We’re still pretty early days in language models, particularly Large Language Models (LLMs). The last year in particular has been pretty crazy. Just under a year ago, on 30 Nov 2022, OpenAI launched ChatGPT – a chatbot wrapped around a OpenAI’s GPT 3.5 model. Just 2 months later they grown to 100 million users and a $29 Billion valuation.
Pretty quickly those 100 million users found that it wasn’t like on Star Trek. You had to type but, more importantly, there are some quirks which needed to be addressed. Two months later, the term prompt engineering caught on. I suspect the term will be with us for a few years but will eventually fade out of existence. For LLMs, I think the change will be due to pre-supplied context and perhaps part of that will be a sort of development of best practices. Small language models will rely on their focus to circumvent a lot of the issues. But for now we’re still stuck with the term of Prompt Engineering along with it all of the connotations of engineering.
One side believes this is brand new world and Prompt Engineers are akin to the Wright Brothers. That is, formal training isn’t really a requirement when you are inventing the field. If the Wright brothers designed an engine, collaborated with their shop mechanic who they tasked to build it, calculated the thrust of their propellors, built a wind tunnel for testing models, then – schooling or not – they were engineers.
The other side recognizes the Wright Bros. as point in time but since then engineering has formalized greatly. And some of the hallmarks of engineering now include things like being able to model a system formally, come up with a set of predicted results. Additionally, many fields also have the requirements of being able to do this with a system in place and also model the costs involved. And, of course, prompt engineering has none of this. It’s more like tinkering. The enthusiasm is great and they may get good results but they shouldn’t think of themselves as real engineers.
From my perspective, I’m not really sure this is a fight worth having either way. Linguistically (and historically) the term has changed over the years, and rather greatly. At one time anyone who operated a siege engine was an engineer. It later primarily referred to someone who applied practical knowledge and ingenuity to solve problems related to construction, design, and invention. It still has branch that essentially means people who operate heavy equipment (like a locomotive engineer). Military engineers can in theory build things like a set of trenches but more often blow things up which is pretty opposite to the building we thing engineers do today. And even amongst today’s engineers, often when a person say they are software engineer they aren’t viewed a ‘real engineer’.
We’ll see. I don’t think the field will last and I think the skill set will be somewhat similar to putting on your resume that you could do email or word processing. That wouldn’t be any sort of differentiator today and you’d look like a fossil were it your resume. I suspect that will be the same in the not too different future.