Stephen Downes

A writer called Paris Marx argues that cyberlibertarianism must die if there's to be any hope of a better future for the internet. "Instead of solely fighting for digital rights, it's time to expand that focus to digital sovereignty that considers not just privacy and speech, but the political economy of the internet and the rights of people in different countries to carve out their own visions for their digital futures that don't align with a cyberlibertarian approach." I remember arguing for something similar back in the late 90s. For some reason, it didn't take.

I wrote about the Social Web Foundation (SWF) last week and suggested there's some controversy about whether the term 'social web' refers to the entire fediverse or to ActivityPub implementations alone. This post is an instance of that controversy boiling over. I'm always wary of attempting to place the governance of a decentralized network under a single entity, despite the calls for governance, coordination and organization around a collective purpose. 

This has nothing to do with online learning (except tangentially with the recent debates about student visas in my country) but I need to create a post to test a bug fix and I do have an opinion on this. Not on the UAE, or even on the United States, but on the debate described here on the subject "Does open borders benefit humanity by reducing poverty and boosting the economy?" I believe it does (and so does Bryan Caplan in this article). And it seems to me that we cannot have freedom - real freedom, which applies equally to all people - until humans can move about the planet at least as freely as goods and capital.

"Descript helps remove filler words and creates clips and descriptions that now reduce the three hours to one. I just saved two hours," writes Dean Shareski. But then the kicker: "what will I do with those 2 hours saved?" The answer? More work. "Those of us who work in education and many other professions have never-ending jobs. All of us could and some of us do work 80 hours a week and we never finish. Our to-do lists never get completely crossed off." There's truth to this, and yet at the same time I can't help but think it's a poor profession that is managed thus.

This article raises the issue of 'pedagogical alignment' of large language models. What does 'pedagogical alignment' even mean in this context? It "involves breaking complex problems into manageable steps and providing hints and scaffolded guidance rather than direct answers." That strikes me as very narrow, but let's persist. The authors "propose a novel approach to achieve pedagogical alignment by modeling it as learning from human preferences (LHP)... to represent desired teaching behaviors as preferences, enabling more nuanced optimization." And this (of course) requires data "to quantitatively measure an LLM's tendency to provide step-by-step guidance versus direct answers, offering a robust metric for pedagogical alignment." And so plays out the paper. I would find such an AI very frustrating. I think we need a model of learning that is something that isn't spoonfeeding.

In a nutshell, "agentic AI... uses sophisticated reasoning and iterative planning to autonomously solve complex, multi-step problems. And it's set to enhance productivity and operations across industries." Potential applications include customer service, content creation, software engineering and healthcare.