SAIL: AI and Learning, Future of Education
Welcome to Sensemaking, AI, and Learning - a weekly look at how AI and technology are progressing and impacting education.
A few thoughts:
-What might universities look like? I don't thinkAI is the leading edge of change. A few reflections here: This time is different. Part 1.
-Over the next week, we (through the Global Research Alliance for AI in Learning and Education - GRAILE) are going to launch a series of workshops, online course on human and artificial cognition, webinars, and AI literacy courses. We are hosting an in-person workshop in Arlington Texas on March 13 during the Learning Analytics Conference where we will be targeting AI leadership development and institutional support for capacity building. If you're interested in your university joining GRAILE, please let me know.
-Foundation models underpin the current explosion of interest in AI. Foundation models are:
"trained on broad data (generally using self-supervision at scale) that can be adapted to a wide range of downstream tasks. These models, which are based on standard ideas in transfer learning and recent advances in deep learning and computer systems applied at a very large scale, demonstrate surprising emergent capabilities and substantially improve performance on a wide range of downstream tasks. Given this potential, we see foundation models as the subject of a growing paradigm shift, where many AI systems across domains will directly build upon or heavily integrate foundation models. Foundation models incentivize homogenization: the same few models are repeatedly reused as the basis for many applications."
Basically, it's a "model as platform" concept, enabling many applications and results to be built from (on) the foundation model. We're already seeing this with the large language model (LLM) GPT-3. Almost two years ago, over 300 companies were building on this model.
My question: What would a foundation education model - i.e. "large learning model"? - look like (i.e. a model built on learning data)? What would it do? For example, would there be value if all the world's university syllabi were mined or all the textbooks used by universities? Or are LLMs and existing foundation models sufficient and don't require domain specific development? The foundation model paper grapples with applications (starting p.67).
Education
Whew. The AI and education hype machine is in full cycle. We are at an existential state it appears. Which is obviously wrong. AI won't make the transformative impact on education we're hearing...at least not over the next year. There will, however, be a rather substantive battle between student knowledge artifacts and assessment.
-How AI could be meaningfully used by teachers and students
-Vitomir Kovanovic insists that “You cannot stop it and, even if you could, it’s a temporary solution. The next one you won’t be able to. It’s futile. And you shouldn’t be doing it, you should be teaching them how to use it – they’re going to use it in the workplace anyway. It’s like having a driving school but teaching people how to ride horses.”
-Could ChatGPT get a Wharton MBA? Yes.
AI Development:
Increasingly, this can occupy this entire newsletter. Here are a few things to note:
-Microsoft considers investing $10b in OpenAI (ChatGPT parent), valuing the company at $29b. Marcus wonders if it's a deal for Microsoft or a steal for OpenAI.
-ChatGPT has a survey/waitlist for paid users. I responded to the survey. I'd pay. Just not sure how much - $50/month?
-We're all going to get sick of ChatGPT real quick, simply due to the hype insanity. Here are alternatives.
-Feel like you've lost the conversation about ChatGPT? Here's a primer.
Social Implications
-You just can't let the technologists interact with social issues, it appears. Koko exposed 4000 people to ChatGPT mental health support. The lack of awareness about possible negative aspects of this leaves me fully convinced that there is no horrible poor decision around AI that someone somewhere will not be willing to make.
Learning
-ChatGPT and GPT3 are different (ChatGPT runs on GPT3.5) and have different avenues for interaction and content creation. Here's an informative tutorial on writing prompts for GPT3.
-Looking at your AI literacy? Open courses.