GenAI: Don’t Silence Your Voice, Amplify It!

I have been writing extensively about collaborative intelligence (CI) and generative AI. You might think that I would be using generative AI to do the writing since it would demonstrate its use as a means of improving productivity (which is one of CI’s goals).

I have not been doing this. Why? Because I have my own voice. I see the world in ways that Chat-GPT and other Large Language Models (LLMs) cannot.

LLM-generated outputs are surprisingly good in most cases. They are often good enough for the job at hand or a great first draft that can be revised and finished much more quickly than starting from scratch. But not in cases when I am writing something unique to me, my knowledge and experience, and my worldview.

I have more than 35 years of work experience, a nuanced understanding of complex topics, and expertise in several domains. If I ever want to use an LLM to help me write something that sounds like me, it's going to have to learn my writing style. If I want it to reflect my expertise and perspectives in its writing, it must learn what I know and learn to see things as I do.

Over the years, I've created many proprietary frameworks, processes, and tools that are unique to me. Someone somewhere may have come up with something similar and written about it publicly so that an LLM might know about these things, but it is not evident from interacting with them in my domains of expertise.

It's cool—and something of a parlor trick—to ask an LLM to write a short story in the style of Ernest Hemingway or a haiku in which every word starts with the letter B. But what if you want it to generate something in the style of YOU?

This is a dilemma we all face. We want to use the new AI tools to help us generate content and be more productive, but these tools don't know enough about our individual styles and perspectives. If we take the time to write in our own voice and then use these outputs to train an LLM, people fear that it is, at best, a waste of time, as much of what we write doesn’t need our “voice”, and, at worst, that we will become redundant once the AI can produce content just as we would.

These fears are unfounded. Yes, AI will change how our jobs are defined and performed today, but it will not take all our jobs or make us all obsolete. There will always be new work to do and new human endeavors that can be supported—but not supplanted—by AI or other machine intelligence.

So, I’ve decided to write down what I know so that I can begin to train LLMs to help me in the future. If we are going to use these tools to produce the things we might produce ourselves, we have to write without the help of AI, learning or honing our ability to express our proprietary knowledge, unique voice, and point of view.

I think this work is valuable for several reasons, including:

• Writing helps me think more clearly, more deeply, and in a more structured way about the topic of interest, something that cannot be “outsourced” to machine intelligence.

• Writing broadens my knowledge as I search for examples or references to augment and support my positions.

• Writing helps me improve my critical thinking skills as I anticipate and respond to readers' questions in advance.

• Writing improves my communication skills by enabling me to learn what resonated with readers and what did not, and then identify ways to improve in the future.

Another valuable result, and my motivation for taking this approach, is that I will be able to use an LLM to help me produce writing that reflects my voice and my unique expertise more quickly, enabling me to amplify my voice and respond more quickly and effectively to future demands.

I encourage you to generate your own content, both as a means of self-expression and to develop a “training set” for your LLM of choice. You may have read that it takes millions of examples to train an AI, so writing even 100 pieces would be a waste of time. However, LLMs have shown a remarkable ability (which will improve over time) to learn quickly from a small set of examples.

If you don’t share my motivation, I’d still encourage you to write without the help of generative AI. I think “Human-Made” will be highly valued in future markets. In this new age of rapidly produced statistical sameness, our individuality and uniqueness will be valued more than ever.

Don’t be influenced by the many “AI bros” on social media trying to convince you that you’d be a sucker to ever write anything on your own again.

If you do not take steps to defend and preserve your self-expression, most of the content you will produce in the coming years will be whatever the latest LLMs spit out because they can certainly write faster than you, but not necessarily better (especially in matters requiring your perspective and proprietary knowledge).

Don't allow your unique and powerful voice to be lost forever in the coming torrent of LLM-generated words.