Matthew Barnett’s Blog philosophy, effective altruism, artificial intelligence, the future, economics
A reply to Michael Huemer on AI 117 implied HN points • 17 Feb 23 🕹 Technology AI Machine Learning Language Models Artificial Intelligence Chatbots We don't fully understand how AI like ChatGPT works and it may have some true understanding. AI models like ChatGPT are not perfect and do have limitations. It's important to differentiate between what an AI model is optimized to do and what it actually does.
GPT-4 takes Bryan Caplan's midterm and gets an A 3 HN points • 15 Mar 23 🕹 Technology AI Education Economics Innovation Automation Even though AI like GPT-4 scored well on an economics exam, progress from the past decade is still astounding Challenging beliefs and bets on AI performance on exams can lead to interesting outcomes and dialogues Specialization based on comparative advantage can increase overall productivity and efficiency in modern economies