On Trust, Context, and Language

Trust

• LLMs are non-deterministic, but human actions are non-deterministic too. But trust factors for humans go beyond logic and cognitive intelligence. (That trust is evolutionary)

Context

• An individual word has no meaning.
• Words gain meaning only in context.
• LLMs output comes from prior of context. The quality of output depends on how well context can be orchestrated.
• Same applies to the Human Language landscape, better priors lead to highly certain action predictions.

Language

• Language is a facilitator of senses.
• Language unites input and output.
• Is there anything that cannot be explained by Language?
• One may say "Can you explain the concept of color to a blind person?"
  • But the concept of color itself is a language.
  • As language facilitates senses, absence of sense does not mean language failed to explain it.