Friday, February 27, 2015

Why Debunked Autism Treatment Fads Persist

The communication struggles of children with autism spectrum disorder can drive parents and educators to try anything to understand their thoughts, needs and wants. Unfortunately, specialists in psychology and communication disorders do not always communicate the latest science so well. These factors make the autism community especially vulnerable to interventions and "therapies" that have been thoroughly discredited, says Scott Lilienfeld, a psychologist at Emory University.  "Hope is a great thing, I'm a strong believer in it," Lilienfeld says. "But the false hope buoyed by discredited therapies can be cruel, and it may prevent people from trying an intervention that actually could deliver benefits." Lilienfeld is lead author of a commentary, "The persistence of fad interventions in the face of negative scientific evidence: Facilitated communication for autism as a case example," recently published by the journal Evidence-Based Communication Assessment and Intervention. Co-authors of the commentary are Julia Marshall (also from Emory) and psychologists James Todd (from Eastern Michigan University), and Howard Shane (director of the Autism Language Program at Boston Children's Hospital). more

Thursday, February 26, 2015

Google Builds An AI That Can Learn And Master Video Games via Operant Conditioning

Google has built an artificial intelligence system that can learn – and become amazing at – video games all on its own, given no commands but a simple instruction to play titles. The project, detailed by Bloomberg, is the result of research from the London-based DeepMind AI startup Google acquired in a deal last year, and involves 49 games from the Atari 2600 that likely provided the first video game experience for many of those reading this. While this is an amazing announcement for so many reasons, the most impressive part might be that the AI not only matched wits with human players in most cases, but actually went above and beyond the best scores of expert meat-based players in 29 of the 49 games it learned ... Next up for the arcade AI is mastering the Doom-era 3D virtual worlds, which should help the AI edge closer to mastering similar tasks in the real world, like driving a car. And there’s one more detail here that may keep you up at night: Google trained the AI to get better at the Atari games it mastered using a virtual take on operant conditioning – ‘rewarding’ the computer for successful behavior the way you might a dog. more

Tuesday, February 17, 2015

Meet Rudy, The Dog Who Is Being Raised With Science

Rudy the staghound gets to enjoy a lot of the things that many pets do: trips to the beach, playtime at the dog park, napping on the couch. But Rudy's adventures and home life are examined a lot more closely than those of an ordinary pet, because he's being raised with science... [We] checked his preference for positive reinforcement. Research has shown that dogs tend to prefer food, then petting, then verbal praise and human company (see: Behavioural Processes and the Journal of the Experimental Analysis of Behavior). [It also shows] that dogs can learn good or bad behaviors from observing each other (see: Applied Animal Behaviour Science and Developmental Psychobiology), so there is great value in having an older 'model' dog to demonstrate good behaviors to a new or young dog. Rudy loves play with a tug toy and food as first choice reinforcers, and we're fortunate to have a very experienced friend with an exemplary Doberman dog named Kade with whom we catch up with regularly. Kade acts in part almost like a guiding chaperone for Rudy on lovely off leash walk adventures, helping us set Rudy up for successful situations (like returning when called) that we can reinforce positively. more

Friday, February 13, 2015

Dogs Can Tell Happy or Angry Human Faces

If you ever get the impression that your dog can "tell" whether you look content or annoyed, you may be onto something. Dogs may indeed be able to discriminate between happy and angry human faces, according to a new study. Researchers trained a group of 11 dogs to distinguish between images of the same person making either a happy or an angry face. During the training stage, each dog was shown only the upper half or the lower half of the person's face. The investigators then tested the pups' ability to discriminate between human facial expressions by showing them different images from the ones used in training. The dogs were shown either the other half of the face used in the training stage, the other halves of people's faces not used in training, a face that was the same half as the training face but from a different person, or the left half of the face used in the training stage. more


Thursday, February 12, 2015

Psychiatry Has Earned Its Stigma

Has there ever been a medical specialty as beleaguered as psychiatry? Since the profession’s founding in 1844, the doctors of the soul have had to contend with suspicions that they do not know what mental illness is, what type their patients might have, or what they should do about it—in other words, that they are doctors who do not practice real medicine... It’s not just diagnostic uncertainty or therapeutic disasters that cast suspicion on the profession. It’s also the bred-in-the-bone American conviction that no one should tell us who we are. For that is what psychiatrists (and the rest of us in the mental-health professions) do, no matter whether we want to or not. To say you know what mental health and illness are is to say you know how life should go, and what we should do when it goes otherwise. You’d better know what to do when you’ve made a grievous error in those weighty matters, or at the very least, how to ask for forgiveness. And you’d better hope that, apologies offered, you can give the public a reason to believe that at long last you know what you are doing. his is the unenviable task that Jeffrey Lieberman, past president of the APA, chairman of psychiatry at Columbia University’s medical school, chief of psychiatry at its hospital, and director of the New York State Psychiatric Institute, has taken on in his book Shrinks: The Untold Story of Psychiatry. “Psychiatry has earned its stigma,” he writes at the outset, and its practitioners must “own up to our long history of mistakes.” Otherwise it will remain “the black sheep of the medical family, scorned by physicians and patients alike.” more

Wednesday, February 11, 2015

Why In-App Purchases Will Bankrupt You

App purchases, on the other hand, are designed to be incredibly seamless. Your credit card is stored with Apple, Google or Amazon, and all it takes is a quick click of a button to instant gratification, no uncomfortable reminders that you just spent real money on that virtual life or game boost. What’s more, games use subtle-but-effective psychological tricks to keep you playing and spending. The Guardian’s Dana Smith writes that “Candy Crush” is exceedingly popular because it uses a series of ingenious mind-games to make it addictive. First, it’s simple. Games start off easy to play with frequent mini rewards that cause a release of “neurochemical dopamine.” This pleasure response in your brain is similar to the “neuro-circuitry involved in addiction, reinforcing our actions.” The next step is to increase the challenge, making the dopamine bursts more intermittent. If it stayed easy forever, you would get bored. And, as it turns out, an unpredictable reward schedule (a “variable ratio schedule of reinforcement”) is more addictive to our brain’s pleasure sensors than one where you keep consistently winning. more

Monday, February 09, 2015

Speak & Spell: A History

The Speak & Spell is a teaching machine specifically in the tradition of B. F. Skinner, reflecting some of both Skinner’s design principles and his theories of learning, decades older than the popular Texas Instruments device. Rather than selecting the correctly-spelled word in a multiple choice quiz, for the example, the Speak & Spell prompts the user to construct the response. It praises; it corrects... The shared features in most definitions of teaching machines include automation, immediate feedback, and self-pacing. The Speak & Spell has all three, using “contingencies of reinforcement” to establish appropriate spelling behavior. (Some of its engineers thought it would be funny if the user received a raspberry or a funny comment when they spelled a word wrong. But this idea was rejected as it would “reward” incorrect spelling.) more

Thursday, February 05, 2015

The Cocktail Party Effect: Birds Hear Like We Do

I worked in Dr. Micheal L. Dent’s laboratory for a few years as I was earning my undergraduate degree in animal behavior. I was interested in studying birds, and Dr. Dent was interested in studying acoustic communication in animals. Dr. Dent is an associate professor in the Psychology Department at the University at Buffalo, SUNY. “In the past, I’ve worked with budgerigars, starlings, canaries (multiple strains), Japanese quail, zebra finches, and barn owls,” she recently told me. “I currently work with budgerigars and zebra finches.” Acoustic communication refers to hearing. Other ways that animals can communicate include visual, olfaction, touch, thermal, and some can communicate even through electromagnetic fields. In Dr. Dent’s lab we used operant conditioning and positive reinforcement to measure birds’ abilities to detect, discriminate, identify, and localize sounds by training them to peck keys. One interesting occurrence she discovered was that parakeets and zebra finches exhibit the cocktail party effect. more

Thursday, January 29, 2015

Using Smartphones and Apps to Enhance Loyalty Programs

Capriotti’s, a 106-store chain of sandwich shops in 16 states, expects to introduce an app-based loyalty program early this year that its chief marketing officer, Jason Smylie, says will enable shop owners to enrich and fine-tune a prior punch card rewards program...The software, developed by the company Punchh, will enable Capriotti’s to award a free drink or a dessert — as an unexpected reward at the cash register — to highly valued customers on perhaps 20 percent of their visits. “You’re not only rewarding the customers who are coming more frequently, you’re also giving people an incentive to show up,” he said. “I can come in and potentially get something for free. That’s awesome.” And effective. Psychologists have a name for this kind of reward — random intermittent reinforcement — and know it as a powerful way to encourage repeat behavior. Think no further than slot machines. Casinos have zeroed in on the gambling habits of their patrons through the use of smart cards rather than coins. Retailers can also now better know their customers through loyalty apps, which may also use data from Facebook profiles. “With apps you now can target specific customers and influence specific behaviors and keep track of all the results and understand the results,” Mr. Smylie said. more

Wednesday, January 28, 2015

Intracranial Stimulation Recovers Learning and Memory in Rats

The research, published in Behavioural Brain Research, was conducted by Pilar Segura and Ignacio Morgado (coordinators), Laura Aldavert and Marc Ramoneda, psychobiologists of the Institute of Neurosciences and the Department of Psychobiology and Health Sciences Methodology of the UAB and by Elisabet Kadar and Gemma Huguet, molecular biologists of the University of Girona, to explore the power of Deep Brain Stimulation treatments in the hypothalamus to recover the ability to learn and remember after a severe lesion of the amygdala...The hypothalamus is a region of the brain in which the most basic impulses are found, helping us to survive and providing us with pleasure. It is part of the brain's reward system, which makes us feel good after carrying out an activity and helps us change our behaviour through positive reinforcement. more

Tuesday, January 27, 2015

What Standardized Testing and Schools Could Learn from Target

The dream of automated learning is even older than computers themselves. In 1924 an educational psychology professor named Sidney Pressey built a mechanical teaching machine that supplied questions, along with the correct answers, when a button was pressed. B.F. Skinner, the famous behaviorist, also introduced a teaching machine in 1954, a clunky thing that looked like a typewriter. He claimed some of the very same benefits for it that you hear from ed-tech entrepreneurs to this day—allowing every student to move at his own pace, supplying immediate feedback, improving motivation...Today, besides the Knewton-powered products, other adaptive software is sold by Dreambox, Scholastic, and Khan Academy, a nonprofit. Most work more or less the same way. They introduce concepts with text, video or animations, ask students to respond in the form of quick short-answer or multiple-choice questions, give them increasingly broad hints when they get stuck, and choose what concept to bring in next based on the student’s responses. What the “adaptive” part means is that the specific selection of questions and order of content presented to each student will vary according to the students’ responses. You take a quick diagnostic test or start off with a medium-hard question. If you get it right, you proceed to harder questions; flub it and you get easier questions. As a result, each student’s path and pace through the material is slightly different...[These] platform[s] and those like it could reduce the need for high-stakes tests. “Knewton allows for much more gentle, passive data collection via ongoing formative assessment in homework/classwork,” he told me in an online chat. “We can predict your score on a bunch of these high stakes tests anyway, lessening the need for so many of them.” “Formative assessment” refers to the feedback that is part of nearly any teaching and learning scenario, such as when a teacher calls on the class during the lecture, or when a student is studying vocabulary with flash cards and flips the card over to see the right answer. It’s opposed to summative assessment, which “sums up” learning at the end of a period of time, ranging from a unit test to a graduation exam. more

Thursday, January 22, 2015

We Lie About What We Eat, And It's Messing Up Science

How many peanuts did you snack on last week? If you don't remember, you're not alone. We humans are notoriously bad at remembering exactly what and how much we ate. And if there's one pattern to our errors, it's that we underestimate — unintentionally and otherwise. And yet, for decades, researchers who want to amass large quantities of data about how much Americans eat and exercise have had to rely on individuals to self-report this information. These self-reported data on diet and exercise have long been called flawed. But a paper published Tuesday in the International Journal of Obesity goes one step further. The authors, led by David Allison, director of the Nutrition Obesity Research Center at the University of Alabama at Birmingham, write, "[The data] are so poor as measures of actual [energy intake] and [physical activity energy expenditure] that they no longer have a justifiable place in scientific research." more

Wednesday, January 21, 2015

The Internet of Me

The so-called “internet of things” has been the talk of the technology world for years now. Consumer electronics firms are exploring ways of connecting all manner of personal and household objects to the internet, thereby extending their functionality...Despite the rush to connect all our household objects to the internet, there is an even more important technical trend which has appeared in the past year or so – the internet of me. Instead of just wanting to hook up our home appliances to the internet, we have started to take our bodies and brains online.In recent times we have seen more and more apps applying the logic of behaviourism. Based on the heavily criticised methods once developed by B F Skinner, these apps alternate rewards and punishment to reinforce positive behaviour. One such example is the GymPact. Users of this app specify how many times a week they intend to go to the gym. In the event of them missing a session, which is easily found out through the GPS, they need to pay $5. This money is then shared equally among those who have followed their individually set goals. A more extreme version of this is Pavlok. To break bad habits, this app sends out electric shocks. Or as the slogan goes: “the habit changing device that shocks you”. If you bite your nails, oversleep, or procrastinate you will be handed out a punishment. The app is marketed as a personal coach on your wrist. These apps do not just shape our behaviour so that we will become more productive people. They also promise to make us healthier and happier. more

Tuesday, January 20, 2015

Measurement: Friend or Foe?

It seems that the world is becoming obsessed with measurement. We are measuring steps, heart rate, calories, sleep—you name it and there is a device to measure it. People are going to waste a lot of time and money on measurement because, for many, it won’t change a thing. It seems that everyone believes the old adage “What gets measured gets done.” Of course that is wrong. Just because my Fitbit measures my steps each day does not cause me to walk 10,000 steps. Just because I weigh myself every day does not mean that I have met my weight goal. As a matter of fact, I have lost almost no weight even though I have had a Fitbit scale for over a year. This may sound strange coming from someone who has helped managers and executives focus time and effort on measuring all sorts of variables in business. However, it is true, and there is a very clear reason as to why. I believe a more appropriate adage is, “What gets measured improves the chances of getting done.” more

Friday, January 16, 2015

Trying to Cure Depression, But Inspiring Torture

To understand the nature of learned helplessness, one needs to travel back to Seligman’s early graduate-school days in the laboratory of Richard Solomon at the University of Pennsylvania. When Seligman began his studies, Solomon’s lab was working with dogs on a phenomenon that Ivan Pavlov had first identified as aversive conditioning or avoidance learning. The researchers administered shocks to the animals, accompanied by tones or lights, so that they would come to associate the tone or light stimuli with the shock’s onset, and, in some cases, then learn to avoid the shock by jumping over a barrier. Solomon would then work to see if he could get the dogs to, in effect, unlearn the association. When Seligman arrived at the lab, he noticed that some of the dogs had started to act rather strangely. Instead of trying to figure out how to avoid a new shock, they just sat there. They didn’t even try to figure it out. Teaming up with fellow graduate student Steven Maier, Seligman began to study what was going on...When Seligman and Maier analyzed the results, they found a consistent pattern. The dogs that had learned to avoid the shocks by pressing their heads against the panels on the first day were quick to jump the barrier on day two. Not a single dog failed to learn to jump quickly after the first go-around. Those that had been unable to escape the shocks, though, weren’t even trying...The effect of the harness experiment was been both severe and lasting. more

Wednesday, January 14, 2015

The "Next Big Things" in Teaching Technology Never Quite Were

While attending a back-to-school event at his youngest daughter Debbie's school in 1953, influential psychologist B.F. Skinner watched as her teacher taught fourth-grade arithmetic. After writing the problem on the blackboard, the teacher would walk up and down the aisle, occasionally pointing out the children's mistakes. Some students finished quickly and sat bored while others continued to work the problems. The teacher collected the papers, graded them and returned them to the students the following day. This immediately gave Skinner insight into some problems in the pedagogy, as well as an idea toward their solution. Skinner knew that a corrected paper seen 24 hours later could not serve as a reinforcer and did not present a good scenario for learning. Understanding the value of using mechanical devices in his experiments with pigeons, he created a crude prototype over the next few days, using a series of cards containing questions, within a box with sliders to "dial in" the answers. It was his first teaching machine. more

Monday, January 12, 2015

Rhesus Monkeys Can Learn to See Themselves in the Mirror

For humans and apes, vanity comes naturally -- homo sapiens and their hairier ancestors both automatically recognize themselves in the mirror. The same can't be said for monkeys. But new research suggests rhesus monkeys can be taught mirror recognition. Scientists say their discovery is likely to help them further understand the neural origins of self-awareness... Scientists have tried to determine whether monkeys could develop mirror recognition before, but nothing seemed to work. In the new study, researchers shined a laser on the monkey's faces -- a mild irritant intended to spur the specimens into using the mirror to their advantage. After several weeks of training the monkeys began recognizing their face and laser in the mirror, moving their hands to the spot of the laser and then often smelling their fingers -- suggesting a recognition of themselves in the mirror. Once the behavior was learned, many of the monkeys were able to use the mirror in a variety of other unprompted ways in order to investigate other parts of their bodies. more

Friday, January 09, 2015

Talk to Your Kids

In the nineteen-eighties, two child psychologists at the University of Kansas, Betty Hart and Todd Risley, began comparing, in detail, how parents of different social classes talked with their children. Hart and Risley had both worked in preschool programs designed to boost the language skills of low-income kids, but they had been dissatisfied with the results of such efforts: the achievement gap between rich and poor had continued to widen... In all, Hart and Risley reported, they analyzed “more than 1,300 hours of casual interactions between parents and their language-learning children.” The researchers noticed many similarities among the families: “They all disciplined their children and taught them good manners and how to dress and toilet themselves.” They all showed their children affection and said things like “Don’t jump on the couch” and “Use your spoon” and “Do you have to go potty?” But the researchers also found that the wealthier parents consistently talked more with their kids. Among the professional families, the average number of words that children heard in an hour was twenty-one hundred and fifty; among the working-class families, it was twelve hundred and fifty; among the welfare families, it was six hundred and twenty. Over time, these daily differences had major consequences, Hart and Risley concluded: “With few exceptions, the more parents talked to their children, the faster the children’s vocabularies were growing and the higher the children’s I.Q. test scores at age 3 and later.” more

Tuesday, January 06, 2015

What Heroin Addiction Tells Us About Changing Bad Habits

Our environments come to unconsciously direct our behavior. Even behaviors that we don't want, like smoking. "For a smoker, the view of the entrance to their office building — which is a place that they go to smoke all the time — becomes a powerful mental cue to go and perform that behavior," Neal says. Over time those cues become so deeply ingrained that they are very hard to resist. And so we smoke at the entrance to work when we don't want to. We sit on the couch and eat ice cream when we don't need to, despite our best intentions, despite our resolutions. "We don't feel sort of pushed by the environment," Wood says. "But, in fact, we're very integrated with it." To battle bad behaviors then, one answer is to disrupt the environment in some way. Even small changes can help — like eating the ice cream with your nondominant hand. What this does is disrupt the learned body sequence that's driving the behavior, which allows your conscious mind to come back online and reassert control. "It's a brief sort of window of opportunity," Wood says, "to think, 'Is this really what I want to do?' " Of course, larger disruption can also be helpful, which brings us back to heroin addiction in Vietnam. more

Monday, January 05, 2015

Getting Hooked: How Digital Firms Create Products That Get in Your Head

The makers of habit-forming products have clearly read the works of B.F. Skinner, the father of “radical behaviourism”, who found that training subjects by rewarding them in a variable, unpredictable way works best. That is why the number of monsters one has to vanquish in order to reach the next level in a game often varies. Faithful Twitter users are rewarded with more replies to their tweets, and more ego-boosting followers, but not according to any predictable formula. These variable rewards come in three forms. The reward of the tribe: people who use Twitter or Pinterest are rewarded with social validation when their tweets are retweeted or their pictures are pinned. The reward of the hunt: users quickly scroll through their feeds in search of the latest gossip or funny cat pictures. And the reward of self-fulfilment: people are driven to achieve the next level on a video game, or an empty e-mail inbox. Should the makers of habit-forming products be praised as innovative entrepreneurs? Or shunned as the immoral equivalents of drug pushers? Ian Bogost, a designer of video games, describes them as nothing less than the “cigarette of this century." more

Thursday, December 18, 2014

Addictive Social Media

Plenty of research has demonstrated that the addictive quality of social media is very real. And according to a new study, heavy social media use may also contribute to a different type of addiction ... Psychologist Julia Hormes, who led the study, said that Facebook was found to have especially addictive properties. The respondents spent an average of one-third of their online browsing time on Facebook, and 67 percent received Facebook push notifications on their phones. "New notifications or the latest content on your newsfeed acts as a reward. Not being able to predict when new content is posted encourages us to check back frequently," Hormes said in a statement. "This uncertainty about when a new reward is available is known as a 'variable interval schedule of reinforcement' and is highly effective in establishing habitual behaviors that are resistant to extinction. Facebook is also making it easy for users to continuously be connected to its platform, for example by offering push notifications to mobile devices." more

Wednesday, December 17, 2014

In Praise of Small Miracles (and Small Nudges)

Most of us don’t save enough. When governments try to encourage saving, they usually enact big policies to increase the incentives. But, in Kenya, people were given a lockable metal box — a simple place to put their money. After one year, the people with metal boxes increased savings by so much that they had 66 percent more money available to pay for health emergencies. It would have taken a giant tax reform to produce a shift in behavior that large. Too many people die in auto accidents. When governments try to reduce highway deaths, they generally increase safety regulations. But, also in Kenya, stickers were placed inside buses and vans urging passengers to scream at automobile drivers they saw driving dangerously. The heckling discouraged dangerous driving by an awesome amount. Insurance claims involving injury or death fell to half of their previous levels. These are examples of a new kind of policy-making that is sweeping the world. The old style was based on the notion that human beings are rational actors who respond in straightforward ways to incentives. The new style, which supplements but does not replace the old style, is based on the obvious point that human beings are not always rational actors. Sometimes we’re mentally lazy, or stressed, or we’re influenced by social pressure and unconscious biases. It’s possible to take advantage of these features to enact change. more

Tuesday, December 16, 2014

Implementation of Contingency Management in Probation Agencies

Contingency management (CM) is an evidence-based practice (EBP) (National Institute on Drug Abuse 2010) that theoretically appears compatible with the basic strategies used by judicial or probation officials within the U.S. criminal justice system. The justice system routinely uses reinforcers to address compliance with behavior for desired drug- and crime-free behaviors. The compatibility of positive reinforcers with the existing system lies in the similarity between the core concepts of CM and the principles of effective punishment: swift, certain, and increasingly intensified responses. Given this consistency with the core functions of justice processing, CM implementation in justice systems should be relatively easy to implement (Rogers 2003). Contingency management has wide applications in the area of behavior change. In substance abuse treatment settings, CM interventions reduce drug use and increase treatment retention for a wide variety of drug abusers (Stitzer et al. 2010). Among the core components of any CM protocol is a focus on reducing or eliminating certain behavior(s) (e.g., abstinence from drug and alcohol use) and the use of structured and transparent rewards or incentives as the primary driver of behavior change. As such, CM protocols consistently use systems in which points are assigned to desired positive behaviors. Clients earn rewards via redeeming earned accumulated points. In prior work on implementing CM within drug and alcohol addiction treatment, clinicians established CM guidelines detailing which behavior(s) needed changing and how rewards could be earned. In such studies, both clients and clinicians understand that the rewards are achievable only with demonstrated positive behavior in line with the target goal (Stitzer et al. 2010). more

Wednesday, December 10, 2014

Keeping Down with the Joneses to Save the Environment

The World Wildlife Fund urges us to think about the poor polar bears and penguins whose homes are swiftly melting away. Meanwhile, documentaries—from Al Gore’s classic An Inconvenient Truth to Showtime’s recent Years of Living Dangerously—warn us of impending natural disasters. Yes, environmentalists tend to go all dramatic on us in order to minimize energy use and treat the planet better. But these efforts, largely, aren’t working. A spate of new evidence suggests that fear-based approaches to environmental change are backfiring, resulting in little more than angry arguments over Thanksgiving dinner between you and your climate-change-doubting uncle. But about a decade ago, behavioral psychologists hit on an important finding: people are actually willing to save energy when they see how much energy their neighbors are using. A groundbreaking 2004 study revealed that, contrary to popular belief, people do not care much about saving money on their electricity bill or protecting the environment; the single biggest motivator to changing our energy consumption is our desire to keep up—or, in this case, down—with the more energy-efficient Joneses. more

Monday, December 08, 2014

Google Deep Mind: A New Artificial Intelligence

Demis Hassabis leads what is now called Google DeepMind. It is still headquartered in London and still has “solve intelligence” as its mission statement. Roughly 75 people strong at the time it joined Google, Hassabis has said he aimed to hire around 50 more. Around 75 percent of the group works on fundamental research. The rest form an “applied research team” that looks for opportunities to apply DeepMind’s techniques to existing Google products.  Over the next five years, DeepMind’s technology could be used to refine YouTube’s recommendations or improve the company’s mobile voice search. They dream of creating “AI scientists” that could do things like generate and test new hypotheses about disease in the lab. When prodded, he also says that DeepMind’s software could also be useful to robotics, an area in which Google has recently invested heavily. DeepMind has combined deep learning with a technique called reinforcement learning, which is inspired by the work of animal psychologists such as B.F. Skinner. This led to software that learns by taking actions and receiving feedback on their effects, as humans or animals often do. In 2013, DeepMind researchers showed off software that had learned to play three classic Atari games - Pong, Breakout and Enduro - better than an expert human. The software wasn’t programmed with any information on how to play; it was equipped only with access to the controls and the display, knowledge of the score, and an instinct to make that score as high as possible. The program became an expert gamer through trial and error. more