Janelle Shane
AI chatbots have linguistic slips when they go off-script adds Shane as the Marketplace Tech guest explaining why chatbots can unexpectedly insert foreign-language words into otherwise fluent English. Her explanation ties multilingual training data, mixed domains, and dialogue formatting to Chatbot Code Switching, while her caution about a chatbot’s own post-hoc explanation creates Chatbot Self-Explanation Uncertainty.
Janelle Shane appears in Uncanny AI: Why AI bots remember random, sometimes useless information as the Marketplace Tech guest explaining why chatbots can remember random personal facts in socially awkward ways. She describes how large language models can draw on chat history or a separate memory file, and why a saved detail may be treated as something to reuse even when it is incidental.
Her role in the source is to separate memory storage from memory judgment. A chatbot such as Claude may correctly retain a detail, but still lack human-like proportion, appropriateness, and sensitivity about when to mention it. That makes her explanation central to Chatbot Memory Salience Failure, Personal AI Memory, and Persistent Agent Memory.
Connections
- Marketplace Tech and [[MeganMcCartyCorino|Megan McCarty-Corino]] - show and interview context.
- AI Weirdness Blog - science-communication project named in the August 2026 source.
- Claude, ChatGPT, Anthropic, and Grok - chatbot examples and adjacent model-behavior branch.
- Chatbot Code Switching, Chatbot Domain Bleedthrough, Chatbot Self-Explanation Uncertainty, Chatbot Memory Salience Failure, Personal AI Memory, Persistent Agent Memory, and Chatbot Safety Guardrail Decay - concepts her explanation sharpens.