🔮 AI & pedagogy; gene-editing for Mars; new robot vocabulary; caterpillars, whales and Bjork++ #178
|Azeem Azhar||Aug 12, 2018|
I am on my summer holidays. While I am away, Linda Liukas has kindly agreed to look after Exponential View this week.
Linda and I first met in Copenhagen in 2016. She mesmerised an audience with her talk on how to teach young children about computing. I hope she will take you on a similar journey of whimsy and pedagogy.
Enjoy her Exponential View,
I’m Linda (@lindaliukas) and I write and illustrate children’s books about technology: code, computers, the Internet. I’m currently finalising a book on how to explain machine learning to a six-year-old.
My work exists at the intersection of three things: early childhood education, technology and play. The first draws inspiration from Montessori, John Dewey and Reggio Emilia, the second from computing culture of the 1960s, the third is the small, soft, communal, subjective side of learning that is often forgotten.
When I first started writing books about technology for kids, I knew almost nothing about pedagogy. For me, computing was magical, charming and imaginative — but the materials teaching it often dull and uninspiring. And if computer science is becoming the way to understand the world, instead of grammar classes, I wish to see more poetry lessons.
Play is the highest form of research. Welcome to my edition of Exponential View.
Dept of the near future
🇯🇵 Japan is the only country to formally decide to make a 100-year life a national project. (See also from Japan, a stunning essay on the role of ghosts in this modern, secular society, and what ghostlore means for its neighbors: “In North Korea, there’s nothing more real than the deceased first and second dictator.”)
💯 “As a species, humans prefer power to truth. We spend far more time and effort on trying to control the world than on trying to understand it,” writes historian Yuval Noah Harari, arguing that post-truth world is modus operandi for Homo sapiens.
👾 Forming human colonies on Mars will require a new breed of genetically-edited humans who are resistant to radiation, can live with less oxygen, have improved memory, and smell better.
👁️ “People in the UK are so conditioned to living in a surveillance state that most of us don’t even realise that we do,” EV reader Stephanie Hare sounds the alarm: tech companies will solve the accuracy problem in their technologies, but it’s up to us to decide what access to biometrics data means in our society.
Dept of kindergarten
If intelligent machines are trained, not programmed, how similar would a kindergarten for AI be to a human one? And what could real kids teach robots? One of my favorite things in the research world of the 1960s and 1970s is how Minsky, Papert and other AI luminaries are referenced in early childhood research and vice versa. I wish these conversations would keep happening.
🤔 Curiosity. Jean Piaget, a famous child development psychologist once said the best preschool curriculum is about keeping children curious, making them wonder, and offering them real problem-solving challenges, rather than trying to transfer the knowledge. Pair with Curiosity-driven Exploration by Self-supervised Prediction (2017).
🤸 Movement. Hajimete No Otsukai is a Japanese TV series of children as young as two running their first errands in their neighborhood. It's exciting and touching to watch the children overcome the obstacles both physical and emotional and develop self-efficacy. Pair with the gym environments of OpenAI (2018) or Carnegie Mellon’s dribbling avatars (2018).
🖥️ Metaphysics. Sherry Turkle wrote in The Second Self in the 1980s about children’s relationship with computers as a new kind of object - psychological, yet a thing. "Childhood animism, this attribution of the properties of life to inanimate objects, is only gradually displaced by new ways of understanding the physical world in terms of physical processes. In time the child learns that the stone falls because of gravity; intentions have nothing to do with it. [...] And so a dichotomy is constructed: physical and psychological properties stand opposed to one another in two great systems. The physical is used to understand things, the psychological to understand people and animals." But the computer is the exception, especially at this age when children grow up having conversations with a computer. Pair with the Allen Institute Research on reasoning and common sense (2018).
In the spirit of life-long learning through play, I invite you to take some time this weekend to play alone or with your family with some of the best resources to understand and experiment with machine learning:
Machine Learning for Kids by Dale Lane. You train machine learning models for text classifying, numbers and recognising images, but with Scratch and kid-friendly context!
Wolfram Alpha's machine learning resources allow for practical and fun exercises around images, color and translations.
We could all learn from the Koreans and the textbooks they use to teach AI.
In a more academic spirit:
Fermat's Library picks an interesting computer science paper each week and allows the community to create annotations in the margins. One of the latest picks was Doug Engelbart’s Developing the Underlying Concepts for Contemporary Computing.
Making learning visible. A lot of the misconceptions and fears around machine learning I come across have to do with thinking it's a black (magic) box. This is why interactive, well-paced and explorable essays are so important. See for example Building Blocks of Interpretability or take part in Explorable Explanations online jam!
How do machines learn? The toy tasks Facebook uses to teach AI include stories, the reading comprehension tests with questions on Apollo landing, private schools and Victoria & Albert Museum, as well as a corpus of sentences your six years old could complete, but that still baffle AI. Not so scary after all!
Short morsels to appear smart at dinner parties
📹 How art influences technology - and how Björk accidentally shaped the way we see humanoid machines in 1999.
As robots become a part of our world we need more vocabulary to describe them. What if the collective noun for robots was school? (“Murder of crows, school of robots”). Here are some other ideas.
🏖️ "Sandbox has taken on a pacifist and constructive meaning in games, but it could have gone the other way. Tabletop sandboxes were used as educational tools for small children in Dewey’s era, but long before, they were military tools for diagramming war strategy." The Design of Childhood by Alexandra Lange.
Competing in a chess tournament as a girl: a revealing short portraying early discrimination; a narrative we need to change.
Old, but gold: once upon a time the Internet could fit on a simple map.
Sneak-peak into the store of the future, which is very much the present in Hangzhou, China.
Silicon Valley is in dire need of ethical guidance. Ethical OS is a toolkit containing important questions and possible scenarios technologists should be aware of before launching their product into the wild. Contains a handy checklist, too!
👄 Wearables that read our thoughts? Possible, but not quite what you imagine.
The purpose of guest curators is to give you something different to my usual ramble through the near future. I think we can agree Linda has done just that.
Have a great week!