October 18, 2016
October 12, 2016
October 8, 2016
October 6, 2016
October 5, 2016
October 4, 2016
October 1, 2016
September 22, 2016
September 17, 2016
Institutionalized | Hapgood (hapgood.us)
Good description of the power of institutions, which I was introduced to via Latour's Missing Mass paper. For open education we need to dismantle the institutions around the traditional textbook
September 16, 2016
It’s Bayes All The Way Up | Slate Star Codex (slatestarcodex.com)
Nice description of two studies of the kind of stuff in Surfing Uncertainty with a psychiatry angle. I want to rethink some of his qualms with a less exactly Bayesian approach to the top-down processing.
September 5, 2016
Human and Artificial Intelligence May Be Equally Impossible to Understand (nautil.us)
Seems obvious to me that we have some fundamental steps between current ML methods and real humanlike intelligence, so imposing interpretability on current models doesn't sound productive. It could be that humanlike AI is an iterative process between deep learning perception and machine-interpretability, but we have little clue what the latter looks like.
The pneumonia/asthma example is a good lesson about using current ML in practice. The ML model is a function of conditions -> long term outcome, but you're using it as conditions -> immediate actions. Yes, interpretability may help bridge that gap, but we have to know those aren't the same thing.
[no title] (hapgood.us)
Is human brain just adding social/argumentative aspects to a logical, planning animal brain?
September 4, 2016
September 3, 2016
What I learned as a hired consultant to autodidact physicists | Aeon Ideas (aeon.co)
"I still get the occasional joke from colleagues about my ‘crackpot consultant business’, but I’ve stopped thinking of our clients that way. They are driven by the same desire to understand nature and make a contribution to science as we are. They just weren’t lucky enough to get the required education early in life, and now they have a hard time figuring out where to even begin."
[no title] (hapgood.us)
Call for tech that supports professional updates. More than the fact that *Facebook* fails here is that professional publications do. in cs, Alan Kay says we have a pop science, which perhaps tarnishes the spread of the most important information. (I find this hard to judge as CS is so broad.) Maybe as a general trend other fields are too pop and/or too broad?
August 24, 2016
Why Do Intervention Effects Fade? - Daniel Willingham (www.danielwillingham.com)
Study suggests that intervention effects fade due to pre existing differences that still take affect after the intervention. Similar to Scott Alexander: genetics explain *more* variation at later ages (but regardless of interventions?). Willingham is more hopeful about education: we need to keep up the extra effort with these students (reminds me of article on good versus great CF treatment)
August 22, 2016
August 21, 2016
August 16, 2016
Tolerance Troubles | Slate Star Codex (slatestarcodex.com)
Medicine is super complicated: depending on the person and whether they've developed tolerance, medicine can take a full range of effects (from intended effect to no effect to opposite of intended effect)
August 15, 2016
"[If designing the Dynabook] Today, we would emphasize not just learning to think well in a complex world of many kinds of evidence, cultures and contexts, but being trained to think well under many kinds of stress, including those of time, scale, opinions, and almost invisible desires that are genetically generated and affect conscious decision making."
August 14, 2016
Making Algorithms Accountable - ProPublica (www.propublica.org)
"At ProPublica, we obtained more than 7,000 risk scores assigned by the company Northpointe, whose tool is used in Wisconsin, and compared predicted recidivism to actual recidivism. We found the scores were wrong 40 percent of the time and were biased against black defendants, who were falsely labeled future criminals at almost twice the rate of white defendants."
"In each class I saw the same basic behaviour – an assumption, implicit in the teaching, that the students either could not, or would not think. Under such conditions it was unreasonable to expect that students were going to spontaneously engage in problem solving enough to get stuck, and then persist through being stuck enough to have an AHA! experience."
Aiming Higher: Bloom and Vygotsky In the Classroom - YouTube (www.youtube.com)
Connects Bloom's (revised) taxonomy with Vygotsky's ZPDs: you don't have to move from bottom to top of the taxonomy, but rather activities with higher-order skills can build up the lower levels *when* the activities are within ZPD.
The Bell Curve - The New Yorker (www.newyorker.com)
The difference between good and great hospital for cystic fibrosis care: knowing the difference between 99.5% and 99.95% survival rates and using ad hoc measures to get patients doing the right things.
"The buzzword for clinicians these days is “evidence-based practice”—good doctors are supposed to follow research findings rather than their own intuition or ad-hoc experimentation. Yet Warwick is almost contemptuous of established findings. National clinical guidelines for care are, he says, “a record of the past, and little more—they should have an expiration date.”"
August 13, 2016
August 8, 2016
August 5, 2016
July 31, 2016
[no title] (computinged.wordpress.com)
C++ lambdas are much harder than C++ iterators for beginners
July 30, 2016
Post-Partisanship Is Hyper-Partisanship | Slate Star Codex (slatestarcodex.com)
"I predict (50% probability) that the progressives most carefully bubbled and separated from any actual threat from Republicans – which disproportionately includes politicians, journalists, and other opinion-makers – will start treating the Trump-voting classes more like Tibetans. I predict when they talk about specific bad Republicans like Trump, they’ll focus more on the ways they are funny and cartoonish (far too easy with Trump, but maybe the next guy will be a better test) instead of the ways they’re threatening. I predict that conflicts within the progressive movement will be increasingly vicious and increasingly likely to use poor whites as a political football (“the other side is bigoted against poor whites!”). I predict this will happen much more if the Democrats win the election than if they lose it; it’s always easier to be gracious toward a vanquished opponent."
July 28, 2016
July 27, 2016
Why We're Different | Edge.org (www.edge.org)
In the example of test scores, twin studies show that 60% of variance in hereditary. Polygenic models that account for variations in DNA can now explain 10% of variance. Still early stages -- these are not yet based on complete DNA sequencing.
"DNA is the best predictive game in town because it's causal. By being able to predict, we can begin to intervene."
Physical analog models (www.johndcook.com)
"Hydraulic engineering gets into some of the most complicated math there is. Allegedly when Albert Einstein’s son Hans said he wanted to study how sediment moves underwater, Einstein asked him why he wanted to work on something so complicated.
The physics involved happen on such a small scale that we still haven’t built equations complex enough to capture them. And so Stanford Gibson, a world-class numerical modeler, is one of the most ardent supporters of physical models."
"When you build a small thing to study how a big thing works, you’ve got to have some theory relating the behavior of the two. If the real thing is 1000 times bigger than the model, does that mean you can simply multiply measurements from the model by 1000 to predict measurements of reality? Sometimes. Some effects scale linearly, but some do not."
July 26, 2016
How The West Was Won | Slate Star Codex (slatestarcodex.com)
There's no dominating western culture, but a universal culture that most of the world is moving toward. I think we could find a powerful definition of universal culture and I don't think it's "better" (nor do I think details of outcomes are inevitable). It may be roughly "least common denominator" -- a mix of better and easier.
July 25, 2016
[no title] (computinged.wordpress.com)
U. Illinois has the most innovative program I have heard of for meeting these new needs. They are creating a range of CS+X degree programs. First, these CS+X programs are significant parts of the “X” departments
Is Listening to an Audio book "Cheating?" - Daniel Willingham (www.danielwillingham.com)
"An influential model of reading is the simple view (Gough & Tumner, 1986), which claims that two fundamental processes contribute to reading: decoding and language processing. “Decoding” obviously refers to figuring out words from print. “Language processing” refers to the same mental processes you use for oral language. Reading, as an evolutionary late-comer, must piggy-back on mental processes that already existed, and spoken communication does much of the lending. "
July 23, 2016
Stuff The Internet Says On Scalability For July 22nd, 2016 - High Scalability - (highscalability.com)
plusepsilon: I transitioned from using Bayesian models in academia to using machine learning models in industry. One of the core differences in the two paradigms is the "feel" when constructing models. For a Bayesian model, you feel like you're constructing the model from first principles. You set your conditional probabilities and priors and see if it fits the data. I'm sure probabilistic programming languages facilitated that feeling. For machine learning models, it feels like you're starting from the loss function and working back to get the best configuration
July 22, 2016
Things Probably Matter | Slate Star Codex (slatestarcodex.com)
"So it looks like happiness can change. It just didn’t change in China over the past thirty years. The apparent paradox of improving economic situation and stable/decreasing happiness is genuinely paradoxical. Intangibles are probably just way more important than money, even amounts of money big enough to raise whole countries out of poverty."
July 16, 2016
[no title] (www.sciencedirect.com)
Scientists now able to reconstruct moving visual scenes from fMRI
[no title] (computinged.wordpress.com)
Two predictions from learning science that don't hold up in the CS context. In CS: giving students subgoals is better than students forming their own, mastery goals are better than performance goals
July 15, 2016
July 13, 2016
July 12, 2016
Frame of Reference - Invisibilia (pca.st)
With transcranial magnetics, woman with autism is able to immediately experience emotional subtleties of human interactions. What are the implications for learning new ways of thinking beyond knowledge acquisition/automaticity/conceptual change? Similar to representation change or beyond?
July 11, 2016
July 5, 2016
Inside the Mirrortocracy (carlos.bueno.org)
"The first step toward dissolving these petty Cultures is writing down their unwritten rules for all to see. The word "privilege" literally means "private law". It's the secrecy, deniable and immune to analysis, that makes the balance of power so lopsided in favor of insiders."
Cargo Cults and Interview Questions (braythwayt.com)
story about cargo cult in interview questions. being able to speak intelligently about *some* detail is much more important than being able to correctly answer about one particular detail.
July 4, 2016
June 22, 2016
[no title] (mathwithbaddrawings.com)
June 21, 2016
What Bill Gates is afraid of - YouTube (www.youtube.com)
Bill Gates: biggest impact and highest probability for major human death is an epidemic that spreads fast and kills. Ebola showed we aren't ready if this combination happens (last time: Spanish flu, but transportation was 50x less).
June 12, 2016