[Tech Talk] Earn 400 Million Won Annually Just by Using ChatGPT Well? ... The Trending New Job

AI Command Refinement Prompt Engineer
AI as a 'Black Box'... Research on Communication Methods
Generative AI Boom Drives Multi-Million Won Salaries in US and UK

Anyone who has conversed with the artificial intelligence (AI) chatbot 'ChatGPT' realizes the importance of 'questions.' This is because the clarity and detail of the questions posed to AI directly affect the reliability and depth of its answers.

Giving AI instructions to generate specific outputs is called a 'prompt.' Computers understand our intentions and execute commands based on the prompts we provide.

However, the way humans use language and the way computers interpret it are very different, which often leads to communication barriers between us and AI. The profession that resolves these errors and helps AI better grasp our instructions is known as a prompt engineer.

A New IT Industry Job Studying How to Communicate with AI

Image created with the demo version of the generative AI 'DALL·E'. The commands rocket, lift off, and dark sky were used to depict a rocket flying in the night sky.
[Image source=DALL·E homepage]

Prompts for generative AI are very diverse. The chat window we use to talk with the chatbot on the ChatGPT website is itself a prompt. One reason ChatGPT has gained popularity is that GPT-series models excel at natural language processing, allowing AI to understand commands even when we phrase them in everyday conversational language.

In the case of 'image generation AI' that converts text into images, prompts tend to resemble more of a 'computer language.' When using image AIs like Stable Diffusion or DALL·E, users describe the image they want to create by combining words in the prompt box. The AI 'understands' these word combinations as commands and generates images accordingly.

However, this command method makes it difficult for users to produce exactly the desired output. There are limits to specifying detailed instructions. Ultimately, users must repeatedly ask the same questions or manually revise prompts, going through trial and error until they get the desired result.

The people who solve this problem are prompt engineers. Their job is to refine AI commands and precisely identify how neural network AI responds to specific prompts. Simply put, they are researchers of AI communication methods and analysts of AI psychology.

Humanity Still Does Not Understand AI’s Way of Thinking

ChatGPT [Image source=Yonhap News]

The reason the prompt engineer profession is gaining attention is that generative AI is essentially a black box. In other words, humanity still does not fully understand how AI learns from datasets and produces outputs.

Current generative AI models are massive, with vast numbers of parameters, trained hundreds or thousands of times on enormous amounts of data. Generally, the larger the model and the more data it has, the better the AI’s performance.

However, why AI performance improves remains a mystery. Therefore, it is difficult to determine which commands AI recognizes better. Prompt engineers delve into this AI consciousness structure to extract more accurate outputs from AI.

As generative AI companies like DeepMind and OpenAI gain prominence, the value of prompt engineers is soaring. The US big tech company Google invested a whopping $400 million (about 510 billion KRW) in the startup Anthropic, which recently posted a job opening for prompt engineers offering an annual salary of up to $335,000 (about 430 million KRW). In AI hubs like the US and UK, skilled prompt engineers reportedly receive offers of $300,000 to $400,000 per year.

In Korea, prompt engineers are also emerging as promising new jobs in the IT industry. 'Lytton Technologies' (Lytton), a platform that generates content using AI, announced an open recruitment for prompt engineers on the 15th, offering an annual salary of up to 100 million KRW.

Essential Talent for AI or Just a Transitional Job?

It is unclear whether prompt engineering is a sustainable career or merely a transitional job.
[Image source=Yonhap News]

So, how about building a career aiming to become a prompt engineer? Many opinions suggest it is still uncertain whether prompt engineering will become an institutionalized profession like general software engineering.

First, the concept of prompt engineers has only recently emerged. Previously, AI company employees themselves tested commands and refined technology, but as the market expanded and demand surged, companies began hiring specialists for this work.

Moreover, the role of prompt engineers is not just about finding the right commands. Based on a deep understanding of the neural network models underlying generative AI, they investigate why AI responds to certain commands, making them closer to 'computer scientists.' For this reason, prompt engineers at well-known companies are often veterans with extensive experience in the AI industry.

Above all, there is much debate within the industry about whether prompt engineers are essential talents for the AI industry or merely transitional workers. In the early days of the internet, when the concept of search portals was unfamiliar, the profession of 'information search specialists' was popular, and related certifications were created. However, now that the internet is widespread and anyone can perform basic searches, companies rarely hire professional information search specialists.

Similarly, once generative AI becomes established in the market, combining commands is likely to become a basic skill that 'anyone can do,' much like how most young people today are familiar with kiosks and smartphones. At that point, the prompt engineer profession may quickly disappear.