No one needs engineers giving orders to AI anymore.

Prompt engineering for AI was one of the most sought-after jobs in 2023. However, the rapid pace of AI development has rendered this profession obsolete.

In 2023, while millions were worried about the possibility of artificial intelligence models like ChatGPT taking their jobs, some companies were willing to pay hundreds of thousands of dollars to hire people capable of leveraging these next-generation AI chatbots.

According to Bloomberg , the emergence of ChatGPT at that time created a new profession called Prompt Engineer, with salaries reaching up to $335,000 per year.

"Speaking to AI"

Unlike traditional programmers, the engineer suggests programming in the form of prose, then sends the plain text commands to the AI ​​system. The system then transforms the descriptive text into actual tasks.

These individuals often understand the flaws in AI, which allows them to enhance its capabilities and develop sophisticated strategies to transform simple input into truly unique results.

Lance Junck once earned nearly $35,000 in revenue from an online course teaching people how to use ChatGPT. Photo: Gearrice.

"To use AI effectively, you must master the skill of designing commands. Without this skill, your career will, sooner or later, be ruined," said Lydia Logan, Vice President of Global Education and Human Resource Development at IBM.

However, with rapid development, AI models now understand user intent much better and can even ask follow-up questions if the intent is unclear.

Furthermore, according to the WSJ , companies are training a range of employees across various departments on how to best use commands and AI models, so there is less need for a single person to hold this expertise.

Specifically, in a recent survey commissioned by Microsoft, 31,000 employees in 31 countries were asked about new roles their companies are considering adding in the next 12-18 months. According to Jared Spataro, Microsoft's AI at Work marketing director, command engineer ranked second from the bottom of the list.

Meanwhile, roles such as trainer, data specialist, and AI security expert top the list.

Spataro argues that major language models have now developed sufficiently to enable better interaction, dialogue, and contextual awareness.

For example, Microsoft's AI-based research tool will ask follow-up questions, let the user know when it doesn't understand something, and request feedback on the information provided. In other words, Spataro argues that "you don't need to have perfect statements."

Prompt being "blind" isn't wrong.

According to Hannah Calhoon, vice president of AI at the job search platform Indeed, the number of job postings for command-line engineers is currently very low.

In January 2023, just a few months after ChatGPT launched, user searches on Indeed for this role skyrocketed to 144 per million searches. However, since then, that number has leveled off at around 20-30 per million searches.

Prompt Engineers are engineers whose job is to create questions or commands for AI tools like ChatGPT. Photo: Riku AI.

Besides declining demand, constrained by tight budgets and increasing economic uncertainty, companies have also become much more cautious about hiring in general in recent years.

Accordingly, companies such as Nationwide Insurance, the workwear brand Carhartt, and New York Life Insurance all stated that they have never hired order engineers. Instead, they found that honing order-setting skills is a better specialization that all current employees can be trained in.

"Whether you work in finance, human resources, or legal, we view this as a capability within a job title, not a separate job title," said Jim Fowler, Chief Technology Officer of Nationwide.

Professor Andrew Ng, founder of Google Brain and lecturer at Stanford University, suggests that users sometimes don't need to be overly detailed when entering prompts for AI.

In a post on X, Ng called this method “ lazy prompting ” —that is, feeding information into AI with very little context or no specific instructions. “We should only add details to prompts when absolutely necessary,” said the Coursera and DeepLearning co-founder.

A prime example given by Ng is programmers when debugging. They often copy and paste entire error messages – sometimes several pages long – into the AI ​​model without specifying the requirements.

"Most large language models (LLMs) are smart enough to understand that you need them to analyze and suggest ways to fix errors, even if you don't explicitly say so," he wrote.

LLMs are gradually moving beyond simply responding to commands, beginning to understand user intent and reason to provide appropriate solutions. Photo: Bloomberg.

According to Ng, this is a step forward showing that LLMs are gradually moving beyond simply responding to commands, beginning to understand user intent and reason to provide appropriate solutions – a trend that AI model development companies are pursuing.

However, "lazy prompting" isn't always effective. It's worth noting that this technique should only be applied when users can quickly test the model, such as through a web interface or AI application, and the model is capable enough to infer intent from limited information.

Mr. Ng emphasized: "If AI needs a lot of context to respond in detail, or cannot recognize potential errors, then a simple prompt will not help."

Source: https://znews.vn/khong-con-ai-can-ky-su-ra-lenh-cho-ai-nua-post1549306.html