Make AI say everything about Golden Bridge without prompt engineering

LLMs are constructed with a number of parameters. And each of them has certain role for each part of output.

Anthropic, who made Claude, published a research about control LLMs by manipulating its paramrters

1. How LLMs learn?

2. The link between parameters and the output.

3. What they developed, the Golden Bridge bot.

4. Applying for AI safety

5. Further research suggestion, similarity with human brain