Welcome to Sensemaking, AI, and Learning (SAIL) a weekly+ look at AI and its impact on education.
What's the best new idea in education that you've encountered in the last decade?
I'm not convinced that there is much new in education lately, other than perhaps scale and the role of analytics. Obviously I place AI in the league of significant systems-level change, but its true potential may be that it allows us to realize existing ideas. Dewey. hooks. Engestrom. Keller. Bloom. We can realize their ideas today in the form of personalized learning, supportive tutoring, adaptive feedback, social and peer-based learning, and self-direct learning. What AI enables is the realization of bold ideas in education that in the past have been too hard or too expensive to implement. We can, at scale, today teach and support students across the entirety of their life's learning needs. Our education systems aren't yet structured to do this, but the technological capability is now here.
AI and Learning
Colleges are doing what they do best when confronted with change: hire faculty in new domain areas and spin up academic programs. Now it's the AI gold rush: "When it comes to artificial intelligence, some universities are going big—very big."
Y'all should join us in Denver July 13 to discuss AI in Education from leadership perspectives. If your institution is a GRAILE member, please reach out and we'll share the discount code.
ChatGPT launches on iPhone (in USA for now). It's fast. An interesting way to engage. The big lesson of ChatGPT remains: a good user experience with speed of development matters (almost) as much as the core model. Google's PaLM was here first and was more powerful than GPT3/3.5. Delivering and letting people play matters. Hence ChatGPT is in the lead.
Wearable AI. Short video with a few technical specs. "AI may automate a majority of knowledge work in a relatively short time period We will soon share the world with a new class of entities - ones which live in servers, phones, and wearables that dot our bodies"
Will we end up with domain specific AI models or will we have a few large everything models? In healthcare, it may be domain specific. "Hippocratic AI tested its model against GPT-4 on 114 different benchmarks, including exams and certifications used for doctors, nurses, dentists, pharmacists, audiologists and medical coders, among others. Hippocratic beat GPT-4 on 105, tied on six and lost on three." Does this mean we will have models like we have courses in universities? A Math model. Biology model. Chemistry model.
How Technology has changed the world. Incredible. Humbling. Awe-inspiring.
Impact of AI
BT to cut 55,000 jobs, 1 in 5 due to AI.
Buzzfeed CEO Jonah Peretti, speaking at Investor Day, just one week after laying off its entire News team: "Over the next few years, generative AI will replace the majority of static content"
47% of Americans have used ChatGPT for investment advice. These AI tools are not only information search and retrieval and generative outputs. They are social, collegial, relational.
61% of Americans say AI threatens humanity's future. What will we do about it? Likely not much. We're still infatuated. ""There's no way to put this genie in the bottle. Globally, this is exploding," said Senator Cory Booker"
Can AI solve the loneliness (or general mental health) crisis? Maybe. "With the advent of generative AI, conversational AI and natural language processing, the thought of using artificial intelligence systems to provide human companionship has now become mainstream." Technology creates problems that only more technology can solve.
Calls for regulation. "Emerging information technologies, including ChatGPT, require proper stewardship. An intergovernmental panel to synthesize the evidence offers the best path forward."