What Is It Like to Be an LLM?
summary
Claims that modern LLMs produce plausible text by pattern‑matching across massive corpora, not by lived experience, so they have no phenomenology and no agency; hallucinations arise when the model’s semantic anchoring links concepts along improbable paths; danger appears if systems are granted real‑world agency without embodiment or responsibility; practical takeaway is using LLMs as an inductive testing tool for deductive software, while keeping any action‑taking strictly mediated by humans.
AI-written summary · about
chapters (10)
- 0:00Introduction
- 1:05AI and LLMs philosophy
- 3:56Thought experiments explained
- 5:39Descartes and radical doubt
- 9:08Nagel’s bat limits
- 13:06Life as an LLM
- 18:21Experience and agency
- 21:51Ethics and responsibility
- 23:18Closing remarks
- 24:00Q&A
related talks
49mA Practical Approach to Using LLMs: Peculiarities and Pitfalls
16mLLMs are not the product
21mThe Coherence Trap: Why LLMs Feel Smart (But Aren’t Thinking)
39mLLMs and Hallucinations: Understand, Measure, Control
39mHallucination-Free LLMs: Strategies for Monitoring and Mitigation
32mWhat can we expect of LLMs as software engineers?
35mReading the Mind of an LLM
01h 00mFrom LLM Theory to Practical Agentic Implementations