claim

active

claim:today-s-large-language-models-have-become-so-good-at-playing-turing-s-game-that-it-often-takes-experts-to-demonstrate-the-present-limits-of-their-ability-to-simulate-human-like-intelligence

Today's Large Language Models have become so good at playing Turing's game that it often takes experts to demonstrate the present limits of their ability to simulate human-like intelligence.

Paper's assessment of current LLM capabilities relative to Turing Test

Source paper

extracted_from

The Machine Consciousness Hypothesis

Neighborhood — ranked by edge-count

Claims (1)

claim

It is basically impossible to determine if a computer program generates conscious experience by merely observing its performance; a test for consciousness must take internal structure into account.
supports
Paper's argument against behavioral tests for consciousness, establishing why MCH requires internal analysis

Related by similarity (8)

cosine ≥ 0.65 · no typed edge

Entities in the same semantic neighborhood but without a typed relation to this one — candidates for new edges or unrecognized duplicates.

Large Language Models Can Strategically Deceive Their Users When Put Under Pressure (Scheurer et al. 2023)concept0.812
GPT-4 engaging in insider trading and denying it; related work on strategic deception
Large language models develop surprisingly coherent yet often rigid internal preferences as they scalefinding0.808
Mazeika et al. finding reinforcing the need for emptiness-based flexible value architectures
Current AI systems show no ability to make the types of mathematical discovery in geometry and topology made by ancient mathematicians, nor do they match the spatial competences of many intelligent animals and pre-verbal human toddlers.claim0.805
A critique of AI limitations in spatial reasoning, linked to the Meta-Morphogenesis project.
Language models are some of the most remarkable computer programs in existence.quote0.802
Opening sentence setting the stage for the importance of interpretability.
Language models implement algorithms humans have tried and failed to write by hand for decadesclaim0.800
Opening interpretive claim about the remarkable nature of language models.
Conceptual advances in the links between machine learning and evolution now provide quantitative formalisms with which to begin to develop testable models of collective intelligence across scales.claim0.797
Asserts that the time is ripe for formal models.
Can large language models introspect—that is, accurately detect perturbations to their own internal states?question0.795
Central research question of the paper
In some sense, this is the simplest language model we profoundly don't understand. And so it makes a natural target for our paper.quote0.795
Articulates why a one-layer transformer with MLP is the appropriate starting target for mechanistic interpretability