Sometime last year I picked up on Kenneth Stanley’s and Joel Lehmann’s 2015 book called Why Greatness Cannot Be Planned - The Myth of the Objective. In the book they develop an argument for an advanced teleology based on experiments with synthetic processes of knowledge acquisition in the context of AI, ALife, and Learning. The argument roughly says, that if you want to reach a goal, that is ambitious in the sense that the exact sequence of steps (the route) which will get you there, is not known, then accumulating possible steps is a better strategy than heading directly into the direction of the goal. That’s because chances are, that some of these steps will turn out, but unforseeably so, to be precisely what is needed to make the next move when negotiating the route. So far so good
Posts tagged learning
Crows aren’t born knowing how to make these tools; they teach the technique to their young. And they can improvise, too. In one lab experiment, a crow bent the end of a wire using the edge of a glass as a cantilever. It used the hooked wire to retreive another stick, which was long enough to reach some food it wanted. So it used one tool to make another tool — and then used that tool to grab still another tool.
The human brain is a sophisticated learning machine, forming rules by memorizing everyday events (“sparrows can fly” and “pigeons can fly”) and generalizing those learnings to apply to things we haven’t seen before (“animals with wings can fly”). Perhaps more powerfully, memorization also allows us to further refine our generalized rules with exceptions (“penguins can’t fly”). As we were exploring how to advance machine intelligence, we asked ourselves the question—can we teach computers to learn like humans do, by combining the power of memorization and generalization? It’s not an easy question to answer, but by jointly training a wide linear model (for memorization) alongside a deep neural network (for generalization), one can combine the strengths of both to bring us one step closer. At Google, we call it Wide & Deep Learning. It’s useful for generic large-scale regression and classification problems with sparse inputs (categorical features with a large number of possible feature values), such as recommender systems, search, and ranking problems.
Next, they designed an abuse-evading algorithm to help the robot avoid situations where tiny humans might gang up on it. Literally tiny humans: the robot is programmed to run away from people who are below a certain height and escape in the direction of taller people. When it encounters a human, the system calculates the probability of abuse based on interaction time, pedestrian density, and the presence of people above or below 1.4 meters (4 feet 6 inches) in height. If the robot is statistically in danger, it changes its course towards a more crowded area or a taller person. This ensures that an adult is there to intervene when one of the little brats decides to pound the robot’s head with a bottle (which only happened a couple times).
So let’s address our children as though they are our children, and let us revel in the fact they are playing and painting and creating; using their first box of crayons, and us proud parents are putting every masterpiece on the fridge. Even if we are calling them all %E2%80%9Cnightmarish%E2%80%9D–a word I really wish we could stop using in this context; DeepMind sees very differently than we do, but it still seeks pattern and meaning. It just doesn’t know context, yet. But that means we need to teach these children, and nurture them. Code for a recognition of emotions, and context, and even emotional context. There’s been some fantastic advancements in emotional recognition, lately, so let’s continue to capitalize on that; not just to make better automated menu assistants, but to actually make a machine that can understand and seek to address human emotionality. Let’s plan on things like showing AGI human concepts like love and possessiveness and then also showing the deep difference between the two.
Nocturnal sleep and daytime napping facilitate memory consolidation for semantically related and unrelated word pairs. We contrasted forgetting of both kinds of materials across a 12-hour interval involving either nocturnal sleep or daytime wakefulness (experiment 1) and a 2-hour interval involving either daytime napping or wakefulness (experiment 2). Beneficial effects of post-learning nocturnal sleep and daytime napping were greater for unrelated word pairs (Cohen’s d = 0.71 and 0.68) than for related ones (Cohen’s d = 0.58 and 0.15). While the size of nocturnal sleep and daytime napping effects was similar for unrelated word pairs, for related pairs, the effect of nocturnal sleep was more prominent. Together, these findings suggest that sleep preferentially facilitates offline memory processing of materials that are more susceptible to forgetting.
“ A umbrella can be grasped like this, using deep_features dgf_164, dgf_64, and dgf_69”
“It seems that, if you just present the correct information, five things happen,” he said. “One, students think they know it. Two, they don’t pay their utmost attention. Three, they don’t recognize that what was presented differs from what they were already thinking. Four, they don’t learn a thing. And five, perhaps most troublingly, they get more confident in the ideas they were thinking before.” Confusion is a powerful force in education. It can send students reeling toward boredom and complacency. But being confused can also prompt students to work through impasses and arrive at a more nuanced understanding of the world.
“It’s not the subject of calculus as formally taught in college,” Droujkova notes. “But before we get there, we want to have hands-on, grounded, metaphoric play. At the free play level, you are learning in a very fundamental way—you really own your concept, mentally, physically, emotionally, culturally.” This approach “gives you deep roots, so the canopy of the high abstraction does not wither. What is learned without play is qualitatively different. It helps with test taking and mundane exercises, but it does nothing for logical thinking and problem solving. These things are separate, and you can’t get here from there.”
We’ve all had those, “Aha!” moments when we finally get an idea. The problem is most of us don’t have a systematic way of finding them. The typical process a student goes through in learning is to follow a lectures, read a book and, failing that, grind out practice questions or reread notes. Without a system, understanding faster seems impossible. After all, the mental mechanisms for generating insights are completely hidden.
There are nine or so principles to work in a world like this: Resilience instead of strength, which means you want to yield and allow failure and you bounce back instead of trying to resist failure. You pull instead of push. That means you pull the resources from the network as you need them, as opposed to centrally stocking them and controlling them. You want to take risk instead of focusing on safety. You want to focus on the system instead of objects. You want to have good compasses not maps. You want to work on practice instead of theory. Because sometimes you don’t why it works, but what is important is that it is working, not that you have some theory around it. It disobedience instead of compliance. You don’t get a Nobel Prize for doing what you are told. Too much of school is about obedience, we should really be celebrating disobedience. It’s the crowd instead of experts. It’s a focus on learning instead of education.