Carnegie Mellon Computer Searches Web 24/7 To Analyze Images and Teach Itself Common Sense


A computer program called the Never Ending Image Learner (NEIL) is running 24 hours a day at Carnegie Mellon University, searching the Web for images, doing its best to understand them on its own and, as it builds a growing visual database, gathering common sense on a massive scale.



How sensors will change our lives from smoother traffic to smarter garbage


The Optimod’Lyon project in Lyon, east-central France, comprises of 450 sensors spread all over the city to measure the density of traffic. The data they transmit allows the city to predict traffic for the next hour with a reliability rate of over 90%. The idea is that in the future traffic light management and information for public transport users can be adapted accordingly.


Top 25 Cities in the World

Happy Friday

Yeast, human stem cells drive discovery of new Parkinson’s disease drug targets
Using a discovery platform whose components range from yeast cells to human stem cells, Whitehead Institute scientists have identified a novel Parkinson’s disease drug target and a compound capable of repairing neurons derived from Parkinson’s patients.
The platform—whose effectiveness is described in dual papers published online this week in the journal Science—could accelerate the discovery of drug candidates that address the underlying pathology of Parkinson’s and other neurodegenerative diseases. Today, no such drugs exist.
Parkinson’s disease (PD) and such neurodegenerative diseases as Huntington’s and Alzheimer’s are characterized by protein misfolding, resulting in toxic accumulations of proteins in the cells of the central nervous system. Cellular buildup of the protein alpha-synuclein, for example, has long been associated with PD, making this protein a seemingly appropriate target for therapeutic intervention.
In the search for compounds that might alter a protein’s behavior or function—such as that of alpha-synuclein—drug companies often rely on so-called target-based screens that test the effect large numbers of compounds have on the protein in question in rapid, automated fashion. Though efficient, such an approach is limited by the fact that it essentially occurs in a test tube. Seemingly promising compounds emerging from a target-based screen may act quite differently when they’re moved from the in vitro environment into a living setting.
To overcome this limitation, the lab of Whitehead Member Susan Lindquist has turned to phenotypic screens in which candidate compounds are studied within a living system. In Lindquist’s lab, yeast cells—which share the core cell biology of human cells —serve as living test tubes in which to study the problem of protein misfolding and to identify possible solutions. Yeast cells genetically modified to overproduce alpha-synuclein serve as robust models for the toxicity of this protein that underlies PD.
“Phenotypic screens are probably underutilized for identifying drug targets and potential compounds,” says Daniel Tardiff, a scientist in the Lindquist lab and lead author of one of the Science papers. “Here, we let the yeast tell us what is a good target. We let a living cell tell us what’s critical for reversing alpha-synuclein toxicity.”
In a screen of nearly 200,000 compounds, Tardiff and collaborators identified one chemical entity that not only reversed alpha-synuclein toxicity in yeast cells, but also partially rescued neurons in the model nematode C. elegans and in rat neurons. Significantly, cellular pathologies including impaired cellular trafficking and an increase in oxidative stress, were reduced by treatment with the identified compound. Enabled by the chemistry provided by Nate Jui in the Buchwald lab at MIT, Tardiff found that the compound was working by restoring functions mediated by a cellular protein critical for trafficking that was previously thought to be “undruggable.”
But would these findings apply in human cells? To answer that question, husband-and-wife team Chee-Yeun Chung and Vikram Khurana led the second study published in Science to examine neurons derived from induced pluripotent stem (iPS) cells generated from Parkinson’s patients. The cells and differentiated neurons (of a type damaged by the disease) were derived from patients that carried alpha-synuclein mutations and develop aggressive forms of the disease. To ensure that any pathology developed in the cultured neurons could be attributed solely to the genetic defect, the researchers also derived control neurons from iPS cells in which the mutation had been corrected.
Chung and Khurana used the wealth of data from the yeast alpha-synuclein toxicity model to clue them in on key cellular processes that became perturbed as patient neurons aged in the dish. Strikingly, exposure to the compound identified via yeast screens in Tardiff’s study reversed the damage in these neurons.
“It was remarkable that the compound rescued yeast cells and patient neurons in similar ways and through the same target—a target we would not have identified without yeast genetics to guide us,” says Khurana, a postdoctoral scientist in the Lindquist lab and a neurologist at Massachusetts General Hospital who recruited patients for participation in this research. Khurana believes that the abnormalities discovered occur in the early stages of disease. If so, successful manipulation of the targets identified here might help slow or even prevent disease progression.
For the researchers involved, these findings are a bit of surprise. Because neurodegenerative disorders like PD are largely diseases of aging, modeling them in a culture dish using neurons grown from iPS cells has been thought to be exceedingly difficult, if not impossible.
“Many, ourselves included, were skeptical that we could find any important pathologies for a neurodegenerative disorder by reprogramming patient cells,” says Chung, a Senior Research Scientist in the Lindquist lab. “Critically, we also validated these pathologies in post-mortem brains, so we’re quite confident these are relevant for the disease.”
Next steps for these scientists include chemically optimizing the compound identified and testing it in animal models. Moreover, they are convinced that this yeast-human stem cell discovery platform could be applied to other neurodegenerative diseases for which yeast models have been developed.
“Using yeast genetics to identify a compound and its mechanism of action against the fundamental pathology of a disease illustrates the power of the system we’ve built,” says Lindquist, who is also professor of biology at MIT and a Howard Hughes Medical Institute investigator. “It’s critical that we continue to leverage this power because as we reduce the rate at which people are dying from cancer and heart disease, the burden of these dreaded neurodegenerative diseases is going to rise. It’s inevitable.”

This Is How Your Brain Becomes Addicted to Caffeine


This Is How Your Brain Becomes Addicted to Caffeine

Within 24 hours of quitting the drug, your withdrawal symptoms begin. Initially, they’re subtle: The first thing you notice is that you feel mentally foggy, and lack alertness. Your muscles are fatigued, even when you haven’t done anything strenuous, and you suspect that you’re more irritable than usual.

Over time, an unmistakable throbbing headache sets in, making it difficult to concentrate on anything. Eventually, as your body protests having the drug taken away, you might even feel dull muscle pains, nausea and other flu-like symptoms.

This isn’t heroin, tobacco or even alcohol withdrawl. We’re talking about quitting caffeine, a substance consumed so widely (the FDA reports thatmore than 80 percent of American adults drink it daily) and in such mundane settings (say, at an office meeting or in your car) that we often forget it’s a drug—and by far the world’s most popular psychoactive one.

Like many drugs, caffeine is chemically addictive, a fact that scientists established back in 1994. This past May, with the publication of the 5th edition of the Diagnostic and Statistical Manual of Mental Disorders (DSM), caffeine withdrawal was finally included as a mental disorder for the first time—even though its merits for inclusion are symptoms that regular coffee-drinkers have long known well from the times they’ve gone off it for a day or more.

Why, exactly, is caffeine addictive? The reason stems from the way the drug affects the human brain, producing the alert feeling that caffeine drinkers crave.

Soon after you drink (or eat) something containing caffeine, it’s absorbed through the small intestine and dissolved into the bloodstream. Because the chemical is both water- and fat-soluble (meaning that it can dissolve in water-based solutions—think blood—as well as fat-based substances, such as our cell membranes), it’s able to penetrate the blood-brain barrier and enter the brain.

Structurally, caffeine closely resembles a molecule that’s naturally present in our brain, called adenosine (which is a byproduct of many cellular processes, including cellular respiration)—so much so, in fact, that caffeine can fit neatly into our brain cells’ receptors for adenosine, effectively blocking them off. Normally, the adenosine produced over time locks into these receptors and produces a feeling of tiredness.

When caffeine molecules are blocking those receptors, they prevent this from occurring, thereby generating a sense of alertness and energy for a few hours. Additionally, some of the brain’s own natural stimulants (such as dopamine) work more effectively when the adenosine receptors are blocked, and all the surplus adenosine floating around in the brain cues the adrenal glands to secrete adrenaline, another stimulant.

For this reason, caffeine isn’t technically a stimulant on its own, says Stephen R. Braun, the author or Buzzed: the Science and Lore of Caffeine and Alcohol, but a stimulant enabler: a substance that lets our natural stimulants run wild. Ingesting caffeine, he writes, is akin to “putting a block of wood under one of the brain’s primary brake pedals.” This block stays in place for anywhere from four to six hours, depending on the person’s age, size and other factors, until the caffeine is eventually metabolized by the body.

In people who take advantage of this process on a daily basis (i.e. coffee/tea, soda or energy drink addicts), the brain’s chemistry and physical characteristics actually change over time as a result. The most notable change is that brain cells grow more adenosine receptors, which is the brain’s attempt to maintain equilibrium in the face of a constant onslaught of caffeine, with its adenosine receptors so regularly plugged (studies indicate that the brain also responds by decreasing the number of receptors for norepinephrine, a stimulant). This explains why regular coffee drinkers build up a tolerance over time—because you have more adenosine receptors, it takes more caffeine to block a significant proportion of them and achieve the desired effect.

This also explains why suddenly giving up caffeine entirely can trigger a range of withdrawal effects. The underlying chemistry is complex and not fully understood, but the principle is that your brain is used to operating in one set of conditions (with an artificially-inflated number of adenosine receptors, and a decreased number of norepinephrine receptors) that depend upon regular ingestion of caffeine. Suddenly, without the drug, the altered brain chemistry causes all sorts of problems, including the dreaded caffeine withdrawal headache.

The good news is that, compared to many drug addictions, the effects are relatively short-term. To kick the thing, you only need to get through about 7-12 days of symptoms without drinking any caffeine. During that period, your brain will naturally decrease the number of adenosine receptors on each cell, responding to the sudden lack of caffeine ingestion. If you can make it that long without a cup of joe or a spot of tea, the levels of adenosine receptors in your brain reset to their baseline levels, and your addiction will be broken.

Neuroscience: How parents see themselves may affect their child's brain and stress level


Self-perceived social status predicts hippocampal function and stress hormones

A mother’s perceived social status predicts her child’s brain development and stress indicators, finds a study at Boston Children’s Hospital. While previous studies going back to the 1950s have linked objective…

Coursera Blog: 30+ Coursera classes and a Google Internship


Feynman Liang is one of Coursera’s ‘top 50’ students by number of courses he has completed. To date, Feynman has taken 34 courses! We have asked him to share tips for the Coursera community on how to tackle taking multiple courses on Coursera

At first, I was skeptical about taking classes…

Peter Thiel’s CS183: Startup - Class 1 Notes Essay


Here is an essay version of my class notes from Class 1 of CS183: Startup. Errors and omissions are my own. Credit for good stuff is Peter’s entirely. 

CS183: Startup—Notes Essay—The Challenge of the Future

Purpose and Preamble

            We might describe our world as having retail sanity, but wholesale madness. Details are well understood; the big picture remains unclear. A fundamental challenge—in business as in life—is to integrate the micro and macro such that all things make sense.

            Humanities majors may well learn a great deal about the world. But they don’t really learn career skills through their studies. Engineering majors, conversely, learn in great technical detail. But they might not learn why, how, or where they should apply their skills in the workforce. The best students, workers, and thinkers will integrate these questions into a cohesive narrative. This course aims to facilitate that process.

I.          The History of Technology

            For most of recent human history—from the invention of the steam engine in the late 17th century through about the late 1960’s or so— technological progress has been tremendous, perhaps even relentless. In most prior human societies, people made money by taking it from others. The industrial revolution wrought a paradigm shift in which people make money through trade, not plunder.

              The importance of this shift is hard to overstate. Perhaps 100 billion people have ever lived on earth. Most of them lived in essentially stagnant societies; success involved claiming value, not creating it. So the massive technological acceleration of the past few hundred years is truly incredible.

             The zenith of optimism about the future of technology might have been the 1960’s. People believed in the future. They thought about the future. Many were supremely confident that the next 50 years would be a half-century of unprecedented technological progress.

              But with the exception of the computer industry, it wasn’t. Per capita incomes are still rising, but that rate is starkly decelerating. Median wages have been stagnant since 1973. People find themselves in an alarming Alice-in-Wonderland-style scenario in which they must run harder and harder—that is, work longer hours—just to stay in the same place. This deceleration is complex, and wage data alone don’t explain it. But they do support the general sense that the rapid progress of the last 200 years is slowing all too quickly. 

Read More

DERiVE: 【New York University】Deep Learningにより学習した画像認識器を試せるデモサイトが公開


以下のサイトに、「Image Classifier Demo 」と銘打った、Deep Learning ベースの画像認識器のデモアプリが公開されているので紹介します。

Image Classifier Demo

このサイトはNYUのMattherw Zeiler氏により製作されたデモサイトです(※ NYUでDeep Learningと言うとYann Lecun先生を思い出すかもしれませんが、彼はRob Fergus研の方です)。彼はこのICCV2011の論文あたりで著名と言えます。


217 plays

Indre Viskontas Daniel Dennett - Tools for Thinking

How Google's Image Recognition Works


Just like Google Drive, Google+ Photos uses some amazing image recognition technology to make photos searchable, even if they don’t have captions or useful filenames. “This is powered by computer vision and machine learning technology, which uses the visual content of an image to generate searchable tags for photos combined with other sources like text tags and EXIF metadata to enable search across thousands of concepts like a flower, food, car, jet ski, or turtle,” explains Google.

Google acquired DNNresearch, a start-up created by Professor Geoffrey Hinton and two of his graduate students at the University of Toronto. They built “a system which used deep learning and convolutional neural networks and easily beat out more traditional approaches in the ImageNet computer vision competition designed to test image understanding.” Google built and trained similar large-scale models and found that this approach doubles the average precision, compared to other object recognition methods. “We took cutting edge research straight out of an academic research lab and launched it, in just a little over six months,” says Chuck Rosenberg, from the Google Image Search Team.

If the government assigned you a grocery card that tracked all your food purchases, it would seem like something out of a dystopian everybody-in-a-jumpsuit movie. But when your grocery store does it in return for a dollar off a bottle of salad dressing, that’s an absolutely everyday occurrence for many of us.

Linda Holmes (via sashareads)