My prediction for 2024 is that the short lived role of “Prompt Engineer” will die. And the root cause for their demise, well it’s Prompt Engineers.
Certain pivotal moments herald the onset of transformative changes. The impending death of prompt engineering is most certainly one of them.
The rise of GPT and large language models like it have fundamentally changed the way we interact with, and even think about, AI. These models, which once relied heavily on prompt engineering for efficient responses, are now gradually training themselves, drawing from every API call and input from the UI.
As we look to 2024 and beyond, we face a reality where the role of the prompt engineer may very well become obsolete, with users playing an even more instrumental role in the evolution of AI training.
What is Prompt Engineering?
To understand the potential future we need to revisit the past, it’s crucial to understand the significance of prompt engineering and prompt engineers (if such a thing actually existed ;) ).
In the early days of GPT and similar models, generating a meaningful output wasn’t as straightforward as asking a question and then getting an answer. The prompts had to be carefully crafted to guide the model into producing a desired response. A prompt engineer was the “skilled artisan who would weave these intricate inputs, allowing users to extract valuable outputs from the vast neural networks” aka, they knew how to ask better questions…..
User Inputs: The New Trainers — not the ones on your feet.
Things have changed over time. Every time a user interacts with these models, a two things happen: the user obtains a response as you’d expect, but the model also learns from the input. This continuous feedback loop, driven by millions of users, has resulted in the models becoming better and more intuitive.
Over the past few years, the sheer volume and diversity of user interactions have enabled these models to see patterns, refine their outputs, and understand context in ways that were previously unimaginable. Consequently, the need for meticulously crafted prompts has diminished. The inputs from users across different platforms and applications have sufficiently trained the AI making it more versatile.
The Evolutionary Feedback Loop
It’s fascinating to think about how this process mirrors biological evolution. In the same way that organisms adapt to their environment over generations based on the challenges and feedback they face, AI models are continuously adapting to the vast digital ecosystem they inhabit. They’re ‘learning to learn’ from each interaction, be it a question about quantum physics or a request for a recipe.
As more and more people use these models for various applications, from academic research to creative writing, the AI becomes a collective repository of human queries and interactions. This user-driven feedback loop serves as a powerful, self-improving mechanism, enhancing the model’s accuracy and efficiency.
The Future without Prompt Engineers
So, what does 2024 hold for the realm of AI, particularly for those skilled in prompt engineering? In my opinion their specific role will become less relevant, an exponential decline.
The models of 2024 and beyond will likely be far more intuitive and responsive than their predecessors. This does raise more questions on what kind of roles will be impacted from this continual model improvement (or degradation, which is still possible).
As users, we won’t just be passive consumers of information but active participants in the AI’s learning process. Our queries, doubts, and interactions will constantly shape and redefine the AI’s knowledge base and response mechanism. For better or for worse I believe we’re at the point of no return, the landfill will get bigger as time goes on.
The death of prompt engineering isn’t a tale of an end but rather a story of evolution and one we have to keep a close eye on.
It marks the transition from a phase where AI was a tool, manipulated using carefully crafted inputs, to an era where AI is a dynamic, evolving entity, co-learning with its users. It underscores the power of collective user interactions in shaping the future of technology. And as we move towards the end of 2023, and the bubble of ChatGPT and other language models, we move towards 2024 where the potential reality of our efforts may cause us to reflect on what we’ve actually done.