Psychology
fromLesswrong
5 days agoA Mirror Test For LLMs - LessWrong
A new measure of LLM self-awareness is proposed, but current models ultimately fall short in demonstrating true self-awareness.
Walking through a field one day, a 17-year-old schoolteacher named George Boole had a vision. His head was full of abstract mathematics - ideas about how to use algebra to solve complex calculus problems. Suddenly, he was struck with a flash of insight: that thought itself might be expressed in algebraic form. Boole was born on November 2, 1815, at four o'clock in the afternoon, in Lincoln, England.
Structural pattern matching excels at... matching the structure of your objects! For the two examples in this article, we'll be using a number of dataclasses that you can use to build abstract Boolean expressions: from dataclasses import dataclass class Expr: pass @dataclass class And(Expr): exprs: list[Expr] @dataclass class Or(Expr): exprs: list[Expr] @dataclass class Not(Expr): expr: Expr @dataclass class Var(Expr): name: str
Which Algorithm Is This? If you step back, this maps almost perfectly to the Top K Frequent Elements problem.We usually solve it for integers in a list. Here, the "elements" are audience profiles age and body-type combinations. First, define what an audience profile looks like: case class Profile(age: Int, height: Int, weight: Int) What we want is a function like this:
In 1983 I asked my parents for an Atari for Christmas, instead I got a Commodore 64... Needless to say, I was very disappointed until I discovered how much cooler Wizard of Wor was than Combat. To their credit, my parents thought a computer was a better investment than a video game. I used that C64 through my sophomore year of college until I replaced it with a 486; my first real investment.
For the first time, speech has been decoupled from consequence. We now live alongside AI systems that converse knowledgeably and persuasively-deploying claims about the world, explanations, advice, encouragement, apologies, and promises-while bearing no vulnerability for what they say. Millions of people already rely on chatbots powered by large language models, and have integrated these synthetic interlocutors into their personal and professional lives. An LLM's words shape our beliefs, decisions, and actions, yet no speaker stands behind them.