My Demons VS Predictive Engines
When Rimbaud became a slave trader, he stopped writing poetry. Chinua Achebe
I find myself wondering what kind of skills I value in my own thinking. It would be nice to develop what's lacking, with new AI tech I start to play defense and maintain. Meditating about my token usage I seriously wonder, can I augment myself with new tools in some type of positive direction? What do I need to make educated decisions with how I adopt it? Draft thoughts ahead.
Human abstraction VS model prediction
Fundamentally models don't possess abstraction the way we do, such as a digital twin or imaginative sculpture. I want to start ignoring some specific patterns from LLMs, especially when they come in first person:
JUDGEMENT (don't trust LLM to introduce absolutes) Language: "... best", "always", "never", "must", "should", "impossible"
FUTURE LOL "will", "going to", "inevitable"
LOG SUMMARY VS DETERMINISTIC OVERREACH In general, LLMs are not deterministic. “root cause is…” “logs prove…” “definitely the bug is…”
I am going to try to move this week ignoring all these patterns and focusing on what value is left. I know my language can be funky, I am definitely benefiting from it correcting my grammar and being my assisting companion. Even so with my grammar mistakes I don't like its stylistic judgement :)
What I found with writing is I need to focus on editing one sentence at a time because when it edits the paragraph it changes the voice to morally painted average. Humans are fundamentally flawed beings, and AI consumes training data with human psychosis of trying to appear fundamentally good.
But that's a whole other post...
Today’s Fuel
Regular: $5.95 Diesel: $90 a tank