large language model (LLM)
AI system trained on vast amounts of text to predict the most statistically likely next word in a sequence; produces coherent text that reproduces patterns from training data but cannot generate truly novel insights
As defined in mVCxfrSMnMQ
Other definitions
AI systems trained on vast text corpora that generate human-like language, e.g. ChatGPT, Gemini, DeepSeek.
309 | Christof Koch on Consciousness and Integrated InformationA generative AI model that processes text input and produces text output, such as OpenAI's GPT, Google's Gemini, or Anthropic's Claude.
Every Essential AI Skill in 25 Minutes (2025)
Where it comes up · 34
- The AI Expert: "Super AI Will Be Unstoppable!" – What’s Coming NEXT | Stephen Wolfram18 claims
- Authors of "THE AI CON" discussing the problematic hype about AI & what those working tech should do17 claims
- AI and Quantum Computing: Glimpsing the Near Future | World Science Festival15 claims
- The Future of Knowledge Graphs in a World of Large Language Models15 claims
- ChatGPT: The Dawn of Artificial Super-Intelligence | Brian Roemmele | EP 35713 claims
- Don’t Believe AI Hype, This is Where it’s Actually Headed | Oxford’s Michael Wooldridge | AI History12 claims
- The Future of Artificial Intelligence11 claims
- Vertical AI Agents Could Be 10X Bigger Than SaaS10 claims
and 26 more
Most important claims · 10 of 189Search all claims for “large language model (LLM)” →
- Language modeling as a technology is old (going back to Claude Shannon in the 1940s) and genuinely useful in narrow applications like automatic transcription and machine translation, where it serves as one component to rank which interpretation of input is most likely given training data—but current large LLMs are trained on undisclosed data that amplifies rather than mitigates bias.Authors of "THE AI CON" discussing the problematic hype about AI & what those working tech should do
- Historical papers from Shane and Hassabis (2009 and 2016) anticipated key developments: Shane's thesis identified compression of Wikipedia as a test for AI (which became the loss function for LLMs), and Hassabis's 2016 neuroscience-AI paper identified attention as a critical capability (which was later crystallized in transformers).Demis Hassabis — Scaling, superhuman AIs, AlphaZero atop LLMs, AlphaFold
- When evaluating whether to use LLMs for a task, the critical questions are: what is the specific task you're trying to automate, what input is available, how will users be using the output, and how will users have the information they need to use the output in an appropriate and reliable way?Authors of "THE AI CON" discussing the problematic hype about AI & what those working tech should do
- Knowledge and accountability in science depend on understanding the positionality and responsibility of people who did the work; when work is attributed to LLMs or generated using opaque processes, the chain of accountability breaks down and interpretation becomes untethered from authorial intention.Authors of "THE AI CON" discussing the problematic hype about AI & what those working tech should do
- People interpret synthetic text from LLMs as meaningful because humans are exceptionally good at making sense of language by imagining what a person was trying to communicate, but this human meaning-making capacity is applied to text that was generated mechanically without intent or understanding.Authors of "THE AI CON" discussing the problematic hype about AI & what those working tech should do
- There is a fundamental misunderstanding of science in claims that LLMs will do science, imagining that anything making progress in LLMs is the only science, when in fact science across multiple fields uses diverse mechanisms and is funded through NIH and NSF grants which are being gutted at the federal level.Authors of "THE AI CON" discussing the problematic hype about AI & what those working tech should do
- LLMs represent a new computing paradigm similar to how cloud and mobile computing were new paradigms in 2005, creating an open question about where value and opportunities lie for startups, with subsequent massive value creation across multiple company categories.Vertical AI Agents Could Be 10X Bigger Than SaaS
- Large language models are not committed to a single answer when playing the 20 questions game; rather, all possible answers consistent with previous responses exist in superposition until the model is forced to collapse that distribution by being asked directly what the answer is, which fundamentally differs from how human consciousness works where we have already committed to an object in mind.Human vs. Machine Consciousness | Imperial’s Murray Shanahan
- A forthcoming bioRxiv paper shows formally (assuming IIT) that two systems can be Turing-equivalent — identical inputs producing identical outputs, indistinguishable functionally — yet radically different phenomenally: a small set of nonlinear gates can have high system-level causal power while a 116-gate von Neumann machine computing the same function has near-zero whole-system causal power, because brain neurons each connect to ~50,000 others with heavy overlap whereas a transistor talks to only two or three, so an LLM can be functionally rich yet not conscious at all.309 | Christof Koch on Consciousness and Integrated Information
- Philosophical questions about AI (does it think? is it conscious? does it understand?) have shifted from pure philosophy into experimental science because we now have systems to test; questions that were purely metaphysical are becoming empirical investigations.Don’t Believe AI Hype, This is Where it’s Actually Headed | Oxford’s Michael Wooldridge | AI History