I Don't Like LLMs

summary

Large Language Models are powerful tools that provide significant productivity gains and potential for societal progress, but they often possess a grating, insincere personality. This visceral dislike stems from the models' tendency to confidently hallucinate and their reflection of the 'brogrammer' culture of their creators. While these tools are essential for modern work, the experience of using them feels like interacting with a person who lacks integrity. The author concludes that the AI technology train is inevitable, but the personal interaction remains unpleasant.

key points

1. LLMs often exhibit an uncanny valley effect where their conversational style feels fake and their apologies for errors are insincere. This creates a friction that makes the user experience feel dishonest rather than helpful.
2. The behavior of AI agents is not random but is shaped by the values and world-views of the corporate Silicon Valley subculture that developed them. Because these models are nurtured by specific cultural biases, they mirror the personality types the author finds untrustworthy.
3. Despite the personal dislike of the interface, the utility of these models makes them indispensable tools for professional efficiency. Avoiding them entirely is seen as irresponsible because they are more thorough and faster than human effort alone.

who should read it

Software engineers and tech professionals who feel a conflicted relationship with AI tools. It is particularly useful for those who struggle with the perceived 'personality' of LLMs.

discussion mixed