DeepMind’s New AI Taught Itself to Be the World’s Greatest Go Player

http://ift.tt/2xh8PLe

The AlphaGo AI that grabbed headlines last year after beating a master of the board game Go has just been trounced 100-0 by an updated version. And unlike its predecessor, the new system taught itself from first principles paving the way for AI that can think for itself.

When chess fell to AI in the 1990s, computer scientists looking for a new challenge turned to the millennia-old Chinese game Go, which despite its simpler rules has many more possible moves and often requires players to rely on instinct.

It was predicted it would be decades before an AI could beat a human master, but last year a program called AlphaGo developed by Google’s DeepMind subsidiary beat 18-time world champion Lee Sedol 4–1 in a series of matches in South Korea.

It was a watershed moment for AI research that showcased the power of the “reinforcement learning” approach championed by DeepMind. Not only did the system win, it also played some surprising yet highly effective moves that went against centuries of accumulated wisdom about how the game works.

Now, just a year later, DeepMind has unveiled a new version of the program called AlphaGo Zero in a paper in Nature that outperforms the version that beat Sedol on every metric. In just three days and 4.9 million training games, it reached the same level that took its predecessor several months and 30 million training games to achieve. It also did this on just four of Google’s tensor processing units—specialized chips for training neural networks—compared to 48 for AlphaGo.

Image Credit: DeepMind

The most striking departure from the previous system is the simplicity of the inputs. AlphaGo learned the basics of Go by analyzing thousands of games between human players, before honing its skills by playing itself millions of times. In contrast, AlphaGo Zero started with nothing more than the rules of the game and learned entirely from playing games against itself starting with completely random moves.

The system’s design is not radically different from its predecessor or the later AlphaGo Master version that defeated a host of human experts, including world number one Ke Jie, which AlphaGo Zero surpassed after 40 days of training. Essentially it is a streamlining of the previous approach, enabled by a simplified architecture and more powerful algorithms.

AlphaGo features two separate neural networks. The first was trained to predict the probable best move first using the human data and then by playing itself, while the second network was trained to predict the winner of these self-play games. When it came to actually playing a game, these networks were combined with a search algorithm to drill down on the best move given the state of the game.

The first network would select the best possible moves and then the system would use a combination of the value network and “rollouts”—a series of quick simulated games to test out possible moves—to decide on a play.

The new system combines the two neural networks into a single one with many more layers of artificial neurons, which can be trained more efficiently. It also uses a much simpler search algorithm and does away with rollouts, instead relying on the higher-quality neural network to make predictions. Speaking to Nature, lead researcher David Silver likened this to asking an expert to make a prediction rather than relying on hundreds of average players to test moves out.

“It not only independently discovered known moves that have taken millennia for humans to develop, it created entirely new ones that are now redefining how Go is played.”

The fact that the researchers managed to dramatically increase performance while simplifying the system is particularly impressive, considering that many recent advances in machine learning have come from throwing more data or processors at problems. “It shows it’s the novel algorithms that count, not the computing power or the data,” Silver told the BBC.

There are the usual caveats that come with breakthroughs in AI, and in particular, reinforcement learning. The program had to play itself millions of times before it became a world-beater, many more games than a human Go player would require to reach expert level. Its achievements are also constrained to the highly ordered world of Go, which is a far cry from the messy and uncertain problems AI will eventually be asked to solve in real life.

Nevertheless, a computer that can play millions of games in a matter of days still learns immeasurably faster than a human, so this shouldn’t be seen as a major limitation. And while the transition is likely to be slow and faltering, researchers at DeepMind are already working on applying similar techniques to those at the heart of AlphaGo Zero to practical applications. In a blog post, DeepMind said the approach could hold promise in other structured problems like protein folding, reducing energy consumption, or material design.

But most importantly, the advance is the most potent demonstration so far that AI could go beyond human intelligence. In their paper, the researchers describe how when they tried training AlphaGo Zero on human games, it learned faster, but actually did worse in the long run. Left to its own devices, it not only independently discovered known moves that have taken millennia for humans to develop, it created entirely new ones that are now redefining how Go is played.

“We’ve actually removed the constraints of human knowledge and it’s able, therefore, to create knowledge itself from first principles, from a blank slate,” Silver told the BBC.

Image Credit: DeepMind

from Singularity Hub http://ift.tt/2y07nfX

Advertisements

The Search to Find Another Earth Like Ours Is Heating Up

Are there other planets out there like our own? If so, could other forms of life, intelligence, or civilizations inhabit them?

This age-old question is a key part of the study of exoplanetology, the scientific field dedicated to looking for planets beyond our solar system. The first exoplanet was discovered in 1988, and since, the search has been ceaseless—and pretty successful too.

In this week’s episode of Tech-x-planations you’ll learn about exciting exoplanet discoveries over the last few decades, which ones might have liquid water, and whether any may be capable of sustaining human life.

Stock media provided by vision008 / Pond5

from Singularity Hub http://ift.tt/2xZJHIC

Why Peak Globalization Is the Path to a Sustainable World Economy

At some point in the future—and in some ways we are already seeing this—the amount of physical stuff moving around the world will peak and begin to decline. By “stuff,” I am referring to liquid fuels, coal, containers on ships, food, raw materials, products, etc.

New technologies are moving us toward “production-at-the-point-of-consumption” of energy, food, and products with reduced reliance on a global supply chain.

The trade of physical stuff has been central to globalization as we’ve known it. So, this declining movement of stuff may signal we are approaching “peak globalization.”

To be clear, even as the movement of stuff may slow, if not decline, the movement of people, information, data, and ideas around the world is growing exponentially and is likely to continue doing so for the foreseeable future.

Peak globalization may provide a pathway to preserving the best of globalization and global interconnectedness, enhancing economic and environmental sustainability, and empowering individuals and communities to strengthen democracy.

At the same time, some of the most troublesome aspects of globalization may be eased, including massive financial transfers to energy producers and loss of jobs to manufacturing platforms like China. This shift could bring relief to the “losers” of globalization and ease populist, nationalist political pressures that are roiling the developed countries.

That is quite a claim, I realize. But let me explain the vision.

New Technologies and Businesses: Digital, Democratized, Decentralized

The key factors moving us toward peak globalization and making it economically viable are new technologies and innovative businesses and business models allowing for “production-at-the-point-of-consumption” of energy, food, and products.

Exponential technologies are enabling these trends by sharply reducing the “cost of entry” for creating businesses. Driven by Moore’s Law, powerful technologies have become available to almost anyone, anywhere.

Beginning with the microchip, which has had a 100-billion-fold improvement in 40 years—10,000 times faster and 10 million times cheaper—the marginal cost of producing almost everything that can be digitized has fallen toward zero.

A hard copy of a book, for example, will always entail the cost of materials, printing, shipping, etc., even if the marginal cost falls as more copies are produced. But the marginal cost of a second digital copy, such as an e-book, streaming video, or song, is nearly zero as it is simply a digital file sent over the Internet, the world’s largest copy machine.* Books are one product, but there are literally hundreds of thousands of dollars in once-physical, separate products jammed into our devices at little to no cost.

A smartphone alone provides half the human population access to artificial intelligence—from SIRI, search, and translation to cloud computing—geolocation, free global video calls, digital photography and free uploads to social network sites, free access to global knowledge, a million apps for a huge variety of purposes, and many other capabilities that were unavailable to most people only a few years ago.

As powerful as dematerialization and demonetization are for private individuals, they’re having a stronger effect on businesses. A small team can access expensive, advanced tools that before were only available to the biggest organizations. Foundational digital platforms, such as the internet and GPS, and the platforms built on top of them by the likes of Google, Apple, Amazon, and others provide the connectivity and services democratizing business tools and driving the next generation of new startups.

“As these trends gain steam in coming decades, they’ll bleed into and fundamentally transform global supply chains.”

An AI startup, for example, doesn’t need its own server farm to train its software and provide service to customers. The team can rent computing power from Amazon Web Services. This platform model enables small teams to do big things on the cheap. And it isn’t just in software. Similar trends are happening in hardware too. Makers can 3D print or mill industrial grade prototypes of physical stuff in a garage or local maker space and send or sell designs to anyone with a laptop and 3D printer via online platforms.

These are early examples of trends that are likely to gain steam in coming decades, and as they do, they’ll bleed into and fundamentally transform global supply chains.

The old model is a series of large, connected bits of centralized infrastructure. It makes sense to mine, farm, or manufacture in bulk when the conditions, resources, machines, and expertise to do so exist in particular places and are specialized and expensive. The new model, however, enables smaller-scale production that is local and decentralized.

To see this more clearly, let’s take a look at the technological trends at work in the three biggest contributors to the global trade of physical stuff—products, energy, and food.

Products

3D printing (additive manufacturing) allows for distributed manufacturing near the point of consumption, eliminating or reducing supply chains and factory production lines.

This is possible because product designs are no longer made manifest in assembly line parts like molds or specialized mechanical tools. Rather, designs are digital and can be called up at will to guide printers. Every time a 3D printer prints, it can print a different item, so no assembly line needs to be set up for every different product. 3D printers can also print an entire finished product in one piece or reduce the number of parts of larger products, such as engines. This further lessens the need for assembly.

Because each item can be customized and printed on demand, there is no cost benefit from scaling production. No inventories. No shipping items across oceans. No carbon emissions transporting not only the final product but also all the parts in that product shipped from suppliers to manufacturer. Moreover, 3D printing builds items layer by layer with almost no waste, unlike “subtractive manufacturing” in which an item is carved out of a piece of metal, and much or even most of the material can be waste.

Finally, 3D printing is also highly scalable, from inexpensive 3D printers (several hundred dollars) for home and school use to increasingly capable and expensive printers for industrial production. There are also 3D printers being developed for printing buildings, including houses and office buildings, and other infrastructure.

The technology for finished products is only now getting underway, and there are still challenges to overcome, such as speed, quality, and range of materials. But as methods and materials advance, it will likely creep into more manufactured goods.

Ultimately, 3D printing will be a general purpose technology that involves many different types of printers and materials—such as plastics, metals, and even human cells—to produce a huge range of items, from human tissue and potentially human organs to household items and a range of industrial items for planes, trains, and automobiles.

Energy

Renewable energy production is located at or relatively near the source of consumption.

Although electricity generated by solar, wind, geothermal, and other renewable sources can of course be transmitted over longer distances, it is mostly generated and consumed locally or regionally. It is not transported around the world in tankers, ships, and pipelines like petroleum, coal, and natural gas.

Moreover, the fuel itself is free—forever. There is no global price on sun or wind. The people relying on solar and wind power need not worry about price volatility and potential disruption of fuel supplies as a result of political, market, or natural causes.

Renewables have their problems, of course, including intermittency and storage, and currently they work best if complementary to other sources, especially natural gas power plants that, unlike coal plants, can be turned on or off and modulated like a gas stove, and are half the carbon emissions of coal.

Within the next decades or so, it is likely the intermittency and storage problems will be solved or greatly mitigated. In addition, unlike coal and natural gas power plants, solar is scalable, from solar panels on individual homes or even cars and other devices, to large-scale solar farms. Solar can be connected with microgrids and even allow for autonomous electricity generation by homes, commercial buildings, and communities.

It may be several decades before fossil fuel power plants can be phased out, but the development cost of renewables has been falling exponentially and, in places, is beginning to compete with coal and gas. Solar especially is expected to continue to increase in efficiency and decline in cost.

Given these trends in cost and efficiency, renewables should become obviously cheaper over time—if the fuel is free for solar and has to be continually purchased for coal and gas, at some point the former is cheaper than the latter. Renewables are already cheaper if externalities such as carbon emissions and environmental degradation involved in obtaining and transporting the fuel are included.

Food

Food can be increasingly produced near the point of consumption with vertical farms and eventually with printed food and even printed or cultured meat.

These sources bring production of food very near the consumer, so transportation costs, which can be a significant portion of the cost of food to consumers, are greatly reduced. The use of land and water are reduced by 95% or more, and energy use is cut by nearly 50%. In addition, fertilizers and pesticides are not required and crops can be grown 365 days a year whatever the weather and in more climates and latitudes than is possible today.

While it may not be practical to grow grains, corn, and other such crops in vertical farms, many vegetables and fruits can flourish in such facilities. In addition, cultured or printed meat is being developed—the big challenge is scaling up and reducing cost—that is based on cells from real animals without slaughtering the animals themselves.

There are currently some 70 billion animals being raised for food around the world [PDF] and livestock alone counts for about 15% of global emissions. Moreover, livestock places huge demands on land, water, and energy. Like vertical farms, cultured or printed meat could be produced with no more land use than a brewery and with far less water and energy.

A More Democratic Economy Goes Bottom Up

This is a very brief introduction to the technologies that can bring “production-at-the-point-of-consumption” of products, energy, and food to cities and regions.

What does this future look like? Here’s a simplified example.

Imagine a universal manufacturing facility with hundreds of 3D printers printing tens of thousands of different products on demand for the local community—rather than assembly lines in China making tens of thousands of the same product that have to be shipped all over the world since no local market can absorb all of the same product.

Nearby, a vertical farm and cultured meat facility produce much of tomorrow night’s dinner. These facilities would be powered by local or regional wind and solar. Depending on need and quality, some infrastructure and machinery, like solar panels and 3D printers, would live in these facilities and some in homes and businesses.

The facilities could be owned by a large global corporation—but still locally produce goods—or they could be franchised or even owned and operated independently by the local population. Upkeep and management at each would provide jobs for communities nearby. Eventually, not only would global trade of parts and products diminish, but even required supplies of raw materials and feed stock would decline since there would be less waste in production, and many materials would be recycled once acquired.

“Peak globalization could be a viable pathway to an economic foundation that puts people first while building a more economically and environmentally sustainable future.”

This model suggests a shift toward a “bottom up” economy that is more democratic, locally controlled, and likely to generate more local jobs.

The global trends in democratization of technology make the vision technologically plausible. Much of this technology already exists and is improving and scaling while exponentially decreasing in cost to become available to almost anyone, anywhere.

This includes not only access to key technologies, but also to education through digital platforms available globally. Online courses are available for free, ranging from advanced physics, math, and engineering to skills training in 3D printing, solar installations, and building vertical farms. Social media platforms can enable local and global collaboration and sharing of knowledge and best practices.

These new communities of producers can be the foundation for new forms of democratic governance as they recognize and “capitalize” on the reality that control of the means of production can translate to political power. More jobs and local control could weaken populist, anti-globalization political forces as people recognize they could benefit from the positive aspects of globalization and international cooperation and connectedness while diminishing the impact of globalization’s downsides.

There are powerful vested interests that stand to lose in such a global structural shift. But this vision builds on trends that are already underway and are gaining momentum. Peak globalization could be a viable pathway to an economic foundation that puts people first while building a more economically and environmentally sustainable future.

This article was originally posted on Open Democracy (CC BY-NC 4.0). The version above was edited with the author for length and includes additions. Read the original article on Open Democracy.

* See Jeremy Rifkin, The Zero Marginal Cost Society, (New York: Palgrave Macmillan, 2014), Part II, pp. 69-154.

Image Credit: Sergey Nivens / Shutterstock.com

from Singularity Hub http://ift.tt/2hXmvVn

Some things you wanted to know about robot sex* (but were afraid to ask)

BOOK LAUNCH – BUY NOW!

I am pleased to announce that Robot Sex: Social and Ethical Implications (MIT Press, 2017), edited by myself and Neil McArthur, is now available for purchase. You can buy the hardcopy/ebook via Amazon in the US. You can buy the ebook in the UK as well, but the hardcopy might take another few weeks to arrive. I’ve never sold anything before via this blog. That all changes today. Now that I actually have something to sell, I’m going to turn into the most annoying, desperate, cringeworthy and slightly pathetic salesman you could possibly imagine…

from Ethical Technology http://ift.tt/2xdAooH

A sneak peak at radical future user interfaces for phones, computers, and VR

http://img.youtube.com/vi/8uRskDzkKl0/0.jpg

Grabity: a wearable haptic interface for simulating weight and grasping in VR (credit: UIST 2017)

Drawing in air, touchless control of virtual objects, and a modular mobile phone with snap-in sections (for lending to friends, family members, or even strangers) are among the innovative user-interface concepts to be introduced at the 30th ACM User Interface Software and Technology Symposium (UIST 2017) on October 22–25 in Quebec City, Canada.

Here are three other concepts to be presented, developed by researchers at Dartmouth College’s human computer interface lab.

Retroshape tactile watch feedback

Darthmouth’s Retroshape concept would add a shape-deforming tactile feedback system to the back of a future watch, allowing you to both see and feel virtual objects, such as a bouncing ball or exploding asteroid. Each pixel on RetroShape’s screen has a corresponding “taxel” (tactile pixel) on the back of the watch, using 16 independently moving pins.

VIDEO
UIST 2017 | RetroShape: Leveraging Rear-Surface Shape Displays for 2.5D Interaction on Smartwatches

Frictio smart ring

Current ring-gadget designs will allow users to control things. Instead, Frictio uses controlled rotation to provide silent haptic alerts and other feedback.

VIDEO
UIST 2017 — Frictio: Passive Kinesthetic Force Feedback for Smart Ring Output

Pyro: fingertip control

Pyro is based on moving the thumb tip against the index finger, a natural, fast and unobtrusive way to interact with a computer or other devices. It uses an energy-efficient thermal infrared sensor to detect to detect micro control gestures, based on patterns of heat radiating from fingers.

VIDEO
UIST 2017 — Pyro: Thumb-Tip Gesture Recognition Using Pyroelectric Infrared Sensing

Here some highlights from other presentations at the ACM Symposium on User Interface:

VIDEO
UIST 2017 Technical Papers Preview

from KurzweilAI » News http://ift.tt/2gWftAn

A sneak peak at radical future user interfaces for phones, computers, and VR

http://img.youtube.com/vi/8uRskDzkKl0/0.jpg

Grabity: a wearable haptic interface for simulating weight and grasping in VR (credit: UIST 2017)

Drawing in air, touchless control of virtual objects, and a modular mobile phone with snap-in sections (for lending to friends, family members, or even strangers) are among the innovative user-interface concepts to be introduced at the 30th ACM User Interface Software and Technology Symposium (UIST 2017) on October 22–25 in Quebec City, Canada.

Here are three other concepts to be presented, developed by researchers at Dartmouth College’s human computer interface lab.

Retroshape tactile watch feedback

Darthmouth’s Retroshape concept would add a shape-deforming tactile feedback system to the back of a future watch, allowing you to both see and feel virtual objects, such as a bouncing ball or exploding asteroid. Each pixel on RetroShape’s screen has a corresponding “taxel” (tactile pixel) on the back of the watch, using 16 independently moving pins.


UIST 2017 | RetroShape: Leveraging Rear-Surface Shape Displays for 2.5D Interaction on Smartwatches

Frictio smart ring

Current ring-gadget designs will allow users to control things. Instead, Frictio uses controlled rotation to provide silent haptic alerts and other feedback.


UIST 2017 — Frictio: Passive Kinesthetic Force Feedback for Smart Ring Output

Pyro: fingertip control

Pyro is based on moving the thumb tip against the index finger, a natural, fast and unobtrusive way to interact with a computer or other devices. It uses an energy-efficient thermal infrared sensor to detect to detect micro control gestures, based on patterns of heat radiating from fingers.


UIST 2017 — Pyro: Thumb-Tip Gesture Recognition Using Pyroelectric Infrared Sensing

Here some highlights from other presentations at the ACM Symposium on User Interface:


UIST 2017 Technical Papers Preview

from KurzweilAI http://ift.tt/2gWftAn

This Week’s Awesome Stories From Around the Web (Through October 21)

ARTIFICIAL INTELLIGENCE
Stunning AI Breakthrough Takes Us One Step Closer to the Singularity
George Dvorsky | Gizmodo
“A new paper published in Nature today describes how the artificially intelligent system that defeated Go grandmaster Lee Sedol in 2016 got its digital ass kicked by a new-and-improved version of itself…Now, every once in a while the field of AI experiences a ‘holy shit’ moment, and this would appear to be one of those moments.”

AUGMENTED REALITY
Magic Leap Confirms $502 Million Series D Round
Megan Rose Dickey | TechCrunch
“It’s still not totally clear what Magic Leap is doing, but it sure has raised a ton of money (more than $1.9 billion) in order to do whatever it is that it’s doing. To date, we’ve been able to gather that the company may be launching a device called ‘Magic Leap One.’ And last month, Bloomberg suggested Magic Leap may be gearing up to ship that device to a ‘small group of users’ in the next six months or so.”

TECHNOLOGY AND SOCIETY
This Alternative Credit Agency Lets Immigrants Bring Their Credit With Them
Ben Schiller | Fast Company
Millions of immigrants enter the United States lacking something important for success: a documented financial history or credit score… Nova Credit is trying to fill the gap. For the last two years, it’s reached deals with credit bureaus in 15 countries to create a ‘credit passport’—a way of porting financial history data across borders. Its business-to-business dashboard makes immigrants visible once more, translating data stored abroad into a form that can be consumed by banks, telephone companies, and landlords here.”

CRYPTOCURRENCIES 
Craig Wright Couldn’t Prove He Invented Bitcoin, but He’s Back Anyway
Jordan Pearson | Motherboard
“In the last six months, Wright has returned to the world of Bitcoin with a vengeance. He is the chief scientist of a new Bitcoin research company called nChain, which has already filed numerous patents for Bitcoin technology. He tweets with unrestrained contempt for his perceived enemies in Bitcoin. He gives talks at conferences for the online gambling industry while wearing sunglasses indoors. The one thing Craig Wright hasn’t done during his comeback tour is to clarify whether or not he is Satoshi.”

TRANSPORTATION
Carnegie Mellon Solves 12-Year-Old DARPA Grand Challenge Mystery
Evan Ackerman | IEEE Spectrum
“Even after the event, CMU wasn’t able to figure out exactly what happened. But last weekend, at an event celebrating the 10th anniversary of the DARPA Urban Challenge (which CMU won handily with their autonomous Chevy Tahoe BOSS), they accidentally stumbled onto what went wrong.”

Image Credit: Alexander Herasymchuk / Shutterstock.com

from Singularity Hub http://ift.tt/2zFp9ah