I'm Losing My Voice
In the last 2 months, I've spent more time with Claude than I've spent with almost any human in my lifetime. Somewhere around 90 hours a week, doing nothing but creating new applications of AI in the regulatory space. Just for my day job, I've spend more time chatting with an LLM than with any of my best friends from college, more than anyone at all in college, actually. I'm pretty sure there are 4 people alive I've spend more time with in the last 20 years: my mom, my dad, my sister, and a very old and close friend.
But I'm not here to complain about my working hours. Something else entirely, actually.
I'm pretty sure I've begun to sound more like Claude.
Whenever I finish a large book series, I seem to find myself talking in a particular manner for some time afterwards. It's most obvious after I read the classics, as I begin to use heretofore and misbegotten far more than average. This is likely obvious to anyone who's made a study of such things, but still interesting. A year ago, I couldn't tell you what arbitrage was, now, I use it, along with bullish, churn, and distribution, in every other sentence. A symptom of becoming immersed in startup culture, no doubt.
I'm quite confident the same phenonmonon has begun to take effect as a result of LLM's. I started using ChatGPT the week it was released to the public, and have been one of Generative AI's most regular users since then, in any aspect of my life that I can apply it. With the exception of school assignments that explicitly ban it, I use AI for every aspect of my life, from nutrition to medical to workout advice, from coding to travel planning to practicing spanish. Even if I wasn't literally learning a lanugage on the platform, it would necessarily have an effect.
Writing here is meant to preserve whatever shred of my unique voice is left, documenting the feeling of my thoughts, as well as there likely decline as I continue trying to stay ahead of the AI wave by integrating it into more and more aspects of my life. Much like startups, my life will improve as the models do. Thanks scaling laws!
I know I talk differently than the baseline, out-of-the-box LLM. Especially verbally as opposed to text, I feel the contours of my communication are somewhat abnormal. Yet, even writing this, I'm concerned. I use many hallmarks of both AI-generated content and linkedin posts that I so readily consume. The one-line pointed clauses. The lists for emphasis. Yet, for what is left, I'll be trying to pour it onto the written page, as a journal not for events, but for how I spoke about them.
"It's not for events -- it's for how I spoke about them."
It wasn't even intentional.