The model can give a sequence of outputs that resemble an answer ..
summary
Claims that modern generative models take a sequence of inputs and produce a sequence of outputs that resemble an answer while often lacking grounding, shows how early products shipped relying on that resemblance, and documents the concrete user failure mode of confusing plausibility with correctness. Offers practical countermeasures: add deterministic grounding steps, automated consistency checks, and adversarial input tests to surface hallucinations before deployment. sequence-to-sequence outputs plausible but ungrounded
AI-written summary · about
related talks
55mAI Showdown: The Good, The Glitchy, and The Risky
19mHumans vs. Machines
06mInterview with Spela Poklukar
41mThe Unreliable Computer Revolution - Embracing the Double-Edged Sword of GenAI
54mDeterministic GenAI Outputs With Provenance
22mGenerative AI - A Pragmatic Business View
29mThe Hater's Guide to Dealing with Generative AI
46mConversations with ChatGPT: illusion or reality?