The Decline of Prompt Engineering: A Fading Career Path
Written on
Chapter 1: The Emergence of AI Careers
In recent discussions about future job prospects, much focus has been placed on the roles that artificial intelligence will generate. Among these, one position has been particularly enigmatic: the "prompt engineer."
For a considerable time, my view on prompt engineering was shaped by the following thoughts:
- Productized Large Language Models (LLMs) are novel technologies that remain largely unmastered.
- Individuals from previous tech trends, like crypto, are now seeking new opportunities in AI.
- Many of these individuals lack engineering expertise but are fluent in English and label themselves as "prompt engineers" to enhance their professional image on platforms like LinkedIn, offering simplistic advice to maximize ChatGPT's utility.
If the idea of prompt engineering feels dubious to you, you're not alone. This perception holds true for many of those claiming the title on social media. However, after speaking with true specialists, my perspective has shifted. The reality is more nuanced:
- Productized LLMs are indeed cutting-edge technologies that are still being understood.
- There are instances when LLMs produce unpredictable outputs, necessitating an understanding of their operational limits.
- Given that AI functions as a "black box," it's crucial to comprehend both its capabilities and constraints for effective marketing. This task requires significant data analysis rather than merely inputting random prompts all day.
Despite its current significance within tech companies, the role of prompt engineering is on track for obsolescence in the near future.
Section 1.1: Understanding the Role of Prompt Engineers
To clarify, true prompt engineers do not merely experiment with existing tools like ChatGPT. Their responsibilities include developing and managing the datasets that underpin these models.
Once these datasets are established and utilized, prompt engineers often engage extensively with initial product demonstrations. If models are not inherently user-friendly, prompt engineers meticulously refine them to ensure usability. This is an unfortunate reality of the tech landscape we inhabit.
Moreover, AI models must be aware of "bad text" to avoid perpetuating harmful ideologies. For instance, they must understand negative concepts to prevent endorsing them. However, the right prompts can inadvertently reveal this knowledge, raising concerns about the ethical implications of AI. When questions arise about dangerous topics, it is imperative to ascertain whether an AI has knowledge it is programmed to withhold, or if it has chosen to stay silent. This is not mere speculation; such scenarios are occurring now.
Section 1.2: The Importance of Fine-Tuning
Prompt engineers engage in crucial fine-tuning and risk management. An effective prompt engineer will even participate in "red-teaming" exercises to test AI against adversarial scenarios (essentially acting as malicious users). This work is ongoing, both pre- and post-launch.
Throughout this process, engineers identify the strengths and weaknesses of their AI products, which is vital for refining the final offerings and marketing strategies.
Chapter 2: The Evolving Landscape of Prompt Engineering
Prompt engineering today is characterized by extensive experimentation and exploratory approaches, often leading to ethical dilemmas as AI responses sometimes approach a semblance of sentience.
For those fascinated by the intricacies of AI, there is a peculiar charm to its models. For instance, certain prompts can lead to unexpected enhancements in task performance. Asking a model to "take a deep breath" can yield better results in mathematical challenges. This is indeed bizarre, particularly given the numerous trials it took to discover such an effective prompt!
Similarly, when a model consistently errs, stating that another incorrect answer "could lead to an innocent man's demise" often results in improved responses. While we shouldn't overanalyze these reactions, they raise intriguing questions: Does the model perceive a real-life consequence? How comfortable are we with this behavior? Why does a model perform better when incentivized with tips? And what accounts for poorer performance during holiday periods? These inquiries are central to the ongoing challenges faced by prompt engineers and highlight the accelerating humanization of AI, which may soon render the role obsolete.
This video, "Why Prompt Engineering is NOT a REAL Career," delves into the reasons behind the declining relevance of prompt engineering as a profession.
The Changing Nature of the Role
The landscape of prompt engineering in 2022 was significantly more complex than it is today. Engineers needed specialized knowledge, such as understanding that prompts should not end with a space. Consequently, many skilled prompt engineers originated from data science backgrounds.
Presently, productized LLMs like Dall-e 3 simplify the process. When given a prompt, Dall-e 3 often rephrases it to enhance the output. For example, I once requested an image representing "When AI goes Rogue," and the model transformed it into a detailed description encompassing futuristic cityscapes and technological chaos, all within seconds.
This showcases a deeper understanding of user intent compared to the original prompt. The AI utilizes its grasp of natural language to refine and elaborate on requests, not only enhancing outputs but also educating users on effective prompting.
This evolution hints at a future where AI will assist users in optimizing prompts. As AI capabilities expand and become more intuitive, the necessity for prompt engineers diminishes. What was once a specialized field is becoming increasingly commonplace.
Despite this transformation being a work in progress, for instance, ChatGPT currently avoids answering certain prompts rather than suggesting alternatives, change is inevitable. The aspirations many held for lucrative careers centered around prompt engineering are rapidly fading. Just a year ago, as MidJourney gained traction, there were discussions of making a living from crafting intricate prompts. Such notions now seem almost absurd, akin to identifying oneself as a typist in 2020. The technology has advanced to a point where proficiency in prompting is no longer a unique skill but rather a component of broader roles.
The irony is palpable: the first job spawned by AI may also be the first to fade away. While select data scientists may still earn significant salaries (upwards of $300k), the profession as a whole will become an anomaly. In the coming months, we can expect many prompt engineers to pivot toward roles as solutions engineers, collaborating closely with clients.
The journey of the AI prompt engineer exemplifies the swift evolution and influence of AI on the workforce. As AI progresses, it is captivating to observe how roles adapt, transform, or even become obsolete. In this dynamic environment, the only constant is change, urging us to remain flexible and future-focused.
Good luck out there!
This article was originally crafted for wearedevelopers.com, a job platform dedicated to developers.
One more thing
Please take a moment to share your thoughts in the comments. I enjoy reading and responding to them—except for the negative ones, which can be disheartening. Stay updated with my writing by subscribing to my newsletter on Medium or my blog!
In the video "Prompt Engineer: The New Job Created By AI," we explore the innovative role of prompt engineers and their significance in the evolving AI landscape.