This is a common topic. Should I learn "prompt engineering"? Should I take a class? It's an essential and important job in the era of artificial intelligence. I often, no, often hear this kind of talk. I always say.
If you think prompt engineering is simply about asking good questions or getting results in a style or format, then you don't need to listen to it at all. It's just seen as a class on how to speak nicely or document well.
In previous posts and the good educational materials I always talk about, the definition of prompt engineering has been given countless times and the educational method has also been established to some extent. Prompt engineering is ultimately a technology that elicits the desired answer from a machine according to human intention. The key here is the premise that "machine", that is, a principle exists and can be predicted to some extent.
Anthropic, the developers of Claude, say that if you are clear on these three things, you are already doing something called prompt engineering.
Clear definition of success criteria for use cases
Here are some ways to test against these criteria that we have already experienced:
The first prompt to improve ← The first prompt can't be perfect!
In fact, in addition to this, we have also heard research results through news and papers that prompt engineering can produce better results faster and more effectively than fine-tuning. In addition, OpenAI and Anthropic, frustrated with this, started to distribute their own prompt writing tips, examples, and training.
OpenAI also releases System Prompt, and Google and Anthropic also release System Prompt (LLM's Instruction), continuously showing why they set this as the default setting and how they use it. In fact, if I was going to talk about this, I wouldn't have written this post. As I said before, I'm tired of saying what I've already said in previous posts, and people who are looking for it have probably already looked it up... I thought I should write a blog post after watching the video below. A podcast run by the actual developers of Anthropic!
Explore Prompt Engineering: Expert Perspectives and Insights
Prompt engineering is a discipline that emphasizes clear communication, iterative improvement, and contextual understanding to optimize interactions with models, thereby maximizing the performance of AI models.
🔍 Prompt engineer Zack Witten defines prompt engineering as the art of efficiently communicating with a model to achieve optimal performance , likening it to a process of trial and error .
⚙ Unlike traditional programming, prompt engineering has the characteristic of learning through repeated experiments and feedback , which can improve the model's response.
🗣Prompt Engineer Jack emphasized the importance of clear communication, the ability to iterate, and anticipating edge cases as the conditions for a great prompt engineer.
💡 Amanda, who is in charge of fine-tuning the language model, emphasized the importance of clear communication between humans and models, pointing out that effective prompts have a big impact on model performance .
The Evolution and Future of Prompt Engineering
🤔 This podcast covers the complexities of AI inference, suggesting that anthropomorphic model interactions can lead to misunderstandings, but structured inference can improve model performance.
💡 It turns out that presenting clear examples and repeatedly inferring improves the performance of the model .
🔍 While there were opinions that good grammar and punctuation improve clarity , there was agreement that the model's ability to understand was more important .
🔑 The future of prompts is already expected to evolve into an interactive relationship where models elicit information from users and improve the prompts .
🤖 As AI technology advances, the role of prompt engineers will gradually change to facilitating conversations with AI .
Among them, Zack Witten, who is currently working as a prompt engineer at Anthropic, explains a bit more clearly what he does. Rather than just giving vague orders and keeping ethics, he talks about how to approach it based on his previous experience as a Machine Learning Engineer and how it is meaningful. You can see the summary of the video by expanding the toggle above.
Personally, I was proud of myself because it contained exactly what I had mentioned in the guide for humans using artificial intelligence . This article was written in the summer of 2023, but many people are still looking for it, and what people who actually work as prompt engineers say seems to be that the essence has not changed.
As I mentioned in my previous study sharing article, users end up spending the most time hesitating about what to ask AI and how to use it, and once the “purpose” becomes clear, the methods and tools for using it will rapidly develop, disappear, and establish themselves.
"If your goal is not clear, it is difficult to achieve results no matter how much you do." This is a saying that we often hear, but we tend to forget as we live. As I felt while talking to Hyunseon today , once you decide what to do, you can change and use tools and directions as much as you want. I hope that those who read this article will take the time to reflect on what they wanted to do and what they will do, rather than being dazzled by the word "prompt engineering" and the theme of artificial intelligence. I also forget every time. If you don't make a promise to yourself.
Subscribe to 'haebom'
📚 Welcome to Haebom's archives. --- I post articles related to IT 💻, economy 💰, and humanities 🎭. If you are curious about my thoughts, perspectives or interests, please subscribe. Would you like to be notified when new articles are posted? 🔔 Yes, that means subscribe. haebom@kakao.com
