Transcription
Let me tell you something you absolutely did not expect. When it comes to AI, you are actually not behind at all. You simply don't know how to use AI properly. Just by watching this 30-minute video today, your understanding of AI will surpass 90% of the people around you. If you can stick around until the end, you will enter the top 1% of global AI users.
I have personally tested over a dozen AI tools and thousands of prompts. I have also helped hundreds of overseas entrepreneurs use AI to become their intelligent agents. In this video today, I will mainly discuss four things with you. How does AI actually work? Secondly, what are the basic skills for using AI? I will also show you how to become an advanced AI user, putting you in an invincible position in this AI movement. Finally, I will show you how to make AI your weapon, so you no longer have to worry about AI replacing your position in the future. I am Henry from Australia. If you find this video useful, remember to click the subscribe button below. Also, turn on the little bell so you don't miss any of our future exciting videos.
Let's start with a very basic but important point: how does AI actually work? Simply put, AI consumes all the content you can see on the internet. Videos, podcasts, articles, books, including posts on various forums. Then, it breaks down all the content into tokens. Based on the content you give it, it predicts what the next segment of text should be. Therefore, all AI models are essentially not problem-solving masters, but are exceptionally good at prediction. You can think of it as a master of the idiom chain game. For example, if I tell AI "In front of the bed, the bright moonlight," and then "suspecting it is on the ground," it can very naturally complete the last word as "frost." This is the logic.
Many people have had similar experiences where you ask AI a question, and AI answers it eloquently, citing classics. But upon closer inspection, you'll find that the answer is entirely fabricated. This isn't AI deceiving you, but a very common phenomenon of AI hallucination. Because AI must give you an answer, even if it doesn't know the answer, it will create a seemingly correct answer for you through understanding the question and restructuring the language. It's like if I only give AI the first word of the poem we just mentioned, "bed." You see, at this point, AI gives many answers that seem reasonable, but none are what I want. Why? Because I gave AI too little information. AI can only make a lot of ineffective predictions.
So, I want to make one thing clear: AI is not as smart as we imagine. AI is just very good at prediction. The more ambiguous your prompt is, the more likely its predictions are to be wrong. What do more precise prompts look like? Remember one thing: the more information you give AI, the longer the prompt you give AI, the more accurate the output from AI will be. Many people actually don't know that when you're talking to AI today, the information content of the questions you ask it can be very large, so large that you can directly feed it the information of two or three books. But unfortunately, most people give it a very simple sentence and expect AI to perform magic. It's like I just gave AI the first word of the opening line of the poem. It's very difficult for AI to give you the entire line of the poem.
A while ago, an entrepreneur came to me for consultation. He accurately identified a market pain point and created an AI product with a strong selling proposition. However, he couldn't figure out the marketing strategy. He didn't know how to acquire customers, such as copywriting, product selling points, and sales scripts. He couldn't handle any of it. When I sat down to help him, I used the format I'm about to teach you. AI took less than 10 seconds to give a very good answer, including marketing copy, product sales pages, and sales scripts. All were designed. This model is the very famous AIM framework. It covers almost all the basic logic for conversing with AI.
What is A? It's Actor, meaning you tell the model what role it needs to play. I is Input, meaning the information and background you want to provide to AI for the question you're asking. M is Mission, meaning what you actually need AI to do. So, don't tell AI "Help me write a great video script." You need to tell AI, "You are now the world's top business consultant and YouTube marketing expert. You have helped hundreds of small and medium-sized business owners acquire customers through content and achieve stable transactions and revenue growth." This is the A, the role you want AI to play.
Next, let's talk about my role. I am a business consultant for overseas Chinese entrepreneurs. Most of my clients are already starting businesses or running businesses, but they encounter problems like very little traffic, difficulty in closing deals, and stagnant growth. I want to attract them to become my users through platforms like YouTube. This is your I, the problem you want to solve and your background. When we are solving a problem in this section, you can give AI all the relevant information, such as your marketing copy, meeting minutes, or even product descriptions. You can throw it all at him at once.
Next, let's talk about the mission. "Please help me design a YouTube video plan that can bring me new customers, including an attractive video title, a captivating opening, 3-5 core content points, a simple case study, and a natural call to action." I will also say, "Your mission is to help me create a video that not only gets watched but also helps me acquire targeted users." This is the M, the task I give to AI. You can also ask AI to present the answer in your preferred format. For example, my favorite is to have AI output in a table format, as it allows me to see at a glance how different answers perform in different dimensions. You can even ask AI to package the answers directly into PDF or Excel charts, or even images.
When you use the AIM prompt framework, you'll find that the core here is to turn prompts into a systematic structure. You can use this framework in almost any environment. You'll magically find that the results AI gives you will be 5-10 times better than before. This is because AI has learned your language. Remember, the quality of the final answer AI outputs depends on the quality of the information you provide to AI when assigning it tasks.
After understanding this basic information, let's discuss a very perplexing issue: which AI software should we use? There are really too many AI software options on the market now. I saw a news report that in the past year alone, the number of AI software launched exceeded 10,000. This can give you the illusion that you'll never catch up with AI. You've just learned to use one, and another one comes out. If you switch to another software, you have to relearn. But if you don't switch, you feel like you're being left behind by the whole world.
My advice here is to choose one model and study it in depth. You will find that once you can master one AI software completely, you will naturally be able to use other AI software. Learning AI is like learning a musical instrument. If you learn a little piano, a little guitar, and a little saxophone, the final result is that you won't master any of them. But if you choose one instrument and study it in depth, you'll find that other instruments are not difficult for you at all. Because by learning that instrument, you've mastered the most important elements behind music, such as music theory, pitch, how to read sheet music, and so on. So, AI tools are the same. Master one first, then explore others. Those who constantly switch AI tools are just delaying their AI learning progress. So there's no need to be anxious about them.
Today, I will quickly introduce you to the three most mainstream AI software on the market: Claude, Gemini, and ChatGPT. I will help you quickly understand which one is right for you.
First, let's talk about Claude. It's the most stable and currently has the best reputation among AI large models under Anthropic. Among all AI large models, I don't know why, but Claude is the best at processing Chinese content and copywriting. So, if you often use AI to help you write copy, revise articles, or write emails, you'll find that much of the content written by AI is recognizable at a glance – repetitive sentence structures, loose logic, and a lot of filler. However, the copy written by Claude is very concise and logically clear, so the cost of revising Claude's copy is very low. Therefore, if you need to produce a large amount of text content in your work, I think Claude should be your first choice.
I also want to emphasize one point: Claude is the AI that is best at guiding people to think actively. Many AI models can't do this, but Claude can. When you ask Claude a question, and Claude feels that the information or background you've provided is insufficient, it will proactively ask you. For example, if you ask it to write copy, it will ask you, "Who is your target audience?" This ability is really important when we deal with complex tasks, because often when we ask questions, we don't even know what problem we are trying to solve.
Second, Gemini can be said to have the strongest background among all AI platforms, because its father behind it is Google. We just said that the essence of AI is to consume all the content you can see, right? You should know what Google's essence is. Google is a company that crawls and searches for information all over the internet. So, the search function itself is embedded in the DNA of Gemini. When you converse with Gemini, you can directly paste web links, news articles, or even YouTube video links. Not only can you ask it to analyze news for you, but you can even ask it to perform linked searches, meaning to search for more related information. Also, if you are already using Google products like Gmail, Google Drive, or Google Calendar, getting started with Gemini will be easier for you. It can not only connect your Google suite of software but also assist you when using several Google software simultaneously, such as helping you write emails. Therefore, Gemini is definitely the first choice for users who frequently perform a lot of searches and are already using the Google suite.
Third, ChatGPT. Many people think ChatGPT is no longer good, but the real data shows that ChatGPT still has the largest number of basic users and is far ahead. I think ChatGPT's biggest advantage is its comprehensiveness. It can draw, search, write code, have voice conversations, and has projects like intelligent agents and an app store. It can be said to have the widest range of skills and the most comprehensive functions. But perhaps precisely because of this, each skill has not been developed to its extreme. Many people complain that the answers given by ChatGPT are often strange. I think ChatGPT is actually most suitable for beginners who are just starting to use AI, because its current functions are very comprehensive, and you don't need to constantly switch between several software.
After discussing these three software, I still want to remind everyone not to constantly switch between these three software. Use one today, another tomorrow, and end up only mastering a superficial aspect of each. This is not efficiency, but creating anxiety for yourself. Truly capable people understand AI itself, not just the language model.
You've probably grasped the previous content well by now. Congratulations, you've entered the top 10% of AI users. But if you can continue watching and take action, you will become the top 1% of AI users. Next, let's talk about advanced AI usage methods, allowing you to become a master of AI invocation and truly have AI work for you. This part will be a little more complex than before, but very interesting. Don't go away, because it's very exciting.
There are actually two levels of using AI. The first level is that you tell AI what to do, and AI follows your instructions step by step. People who use it this way usually let AI do 80% of the work and they do 20%. This sounds fine, but in reality, using AI is very tiring. This is the state of most people using AI now. Let me give you an analogy. It's like you want to go to a supermarket and don't know the way. So you ask a passerby on the street. The passerby gives you instructions: "Turn left ahead, then turn right after the traffic light, and you'll see a post office. The supermarket is opposite." AI is indeed helping you, but you'll find that you still have to complete most of the operations.
The more advanced level is to let AI become your navigator. You just need to tell AI your goal, and let AI design and plan how to achieve that goal. It's like when we use GPS. We just need to enter the address we want to go to, and then the route planning is up to the GPS. We just need to follow it. At this point, you'll find that you've gone from being a driver to a passenger.
How to do it specifically? Let me give you an example. Suppose you want to create a marketing plan. First, we still give AI a role, an actor. We can tell it, "You are now a senior business consultant." Second, we clarify what we actually want. "Let AI give me a marketing plan on how to turn strangers into customers willing to pay high unit prices." But the third step is a bit different. We don't need to rush AI to give a plan, but let AI interview you instead. We will tell AI, "You don't need to rush to give the final answer. You can ask me questions until you fully understand my core business, my customers' pain points, and their true underlying needs. Let's work together to outline a clear execution path." At this point, you'll find that AI will ask you many questions.
Here's a little trick: answering AI's questions one by one is too tiring. So, when I converse with AI, 99% of the time I have the voice input turned on and speak directly to it. You'll find that your efficiency will skyrocket. Then, we just wait for AI to give an execution plan. If you're still not satisfied with the plan AI gives you, what should you do? It's okay, you can ask AI to continue optimizing. You can tell it, "I need a more precise and brilliant answer. I don't think your answer is right or satisfactory, but I can't say exactly where the problem is. Please ask me 5 more questions based on the current situation to further improve the quality of the answer." This process is, on the one hand, constantly letting AI understand my true intentions, and at the same time, it's also letting AI think about where its answer path might have gone wrong. This is the difference between teaching AI to work and AI leading you to work. You don't need to think of the answers; you just need to answer the questions AI asks.
This method is not new. Most programmers solve problems this way. When they write code, they often don't know what the final desired code answer looks like. So they describe a result and let the program run itself. Now, this logic has entered the AI circle and is becoming a mainstream way for experts to use AI. Letting AI become your navigator is a very powerful way of asking questions, but it's not perfect yet. Because even if your way of asking questions is perfect, AI still doesn't understand you well enough. It doesn't know who you are, what you do. Every time you chat with it, it has to spend a lot of time getting to know each other again.
If you want to improve the efficiency of using AI and turn AI into your exclusive assistant, you need to give it a background prompt about yourself. Once, during a team meeting, I was using a team member's computer to converse with AI, and I found that the quality of the answers given by his AI was much worse than mine. I was puzzled, and later researched it to find out the problem. Their AI simply didn't understand them, didn't understand what their work was, didn't understand who their users were, and they had never set any standards for AI's answers. So the answers AI finally gave were generic answers that anyone asking it would receive, which couldn't solve problems at all.
What is a background prompt? It's a user manual about you. Imagine a book called "Henry from Australia User Manual." It would state where he is from, what business he does, what products he sells, what kind of people his customers are, his personality, what kind of answers he needs, and how he usually expresses himself when he speaks. Once AI has this manual, all the answers AI gives will be tailored to you. You will no longer receive generic answers from AI.
Remember the "I" in the AIM format I mentioned earlier? That's the input information. If you have your own background prompt in the future, you can save a lot of time when inputting information. You just need to prepare all your background information once and feed it to AI. In the future, when AI answers your questions, it will automatically call it up. You won't have to explain who you are to AI anymore. AI will remember you for life.
So, how to achieve this effect? It's very simple, with just three steps.
Step 1: Let AI interview you. You can say to AI, "I want to create a background prompt about myself. What is my identity? What kind of work do I do? Please ask me all the background questions you need to know about me."
Step 2: Turn on your voice input and patiently answer its questions one by one.
Step 3: After AI generates your background prompt, carefully check it. Sometimes AI might misunderstand, so correct any misunderstandings until you feel it's perfect. Then, ask AI to save it as a PDF file. You can also generate different manuals based on your different identities, such as "Henry from Australia Self-Media Master Background Prompt" and "Dad of Two Children Background Prompt," etc. This way, no matter which AI you use, you can directly give it your user manual about yourself, and it will immediately know who it's talking to. So, if you feel that your conversations with AI are inefficient, it's because it has never understood who you are. The reason I can efficiently create a large number of my intelligent agents is also because I have a complete background prompt about myself. Whenever I need to design and solve a problem, the first step is always to give it my background prompt. This saves me a lot of time. Background prompts help AI transform from a stranger into a partner who deeply understands your business.
Okay, now AI knows you, and you can finally get perfect output answers from your conversations with AI. But have you noticed that many problems in our lives are actually repetitive? If you don't want to keep encountering the same problem and having to converse with AI repeatedly, then we need to talk about the next advanced tool: system prompts. We often dislike AI because when you solve a complex problem with AI, you look back and find that it took you longer than solving the problem yourself. But don't worry at this point. You need to think about what AI's strongest ability is. It's very good at solving repetitive work. So, even if you spend a little more time this time, next time it might take tens of seconds to solve a problem that took hours.
How to understand a system prompt? It's like a recipe for cooking. You see, with the same ingredients, if you put them in the right order when cooking, it might be a delicious fried egg rice. But if the order is wrong, you might end up with undercooked rice. A system prompt is to systematically organize the logic behind the perfect answer we get from conversing with AI, fix it, and turn it into a reusable recipe. This way, next time we make fried egg rice, even if the person hasn't made fried egg rice before, they can make it more or less by following this recipe. Once this step is completed, you will get a prompt with your exclusive imprint, or an intelligent agent that can repeatedly solve the problem.
So, how to do it?
Step 1: Tell AI, "You are my AI engineer. I want to create a system prompt. Its function is to solve what kind of problem." Then, throw all the answers you just got to the problem to AI and ask it to help you organize them into a reusable system prompt. Of course, if you encounter any difficulties during this process, remember to stop and let AI ask you questions. Then you will get a system prompt. This system prompt can be used repeatedly. But wait, test it in a different scenario first. Once you ask it another question and find that it can answer perfectly, then congratulations, your system prompt is complete. From now on, no matter which platform you use, you don't need to repeatedly explain your problem. With this system prompt, you can even have your first intelligent agent. A complex problem-solving process is just limited by the prompt to define its function. Then you can use it to repeatedly and simply get high-quality answers. You are no longer chatting with AI, but building a machine that works for you with text. You can use this system prompt to handle work, conduct research, or even give this system prompt to your team. When they encounter problems, they can directly use this system prompt. At this point, you will find that the quality of the answers your team gets is the same as yours. Complex things become simple, and simple things become automated. This way, you will have time to do what you should truly focus on.
Here's a bonus for you. There have been many system prompts used by AI experts circulating online. Many of these experts are even engineers behind AI software like Perplexity and Notion. Some people have compiled these excellent system prompts and put them on GitHub. If you want to know what these world-class system prompts look like, you just need to Google this phrase: "system prompts AI TOOLS GitHub." Then you will get a large number of free, advanced system prompts. You'll find prompts to help you write content, do finance, and even teach you how to invest. After reading them, you'll have a feeling: "Wow, today's AI prompts are already outrageously powerful."
At this point, your understanding of AI has surpassed most people. But I want to tell you the truth: just knowing how to use AI is far from enough. Because AI is evolving every day. So the real problem we need to face is how to establish an irreplaceable advantage in front of AI. This is a question I often ask myself: what skills can AI never learn, and what is the value of my existence? The answer is three words: values, imagination, and care.
First, let's talk about values. Generating content with AI is not difficult today, but the challenge is how to judge whether the content generated by AI is good or garbage. I've seen many people whose content becomes increasingly mediocre after having AI assistance. The core reason is that they haven't infused their content with values. They use whatever content AI provides, and such content is essentially garbage. In today's world, we can access almost all information. So, if you want your content to influence others, you must have your own strong values, whether you are an individual, a business, or a brand. Look at Li Ziqi. Her values are about returning to nature and supporting traditional crafts. So, white-collar workers in cities who yearn for freedom particularly like her. Let's look at Apple. Why is Apple different? Because Apple's values believe that simplicity is the best, and simplicity becomes sophistication. To truly use AI well, you should be clear about who your content is for and whom you want to influence. Then you will understand how to adjust the content written by AI when it reaches you, making it more vivid and human.
Second, imagination. We've said before that AI predicts the next word based on tokens that make up text. But the way humans think is not formed by text. If I ask you to imagine a black and white panda, can you imagine it immediately? We don't need any text to combine the image. This is actually the biggest difference between us humans and AI. The essence of AI is to learn from the past and predict the present. But our advantage as humans is that we can imagine things that don't exist at all. The problem is that most people are slowly letting this ability atrophy. We scroll through information every day, looking at other people's worlds, and slowly we lose our own imagination. In today's era, the most precious resource is actually the ability to stop yourself, find some time to do nothing, and think. You'll find many strange thoughts rushing into your head. Many great scientific discoveries and great inventions in human history have arisen from this kind of daydreaming, from doing nothing. In the future world, AI will be responsible for execution, and humans should truly be responsible for unleashing their imagination and expanding possibilities.
Third, care. I think this is also the biggest opportunity. AI can help you handle a lot of boring and draining work. This way, you can use the saved time for more important things, such as accompanying your clients, motivating your team, and spending time with your family. Have you noticed a phenomenon? Today, we can listen to almost all music on the apps on our phones. I can contact all the friends I want to talk to around me with a single phone call. But a magical thing is that offline activities are becoming increasingly popular. Look at concerts, business events, and sports competitions. Ticket prices are getting more and more expensive. I heard that the final of this year's FIFA World Cup in North America has been scalped to $2.3 million per ticket. This North American World Cup is really serious about making money! As AI becomes more widespread in the future, our human desire for connection with the real world will only become stronger. AI can simulate all human care, but the person being cared for will eventually find that it's all fake. The essence of care is that it must come from the heart, right? So, do you feel the thoughtfulness of this video? To satisfy this need to go out and see the real world and connect with real people, AI can never satisfy it.
Therefore, values, imagination, and care. Do these three things, and AI will be your strongest weapon, forever irreplaceable. Whether you believe it or not, the development of AI is still in its early stages; it's still a baby. So, none of us are behind in using AI today. You don't need to be anxious or panicked at all. Look at most small and medium-sized enterprises today; they are still trying to resist AI. Even those that use it don't use it well at all. Just watching today's video is enough to put you ahead of others. I highly recommend that everyone use AI to handle a small task around them every day, whether it's letting AI plan a trip for you or write a bedtime story to coax your child to sleep. These are not important. But in the process of these operations, you can utilize the methods I just told you, and you will discover the beauty of AI, which is far superior to any tutorial you can find online.
If you, like me, are starting a business overseas and have felt the value AI can bring to your business, and you want AI to revamp your business, I want to share a tutorial that we have summarized after years of using AI internally. In this tutorial, I will show you how I use AI as our customer acquisition machine internally, help us with sales, and even help us manage our team in the later stages. I have summarized the entire process for you. In the future, in any industry, AI essentially cannot create any demand. However, AI can completely subvert the method of doing business in that industry, meaning to do it all over again. I suggest all entrepreneurs try to look at your business in a new way with AI, acquire customers, do sales, and help you make money. I've put this benefit at the end of my video, hoping to help fellow entrepreneurs overseas achieve rapid progress in their careers with AI. I am Henry from Australia, bringing you a business perspective on our overseas life. If you find this video helpful, remember to click the subscribe button below, and also turn on the little bell so you don't miss any of our future exciting videos. See you in the next video. Goodbye.
Most people, the more precise the prompt, the quality of information you input to AI when assigning it tasks determines the quality of the output. To use an AI large model to its fullest extent, AI's predicted answers will be more accurate.