Friday, 9 June 2017

#oeb_midsummit Cognitive neuroscience and learning by @BekkeringHarold



http://www.ru.nl/publish/pages/792606/harold_bekkering.jpgHarold Bekkering reminds me of some of my Dutch family, funny feeling. Once he took the stage, it really was like listening to cousin Folkert.
Everything is connected in the brain, and the brain is a predictive machine.
A neuron at work is incoming data is output in just one unit. So in neuroscience we see that multiple inputs triggers a process into one output.
Hebbian learning: everything is connected (Hebb’ theory).
A human can adapt to one specific tone of voice and timbre. Our brain adjusts.
The brain is a prediction machine. Read Karl Friston and Andy Clark.
Distributed knowledge in the brain. Exteroceptive, autonomic (interoceptive), motoric (proprioceptive) given the conceptual multimodel representations. The brain tries to make the best things for you. It is a multimodal representations, which makes it of interest.
The whole brain is summed up in error correction.
E.g. you walk to a door (prediction is going through the door), but if a door is closed you adjust your action based upon the perception after the prediction.
Amal and Giraud, TICS 2012, beta and gamma power. The brain is only active when you make errors. Only after making an error your model is updated [interesting].
We learn more from negative feedback. Oh no! 😃
Creating a safe environment to learn from errors.
Social learning: humans are by nature social, homo imitans.
Dunbar (1998) The social Cortex, it is big because we have huge networks. The Dunbar’s number.
Social learning is needed due to the brain needing to be matured over time (baby to adult).
Good tip for learning: saying: hi (insert name) and then say the action. The brain does not ignore the calling of a name to become active.
Mirror neurons: have revolutionised neuroscience. Motorcortex was investigated, and in the lab, monkeys mimicked humans picking up peanuts. There are cells that only get active if we see someone else moving. Your brain cannot help to observe and react.
Learning analytics: extracting data, predicting data, … data driven education.
It is not useful to only learn in one learning style is the real point of interest, if you want to really learn the best thing to do is to use all styles for learning.  
Lessons for online learning
Hebbian learning: be careful with what you offer
Predictive learning: structure and provoking errors (job aids similar)
Social learning

#OEB_midsummit teaching critical thinking by Ben Nelson @MinervaSchools



https://d33z52hfhvk3mr.cloudfront.net/s3_proxy/minervaschools-production-cms-uploads/images/ben-nelson-large-flipped.width-1200.jpg Ben Nelson takes the stage. What is critical thinking? Critical thinking in itself does not exist. There are many things that together make up critical thinking. Evaluating a claim is something else than critical thinking, which requires arguments and proof.
Famous study with air traffic controllers, a manual job and taking in enormous amounts of data at split seconds. Critical thinking was similar because of far transfer. Far transfer is the holy grail of wisdom, of learning.
Point of wisdom is when you encounter a novel topic, you are able to weave your previous together to come up with a logical response. But we are seldom able to do this.
Contexts are influencing learning and recall of learning, which is why people do not often far transfer.
At Minerva far transfer is practiced in complex systems, applicable to many different fields. There is a large set of rules that underpin what creative and critical thinking is. At Minerva we took each of these rules and broken it down into approx. 100 parts and we teach them in multiple contexts. This way the students learn to use far transfer more often.
Eg. Foundational concept: Bayesian formulas for choices. Learning the concept, and distribution, the hard part is to develop the instinct of when to apply what to which data set. The bulk of the time is to train responses to triggers that point towards one or the other concepts.
Once you understand these concepts you can continue to use them over time.
Technology is used for tracking data to see whether the students interact based on the data and results in critical thinking. To do this there are also highly immersive seminars with fully active learning: at least 75% of the time need to be active in these seminars. They can debate, then others moderate the debate, interact or fill in the next part of a thought… all of these techniques enable students to be fully engaged [Inge: wondering whether this is not simply Pavlov combined with engagement]. Also using Deep Processing (repeating for deep remembrance).
A bit like the Flow, as the learner learns more by being engaged, motivated, willing to learn and focus on something specific.

#OEB_midsummit Humanising machine learning by @nellwatson

Nell Watson spoke in her natural, slow paced way, bringing her message across in a steady, transparent voice.

Image result for nell watson twitter
Essentially any task a human brain can do in roughly one second processing time can now by replicated by a machine. Machines can recognise people, transcribe between languages, esthetic interpretation, make predictions. What we have today is something of a revolution, leading us towards intuitive intelligence is a force moving us to the next revolution. Increasing human intelligence, as most geniuses are alive today. We never had so many powerful brains with enormous outputs. The trouble is that the children that are entering the school today, 65 % will take up jobs that will not even exist today. So, how can we prepare our children. Certainly we can learn, but instead of chosing a role like a job, the children will have to learn from their strengths and skills to solve challenges or problems. We are moving towards the 3 c’s critical thinking, collaboration, complex problem solving. Scratch is an amazing tool, and one million projects are built and shared every single month. These tools can pluck together different ideas and make them real, it provides new insights by becoming a maker. It is also about sharing new ideas, it is like creating a gift, and see others remix it. Originality is overrated, as remixing is one of the greatest skills of this new age. To amplify and improve these skills are essential to help people to grow from each other and let objects grow.
Creativity and fantasy are essential for growth, the capacity for fantasy can help with machine learning and learning from machines. Any kind of creativity, can now be turned into something recognizable by machines [inge: why would you want this to happen?]
Machine intelligence helps to create new layers (cfr augmentation). Real time mash-ups between fantasy are resulting in new creativity [Inge: but examples are known things turned into other known things, so how is this interesting]. There are algorithms that can fill in the gaps in knowledge (like finding the dates of a specific photograph). This leads to further thoughts of children [Inge: but why would a child want a machine to tell it, how her or his drawing would be better, how does this promote fantasy?].
Machine learning can be a mediator between humans: machines can help to create a persuasive message: empathy is now growing in machine learning [Inge: wondering whether this is personal profile diversity organized?]. Staff empathy is the most searched for quality in communication.
Teaching children moral agency is important. Questioning assumptions in society.
There will be more and more machine learning entering all of our lives: elderly relatives helped by enhanced pets – robocat by mattel. But sometimes such research are spooky, like finding vulnerable children to turn them into consumers feeling better by materials for instance.
Machine driven panopticon becomes a reality. Now 3D sensors are embedded in smartphones to scan and understand the world and merge them together. It is now able to scan an object, analyse it and to put it back into reality again (shows funny cat movie).
Good stuff: text transformed into sign language. Or impactvision, looking inside fruit to see if fruit is ok. Same for health: looking into the body with smartphones (e.g. Koen Kas). Or making the heartbeat visible to the naked eye. Looking at fotos we can find gene combinations, or finding stress in animals and humans.
Machines can help to unbundle our own personal complexity, which we not seem to be able to do.
Intelligence is going to be embedded everywhere. Machines are helping us to ‘how best to spend our money’, so machines are already supporting and advising us in certain ways. Replika is a machine, and research shows that humans do not care talking to machines. One example is a dead person turned into AI, enabling the partner to talk to their dead partner after dead.
In some of our imagination we need to destroy the bots, but in truth we find that we can outgrow our human shyness as humans. The more we get familiar with machine learning, we become more open in opening up to machines in our lives, embedding them in our lives. There is of course the good and the bad. We cannot help to put on ‘personhood’ onto those robots. So if an identity is felt, we cannot help to anthropomorphize quit robots. Brains scans show humans feel for robots (shows movie with robot being hurt). Our species have sit around campfires, sharing stories… we are driven to connect with others, it is an intrinsic part of being humans. We are bound to connect with humans and non-humans.  USC institute for creative technologiesmultisense and simvoice interaction.
We are entering a world where the humans build sensitive relationships with machines. Technologies will make scarce things abundant (e.g. to respond to loneliness). But we need to learn from the lessons of the past.  
AI needs good influences and great role models. AI is growing up and is shaping the nature of humanity. Computational ethics or machine ethics, letting machines
OpenEth.org, helping to map ethics and explore dilemma’s [Inge look this up]. Nell thinks that non-fiction is for facts, and fiction is for values.
3 billion machines will come online within before 2020… and it will increase exponentially. We want to make it easy, working on scratch for ethics which uses emoji’s to help people link together ethics to emotions. Nell believes in the kindness era, if only we can come together collaboratively, building a human heart for machines.