Post has attachment
Why the "Unknown Unknowns" Critique of Science Gets Us Nowhere
When challenged that a cherished belief is contrary to established scientific knowledge, many fall back on "we don't know what we don't know, so all that could change." In my latest Medium post I explain why that doesn't get us anywhere.
When challenged that a cherished belief is contrary to established scientific knowledge, many fall back on "we don't know what we don't know, so all that could change." In my latest Medium post I explain why that doesn't get us anywhere.
Post has attachment
Entropy and Complexity: The surprising Paradox Behind Our Universe
One of the many big ideas in physicist Sean B. Carroll’s The Big Picture: On the Origins of Life, Meaning, and the Universe Itself is the concept that entropy can drive increasing complexity. In fact if our universe did not have increasing entropy as one of its fundamental components, we would not have the complex world we see today, including you and me.
Continue reading at the link below
One of the many big ideas in physicist Sean B. Carroll’s The Big Picture: On the Origins of Life, Meaning, and the Universe Itself is the concept that entropy can drive increasing complexity. In fact if our universe did not have increasing entropy as one of its fundamental components, we would not have the complex world we see today, including you and me.
Continue reading at the link below
Post has attachment
Physicists have created an “impossible” state of matter that could power quantum computers http://flip.it/QuUDRq
"Quantum computers require atoms to exist in entangled states, where changing the state of one automatically causes the other to change state too. At present, such states can be achieved only at extremely low temperatures. Lurkin got all the nitrogen atoms in his dirty diamond to change position together at a constant frequency—meaning they were held in quantum entanglement—and he did it at room temperature."
"Quantum computers require atoms to exist in entangled states, where changing the state of one automatically causes the other to change state too. At present, such states can be achieved only at extremely low temperatures. Lurkin got all the nitrogen atoms in his dirty diamond to change position together at a constant frequency—meaning they were held in quantum entanglement—and he did it at room temperature."
Post has attachment
EPA head Scott Pruitt: "There's this one guy on YouTube who says the earth is flat, so we can't say there's scientific consensus on a round earth yet. We must continue the debate."
Post has attachment
Post has attachment
Post has shared content
As usual, a great summary and analysis by +Eli Fennell
The Physics of Why Neural Networks Are So Good At Problem Solving
Neural network AI's are really good at problem solving. Really, really good. Too good, in fact. In seeming violation of the laws of probability, neural networks can be trained to provide very accurate solutions to problems whose pure mathematical odds are well beyond their computational power.
This isn't nitpicking. This is a fundamentally important issue in science and computing: how do artificial neural networks achieve high accuracy despite the overwhelming probabilities against doing so? It seems impossible.
Now, a team of researchers from Harvard University and MIT believes they've solved the problem, and the key to their power to address complexity is, in fact, simplicity. While the 'pure math' of physical probabilities can seem overwhelming, in reality the laws of physics constrain the real outcomes to a much smaller subset.
By analogy, consider Google's DeepMind's growing ability to simulate realistic human speech patterns (https://deepmind.com/blog/wavenet-generative-model-raw-audio/): on paper, human voices could produce an almost infinite number of patterns of tone and frequency, but in reality human speech patterns tend to fall within a narrower subset of tones and frequencies. As a result, an AI programmed to produce human sounding speech based on abstract mathematical models would struggle, whereas an AI trained by studying millions of real human speech patterns learns to identify a narrower subset and thereby produce a passable fascimile.
In addition, complex configurations can often be reduced to a simpler set of physical properties, e.g. by simply knowing that atoms are configurations of protons, neutrons, and electrons you can work out on paper the entire Periodic Table of Elements, but even without going to that length, one can easily tell whether something is an atom or not, based on the presence or absence of those fundamental building blocks. Configured of protons, neutrons, and electrons? Atom. Not configured of those things? Not an atom.
Reduced even further, one need not know more than a few of the 'sensory' properties of matter to know if something is matter or not, e.g. a rock is clearly matter because it has height, width, depth, solidity, and texture, whereas the vacuum of space, having none of these, is clearly not any form of matter. It is possible to imagine other things besides matter possessing similar properties to matter, but no human alive today can credibly claim to have encountered such exotic substances.
This may help to explain the discovery of 'Slime Mold Learning' (https://www.washingtonpost.com/news/speaking-of-science/wp/2016/04/30/this-weird-slimy-single-cell-organism-can-learn-without-a-brain/), whereby single celled slime molds actually demonstrate the learning behavior of Habituation. Classically, slime mold is considered insufficiently complex to 'Learn' anything due to its lack of neurons or memory, but if a very basic form of learning like Habituation depends only on these cells chemically responding to a subset of probabilities, then 'Learning' becomes something, at its lowest levels of sophistication at least, that doesn't require complex neural networks, organic or artificial.
This could have important ramifications for our understanding of how the human brain learns, as well.
#AI #ArtificialIntelligence #MachineLearning
Neural network AI's are really good at problem solving. Really, really good. Too good, in fact. In seeming violation of the laws of probability, neural networks can be trained to provide very accurate solutions to problems whose pure mathematical odds are well beyond their computational power.
This isn't nitpicking. This is a fundamentally important issue in science and computing: how do artificial neural networks achieve high accuracy despite the overwhelming probabilities against doing so? It seems impossible.
Now, a team of researchers from Harvard University and MIT believes they've solved the problem, and the key to their power to address complexity is, in fact, simplicity. While the 'pure math' of physical probabilities can seem overwhelming, in reality the laws of physics constrain the real outcomes to a much smaller subset.
By analogy, consider Google's DeepMind's growing ability to simulate realistic human speech patterns (https://deepmind.com/blog/wavenet-generative-model-raw-audio/): on paper, human voices could produce an almost infinite number of patterns of tone and frequency, but in reality human speech patterns tend to fall within a narrower subset of tones and frequencies. As a result, an AI programmed to produce human sounding speech based on abstract mathematical models would struggle, whereas an AI trained by studying millions of real human speech patterns learns to identify a narrower subset and thereby produce a passable fascimile.
In addition, complex configurations can often be reduced to a simpler set of physical properties, e.g. by simply knowing that atoms are configurations of protons, neutrons, and electrons you can work out on paper the entire Periodic Table of Elements, but even without going to that length, one can easily tell whether something is an atom or not, based on the presence or absence of those fundamental building blocks. Configured of protons, neutrons, and electrons? Atom. Not configured of those things? Not an atom.
Reduced even further, one need not know more than a few of the 'sensory' properties of matter to know if something is matter or not, e.g. a rock is clearly matter because it has height, width, depth, solidity, and texture, whereas the vacuum of space, having none of these, is clearly not any form of matter. It is possible to imagine other things besides matter possessing similar properties to matter, but no human alive today can credibly claim to have encountered such exotic substances.
This may help to explain the discovery of 'Slime Mold Learning' (https://www.washingtonpost.com/news/speaking-of-science/wp/2016/04/30/this-weird-slimy-single-cell-organism-can-learn-without-a-brain/), whereby single celled slime molds actually demonstrate the learning behavior of Habituation. Classically, slime mold is considered insufficiently complex to 'Learn' anything due to its lack of neurons or memory, but if a very basic form of learning like Habituation depends only on these cells chemically responding to a subset of probabilities, then 'Learning' becomes something, at its lowest levels of sophistication at least, that doesn't require complex neural networks, organic or artificial.
This could have important ramifications for our understanding of how the human brain learns, as well.
#AI #ArtificialIntelligence #MachineLearning
Post has attachment
An excellent, if disturbing read, I found via +David Amerland's Weekly Read today.
I've dedicated the rest of my life to embracing uncertainty and ambiguity (after spending way too many years trapped in dogmatism). I still fail often, because the evolutionary forces toward embracing the easiest certainty at hand, and our weak ability to detect ourselves dong that, war against it.
But I'm convinced embracing uncertainty, epistemological agnosticism, is the way forward.
In other words, intuition, like attention, is “an intentional, unapologetic discriminator [that] asks what is relevant right now, and gears us up to notice only that” — a humbling antidote to our culture’s propensity for self-righteousness, and above all a reminder to allow yourself the uncomfortable luxury of changing your mind.
I've dedicated the rest of my life to embracing uncertainty and ambiguity (after spending way too many years trapped in dogmatism). I still fail often, because the evolutionary forces toward embracing the easiest certainty at hand, and our weak ability to detect ourselves dong that, war against it.
But I'm convinced embracing uncertainty, epistemological agnosticism, is the way forward.
In other words, intuition, like attention, is “an intentional, unapologetic discriminator [that] asks what is relevant right now, and gears us up to notice only that” — a humbling antidote to our culture’s propensity for self-righteousness, and above all a reminder to allow yourself the uncomfortable luxury of changing your mind.
Post has attachment
#BOOM
Science isn't a perfect process. There are still human beings involved. But it is by far the best methodology we've ever devised for pushing toward the truth of things.
Science isn't a perfect process. There are still human beings involved. But it is by far the best methodology we've ever devised for pushing toward the truth of things.
Post has attachment
Combatting the Post-Factual Era
Christopher Penn shares two great dangers that feed our human tendency to privilege belief over rational analysis:
1. Innumeracy: a general lack of understanding of how to absorb and analyze data
2. Incuriosity: a general lack of curiosity and healthy skepticism on the part of both information providers and consumers.
He ends with a call for those who see this crisis and understand its implications, whether its for our businesses, our industries, or our world, to be passionately and personally involved in trying to reverse the fall into a post-factual era.
Christopher Penn shares two great dangers that feed our human tendency to privilege belief over rational analysis:
1. Innumeracy: a general lack of understanding of how to absorb and analyze data
2. Incuriosity: a general lack of curiosity and healthy skepticism on the part of both information providers and consumers.
He ends with a call for those who see this crisis and understand its implications, whether its for our businesses, our industries, or our world, to be passionately and personally involved in trying to reverse the fall into a post-factual era.
Wait while more posts are being loaded