Paired Texts – Google vs. History re: AI

Code.Org exec, Pat Yongpradit, shared a paper written by Google sharing their LearnLM efforts. LearnLM purports to be Google’s “family of models fine-tuned for learning, and grounded in educational research to make teaching and learning experiences more active, personal and engaging.”

Mr. Yongpradit’s LinkedIn post shares the following…

Hidden underneath all the LearnLM buzz from Google is this paper from Google DeepMind. It is a fascinating, technical read and peels back the curtain on how these tools are made and evaluated, and includes a history of related efforts.

The paper covers:

• the difficulty of translating pedagogy into gen AI prompts and the lack of good evaluation practices, compounded by the challenge of defining excellent pedagogy
• their work collaborating with students and teachers to translate high-level principles from learning science into a pragmatic set of seven diverse educational benchmarks for Gen AI in education
• a new set of fine-tuning datasets to improve the pedagogical capabilities of
  Gemini

They hope that this work can serve as a step towards developing a framework for evaluating Gen AI in education and help the AI and EdTech community realize the benefits of Gen AI in education.

Pat Yongpradit

Funny that they should mention history since the paper does not even acknowledge how much of the best of educational computing (EdTech if you must) was born in the MIT Artificial Intelligence Laboratory of the late 1960s and 70s. Seymour Papert, Cynthia Solomon, Marvin Minsky, nor any of their colleagues warrants a mention in Google’s “history of related efforts.”

Is this a hallucination or just more Silicon Valley narcissism? It’s difficult to define “excellent pedagogy” when much of your industry has contempt for teachers and schools.

So, here is the Google paper you can and should read until your eyes glaze over and my recent efforts to situate AI in a long progressive constructionist tradition.

Paired Texts

• Towards Responsible Development of Generative AI for Education: An Evaluation-Driven Approach (Google)
• CMK’s Roots: Logo, Piaget, and AI (Gary Stager – shameless, but much more readable)

Gary Stager

Veteran educator Gary Stager, Ph.D. is the author of Twenty Things to Do with a Computer – Forward 50, co-author of Invent To Learn — Making, Tinkering, and Engineering in the Classroom, publisher at Constructing Modern Knowledge Press, and the founder of the Constructing Modern Knowledge summer institute. He led professional development in the world’s first 1:1 laptop schools thirty years ago and designed one of the oldest online graduate school programs. Gary is also the curator of The Seymour Papert archives at DailyPapert.com. Learn more about Gary here.