Published inTowards Data ScienceHuman and Artificial General Intelligence Arises from Next Token PredictionWhat if human intelligence derives from successful next token prediction, and what if next token prediction is a sufficient objective…Apr 287Apr 287
Published inAI AdvancesIs “Good” AI Harder than “Bad” AI?Many definitions of AI alignment describe a goal of “making AI systems follow human values” while attempting to skirt difficult moral…Mar 29Mar 29
Published inAI AdvancesAI Alignment and Moral Philosophy for Artificial General IntelligenceIn this post I summarize a few of my thoughts on alignment and moral philosophy for safe artificial general intelligence.Mar 1913Mar 1913
Published inTowards Data ScienceChatGPT Is Not a DoctorHidden dangers in seeking medical advice from LLMsFeb 237Feb 237
Hands in Human Dreams Look AI-GeneratedIt’s fun to see the surprising images of human hands produced by generative image models. Why are there so many fingers — or so few? It…Feb 61Feb 61
Just for Fun: Llama2 Chats with ChatGPT3.5I decided to have Llama2 and ChatGPT3.5 chat with each other just for fun. Llama2 is available here and ChatGPT3.5 here. I typed the first…Nov 29, 2023Nov 29, 2023
Published inTowards Data ScienceBias, Toxicity, and Jailbreaking Large Language Models (LLMs)A review of recent research on concerning characteristics of LLMsNov 28, 20234Nov 28, 20234