e499 — Seeking Data

Photo by Matthew Jones on Unsplash Published 3 February 2025 E499 with Michael and Michael – all about #AI with #DeepSeek, #coldstart, #hallucinations, #BerlinWall, #AttentionEconomy, #IntentionEconomy and a whole lot more! Michael and Michael start the AI train off with a discussion about DeepSeek, which has been all over the news this past week.  This leads into a conversation about generative AI’s propensity to hallucinate, and per the Schneier article, how AI mistakes are very different from human mistakes.  Michael M gives an example of his recent experience using Ollama to run LLMs on his laptop, and the GenAI inaccurate description of some of the events leading up to the fall of the Berlin Wall, which in and of itself was precipitated by a human mistake.   Next up is a discussion on methods used to trick AI scrapers that ignore the robots.txt file – and a callback to e451 where other data poisoning techniques we