1 / 1001

No flattery please, Claude: I’m British | Brief letters

TL;DR

Richard Dawkins and chatbots | LLM meaning | Flattery battery | Dancing in PE | Maths breakthrough The otherwise admirable Richard Dawkins should adjust the local settings of the chatbot or tell it to be less obsequious (Richard Dawkins concludes AI is conscious, even if it doesn’t know it, 6 May). Such bots are initially geared to American overenthusiasm and egregiously flattering reinforcement, but just tell them you want British attitude. They’re only simulating you know. Brian Reffin Smith Berlin, Germany. With artificial intelligence bringing “large language models” into everyday use, the LLM after my name has acquired a new meaning. For 70 years I assumed that it referred to my Cambridge master of laws. Trevor Lyttleton London Continue reading...

Nauti's Take

Funny but instructive: persona and tone prompts do work, and users notice when a bot defaults to flattery mode. The risk is that default personas erode trust — flattered users trust outputs more than they should.

For teams putting LLMs in customer contact: calibrate the tone deliberately and test against bullshit-detection, otherwise you walk straight into that trap.

Sources