.jpg)
It just crossed my mind that people may not be able to distinguish my writing from an AI acting like me. I mean, it is pretty easy to make an LLM write some article as if it was "me".
I can just feed all my existing writings and the bot can write, on any topic. I can even train it, now. I know how to fine-tune and create specialized LLM models.
Therefore, it can be done.
and... kaboom, we have a Robo-Yunus!
Which seems to me as a terrible idea with zero sense of creativity.
I can make a bot mimic me. Will I do it? No. That would be stupid to be honest. Seems like a cringe idea to begin with.
I don't need my AI to be replicating my chain-of-thought, I need it to bring newer combinations to the table so I can benefit from them. A digital friend that is not me is better than a digital me engaging with me, like me.
I need a robot to bring a new perspective combined from billions of data snippets and not just replicating my way of expressing myself.
Trying to write a poem in the style of Einstein being angry at a donut is fun. Making a fake physics paper as if Einstein, is NOT fun.
Why not use LLM tools to better understand a topic with the perfect tone and choice of words for you?
I LOVE when I can tell an LLM to rewrite it's response so I can better understand a topic.
My number #1 personal rule is this: Never ever use an AI to write something you are going to publish on your personal blog. That is YOUR blog, not a dumpster of information.
I would suggest the same to everyone really.
If it is your blog, it is meant to be about you and your work.
I want to read YOUR perspective of things. That's why I visited your website or twitter etc. Otherwise I would visit ChatGPT, NOT you.
And I really hate to see the "Rewrite with AI" button right here on LinkedIn as I am typing this. To give them credit though, it is probably there to rewrite marketing or very formal content mostly but still it looks annoying.
For example, I always use LLMs to rewrite not-so-unique texts such as fixing my email structure, fixing my spelling, fixing a list I made but didn't put proper spacing or symbols. I also use LLMs to generate marketing content from my little paragraphs of examples. They write pretty good informative texts.
As long as a computer is acknowledged for what it is, there isn't a problem.
If a computer is programmed to NOT acknowledge what it is, that IS a big problem.