Talking About Large Language Models

2022/12/07 by Murray Shanahan, Shanahan, Murray · 11 voices · 38 citations
Computer Science · #Topic Modeling #Explainable Artificial Intelligence (XAI)

paper · pdf · doi:10.48550/arxiv.2212.03551

Abstract

Thanks to rapid progress in artificial intelligence, we have entered an era when technology and philosophy intersect in interesting ways. Sitting squarely at the centre of this intersection are large language models (LLMs). The more adept LLMs become at mimicking human language, the more vulnerable we become to anthropomorphism, to seeing the systems in which they are embedded as more human-like than they really are. This trend is amplified by the natural tendency to use philosophically loaded terms, such as "knows", "believes", and "thinks", when describing these systems. To mitigate this trend, this paper advocates the practice of repeatedly stepping back to remind ourselves of how LLMs, and the systems of which they form a part, actually work. The hope is that increased scientific precision will encourage more philosophical nuance in the discourse around artificial intelligence, both within the field and in the public sphere.

Cited by

• Deterministic Replay for AI Agent Systems
• Does neural computation feel like something?
• Not Minds, but Signs: Reframing LLMs through Semiotics
• Exploring the integration of large language models in industrial test maintenance processes
• Alignment Is Not Enough: A Relational Framework for Moral Standing in Human-AI Interaction
• Reading Without a Reader: Large Language Models Collapse Reading and Writing into a Single Entangled Code
• Coherent without Grounding, Grounded without Success: The Bidirectional Coherence Paradox in Artificial Epistemic Agents
• How Tech Workers Contend with Hazards of Humanlikeness in Generative AI
• Humanlike AI Design Increases Anthropomorphism but Yields Divergent Outcomes on Engagement and Trust Globally
• Large Language Models have Chain-of-Affect
• Tracing the ongoing emergence of human-like reasoning in Large Language Models
• Invasive Context Engineering to Control Large Language Models
• Difficulties with Evaluating a Deception Detector for AIs
• LLM Reinforcement in Context
• On the Creativity of AI Agents
• A Standardized Benchmark for Multilabel Antimicrobial Peptide Classification
• Cognitive Convergence: Deep Similarities Between Large Language Models and Human Cognition
• Large models of what? Mistaking engineering achievements for human linguistic agency
• Exploring students’ (mis)conceptions about ChatGPT-generated text: A qualitative study
• Discovering Heuristics with Large Language Models (LLMs) for Mixed-Integer Programs: Single-Machine Scheduling
• Network Intrusion Detection: Evolution from Conventional Approaches to LLM Collaboration and Emerging Risks
• Resilient Radio Access Networks: AI and the Unknown Unknowns
• Creating a large language model of a philosopher
• LLaVAShield: Safeguarding Multimodal Multi-Turn Dialogues in Vision-Language Models
• Tracing Uncertainty in Language Model "Reasoning"
• Asymmetric Communication: Large Language Models and Language Games
• Mind Over Model? Students' Evaluation and Use of <scp>ChatGPT</scp> ‐Generated Versus Human‐Generated Texts
• Symbol ungrounding: what the successes (and failures) of large language models reveal about human cognition
• AI for social science and social science of AI: A survey
• Generative AI, large language models, and their agentic framing in news media
• Intermediate Languages Matter: Formal Languages and LLMs affect Neurosymbolic Reasoning
• What Would an LLM Do? Evaluating Policymaking Capabilities of Large Language Models
• The Personality Illusion: Revealing Dissociation Between Self-Reports &amp; Behavior in LLMs
• Data Augmentation Improves Machine Unlearning
• Why Cannot Large Language Models Ever Make True Correct Reasoning?
• <scp>AI</scp> Mimicry and Human Dignity: Chatbot Use as a Violation of Self‐Respect
• Explainability Through Systematicity: The Hard Systematicity Challenge for Artificial Intelligence
• AI anthropomorphism [wikipedia]

Discussions

• Talking About Large Language Models [hn, 146 points, 149 comments]
• Here's a great paper that sorta gets at it from the opposite direction (I'd talk about the results, but this paper will help you understand the *processes*). It was written in the Bronze Age of LLMs [bsky, 15 points, 1 comments]
• This draws upon @mpshanahan.bsky.social 's work regarding how we should talk about and understand large-language models. The highly readable article I've linked to here is one of the best guides to co [bsky, 9 points, 0 comments]
• If you have a loved one writing for the NYT, please make them read this (very accessible) paper: arxiv.org/abs/2212.03551 [bsky, 3 points, 1 comments]
• I feel like you might enjoy this essay (it will be preaching to the choir in this thread, but I just think it's really neat) arxiv.org/abs/2212.03551 [bsky, 2 points, 0 comments]
• Talking About Large Language Models (LLMs) [hn, 1 points, 0 comments]
• Here's a (short) and interesting paper about it: arxiv.org/abs/2212.03551 :D Sorry. I'm done now. :D [bsky, 1 points, 0 comments]
• If you're new to Shanahan's work, I recommend starting with: - Talking About Large Language Models: arxiv.org/abs/2212.03551 - Its short follow-up: arxiv.org/abs/2412.10291 - Simulacra as Conscious E [bsky, 1 points, 0 comments]
• Understand generative Machine Learning and Large Language Models. #LLM https://arxiv.org/abs/2212.03551v2?utm_medium=social&utm_source=twitter [bsky, 0 points, 0 comments]
• The claim was that chatbots simulate *conversation*, not that LLMs simulate *brains*, which is obviously false. But if you don’t like the term, here is a more detailed working out of a similar idea: a [bsky, 0 points, 0 comments]
• This conversation focuses on Dr. Shanahan's recent paper, 'Talking about Large Language Models,' available here: https://arxiv.org/abs/2212.03551 [bsky, 0 points, 0 comments]

Related

• A Philosophical Introduction to Language Models - Part II: The Way Forward
• Understanding Conversational AI: Philosophy, Ethics, and Social Impact of Large Language Models
• The Philosophy of Language Models
• A Philosophical Introduction to Language Models -- Part I: Continuity With Classic Debates
• The Quo Vadis of the Relationship between Language and Large Language Models