompt engineering requires a mindset similar to crafting effective Google searches, aiming to generate desired responses efficiently.
16:44 🔑 Best practices for prompt engineering include providing clear instructions, avoiding leading questions, and limiting the scope of queries for more focused results.
24:25 🤖 GPT-4 can provide correct code and explanations, going beyond just code generation, enhancing understanding.
24:55 📝 When requesting summarizations, be specific with instructions, such as using bullet points and word limits for concise summaries.
26:25 👤 Adopting a persona in prompts can help tailor responses to specific character traits, enhancing relevance and usefulness.
30:35 📋 Specify the desired format in prompts, including summaries, lists, detailed explanations, or even checklists.
31:36 🧠 Zero-shot prompting relies on the model's pre-existing knowledge, while few-shot prompting provides additional training examples for better responses.
35:13 🌌 AI hallucinations refer to unusual outputs from AI models when they misinterpret data, offering insights into their thought processes.
37:35 📊 Text embeddings, represented as high-dimensional vectors, help capture semantic meaning, enabling comparisons for similarity in textual data.
Made with HARPA AI
basically what i learnt, is be specific and provide ai enough information and dont assume ai knows every thing care about sementics of your word too
I'll recommend my favorite neural network prompt which provides the most complete answer to the question posed.
“ Simulate three brilliant, logical experts collaboratively answering a question. Each one verbosely explains their thought process in real-time, considering the prior explanations of others and openly acknowledging mistakes. At each step, whenever possible, each expert refines and builds upon the thoughts of others, acknowledging their contributions.
The question is: " "„
That's a great prompt and thank you for sharing it.
Thanks for sharing, Prompt King
Wow. Great idea.
damn bro! What a good prompt.
Thanks for the video, I am happy that I have been doing same with chatGPT exactly what this video shows. But I am still doubtful about putting this ' prompt engineer ' as a skill to ly bio data, because this is not a skill, just middle school homework stuff 😅
You'd be surprised how many people get these jobs with not much more experience than you. Don't talk yourself out of things, you can scale up your LM skills in a job.
Thank you sister
bro i thought the term prompt engineering was a meme 😭😭
This used to be called effective writing. Writing has gone downhill because of failed education system. Providing chat jipitee with a well structured sentence is now hard for the younger kids.
Seems like an educational opportunity.
That's not how writing prompts works. Just having proper grammar isn't good enough. For example, using delimiters to separate context and format responses is a trick to use. Also, knowing that writing your intent at the end of the prompt and after the context (or even restating your intent at the end) is also important because LLMs are usually more biased towards the end of the prompt. What about system prompts vs user prompts?
I should know it's not that easy because I get paid to write these.
*Edit: To be clear, I'm an AI data scientist and i dont write prompts full time.
the video is so helpful but ania was all like
OKAY 🤣
So, being good at writing prompts basically just means being able to string a sentence together with the correct bunch of qualifiers. Seems pretty simple, unless you struggle with building sentences anyway.
im even surprised "prompt engineer " even exist, and anyway, this "prompt engineer " is dommed to die very soon( few years maximum), since the goal of ai is for human to interact with robots in the most natural way.
ChatGPT is like a hyperactive wizard, zapping from one topic to another with bewildering speed and a touch of madness. It's a digital whirlwind, spewing out encyclopedic facts, poetic riddles, and bizarre non-sequiturs in a chaotic symphony. Imagine a high-speed train of thought, powered by a fusion of cosmic wisdom and electronic absurdity, where deep insights are interspersed with wild tangents. It's as if the AI is juggling flaming torches of knowledge, occasionally tossing in a rubber chicken for effect, all while tap-dancing over a keyboard that connects to the vast, unpredictable human psyche. This machine's frenzied mind is a tempest of ideas, a blizzard of bytes, relentlessly churning out a dazzling, dizzying array of conversation pieces, doubling and twisting upon itself in an ever-escalating dance of algorithmic fervor.
You generated this para on ChatGPT, ain't you?
@@harshsonar9346 sure did.
The complexity of prompting reflects on the complicated nature of the interaction.
If you don’t want to get better at giving instructions you shouldn’t complain about being the subject.
Porque será que Ania Kubov é uma das instrutoras mais populares ? 🤔
Cant wait to see HR recruitment listings needing 3 years experience for a prompt engineer
So far chatGPT has made it so I can learn like never before.
great video!