Science and IT news

Deepstack AI has beaten professional poker players in heads-up no-limit Texas hold’em

Artificial intelligence has seen a number of breakthroughs in recent years, with games often serving as significant milestones. A common feature of games with these successes is that they involve information symmetry among the players, where all players have identical information. This property of perfect information, though, is far more common in games than in real-world problems. Poker is the quintessential game of imperfect information, and it has been a longstanding challenge problem in artificial intelligence. In this paper researchers introduce DeepStack, a new algorithm for imperfect information settings such as poker. It combines recursive reasoning to handle information asymmetry, decomposition to focus computation on the relevant decision, and a form of intuition about arbitrary poker situations that is automatically learned from selfplay games using deep learning. In a study involving dozens of participants and 44,000 hands of poker, DeepStack becomes the first computer program to beat professional poker players in heads-up no-limit Texas hold’em. Furthermore, they show this approach dramatically reduces worst-case exploitability compared to the abstraction paradigm that has been favored for over a decade DeepStack was evaluated against 33 professional poker players from the International Federation of Poker. Each participant was asked to play a 3,000-game match over a month. DeepStack takes a fundamentally different approach. It continues to use the recursive reasoning of CFR to handle information asymmetry. However, it does not compute and store a complete strategy prior to play and so has no need for explicit abstraction. Instead it considers each particular situation as it arises during play, but not in isolation. It avoids reasoning about the entire remainder of the game by substituting the computation beyond a certain depth with a fast approximate estimate. This estimate can be thought of as DeepStack’s intuition: a gut feeling of the value of holding any possible private cards in any possible poker situation. Finally, DeepStack’s intuition, much like human intuition, needs to be trained. They train it with deep learning using examples generated from random poker situations. We show that DeepStack is theoretically sound, produces substantially less exploitable strategies than abstraction-based techniques, and is the first program to beat professional poker players at HUNL with a remarkable average win rate of over 450 mbb/g. A rival AI poker team of researcher from Carnegie Mellon University announced a $200,000 match between its system, Libratus, and four poker pros: Jason Les, Dong Kim, Daniel McAulay, and Jimmy Chou. Collectively, the four human pros will play 120,000 hands of heads-up no-limit Texas hold 'em over 20 days against Libratus. At the end of day two, however, Libratus was up by $150,126. It was winning against three players and losing against one. nextbigfuture.com  

'Tooth repair drug' may replace fillings

Teeth can be encouraged to repair themselves in a way that could see an end to fillings, say scientists. The team at King's College London showed that a chemical could encourage cells in the dental pulp to heal small holes in mice teeth. A biodegradable sponge was soaked in the drug and then put inside the cavity. The study, published in Scientific Reports, showed it led to "complete, effective natural repair". Teeth have limited regenerative abilities. They can produce a thin band of dentine - the layer just below the enamel - if the inner dental pulp becomes exposed, but this cannot repair a large cavity. Normally dentists have to repair tooth decay or caries with a filling made of a metal amalgam or a composite of powdered glass and ceramic. These can often need replacing multiple times during someone's lifetime, so the researchers tried to enhance the natural regenerative capacity of teeth to repair larger holes. They discovered that a drug called Tideglusib heightened the activity of stem cells in the dental pulp so they could repair 0.13mm holes in the teeth of mice. A drug-soaked sponge was placed in the hole and then a protective coating was applied over the top. As the sponge broke down it was replaced by dentine, healing the tooth. New treatment Prof Paul Sharpe, one of the researchers, told the BBC News website: "The sponge is biodegradable, that's the key thing. "The space occupied by the sponge becomes full of minerals as the dentine regenerates so you don't have anything in there to fail in the future." The team at King's is now investigating whether the approach can repair larger holes. Prof Sharpe said a new treatment could be available soon: "I don't think it's massively long term, it's quite low-hanging fruit in regenerative medicine and hopeful in a three-to-five year period this would be commercially available." The field of regenerative medicine - which encourages cells to rapidly divide to repair damage - often raises concerns about cancer. Tideglusib alters a series of chemical signals in cells, called Wnt, which has been implicated in some tumours. However, the drug has already been trialled in patients as a potential dementia therapy. "The safety work has been done and at much higher concentrations so hopefully we're on to a winner," said Prof Sharpe. This is only the latest approach in repairing teeth - another group at King's believe electricity can be used to strengthen a tooth by forcing minerals into the layer of enamel. Minerals such as calcium and phosphate naturally flow in and out of the tooth with acid, produced by bacteria munching on food in the mouth, helping to leach out minerals. The group apply a mineral cocktail and then use a small electric current to drive the minerals deep into the tooth. They say "Electrically Accelerated and Enhanced Remineralisation" can strengthen the tooth and reduce dental caries. bbc.com

Astronomers have found the source of a deep space radio wave burst for the first time

After a decade of bewilderment, astronomers have pinpointed the source of a mysterious blast of radio waves coming from deep outside the Milky Way: a dwarf galaxy located 3 billion light years from Earth. It’s a remarkable first in the study of what has been a tremendous astronomical puzzle. Scientists still don’t know what causes these deep space pulses, but locating the galaxy that spawned one brings us closer to figuring out where they come from. First discovered in 2007, only 18 of these phenomena have ever been detected. They’re called fast radio bursts, or FRBs, because they occur for just milliseconds; their fleeting nature makes it tough to catch one in action, and even tougher to figure out the exact spot in the sky they’re coming from. "A remarkable first in the study of what has been a tremendous astronomical puzzle" But astronomers got lucky when they found a particular burst known as FRB 121102: it is the only one known to repeat, meaning multiple radio bursts have been detected coming from the same location in the sky. That makes it easier for scientists to catch again, Shami Chatterjee, an astronomer at Cornell University who discovered the repetition, tells The Verge. That discovery gave Chatterjee the idea to continually observe FRB 121102 with a huge network of radio telescopes. And sure enough, he and his team were able to get high-resolution images of multiple bursts after many hours of observation, allowing them to track down the source of FRB 121102. Their work is detailed today in three studies published in Nature and The Astrophysical Journal Letters. Now that researchers know the cosmic neighborhood generating the FRBs, they can study the galaxy more closely — which may help figure out the origins of these fast pulses. There may be other uses for FRBs, too. Since FRBs originate so far away, they have to pass through a lot of interstellar junk, such as gas and plasma, to reach Earth. Once scientists know which exact galaxy the signal comes from, the radio waves could help scientists determine just how much gas and plasma they had to pass through to get here. “The analogy I use is that until now, we could potentially know the country it came from,” Heino Falcke, a radio astronomer at the Radboud University Nijmegen who wrote an accompanying editorial about the discovery, tells The Verge. “Now we know the home address.” The mystery of fast radio bursts When FRBs were first discovered, there was debate over whether or not these signals were actually coming from space at all. Astronomers wondered if they were just bizarre interference of some kind. But after a closer look, researchers realized FRBs are unique. Typically, a burst of radio waves will have different wave frequencies occurring at once, but FRBs have frequencies that are spread out. The highest frequencies of each FRB arrive slightly earlier at Earth while the lowest frequencies arrive slightly later. It’s a sign that the these FRBs are weary travelers, having journeyed through a lot of interstellar gas and plasma that’s mucking up their signals. A rendering of an FRB reaching Earth, with different frequencies arriving at different times. Jingchuan Yu, Beijing Planetarium And FRB signals are so mucked up that astronomers are convinced they’re coming from outside the Milky Way Galaxy. But that creates another problem: these bursts must come from a super bright source. “Like absolutely, incredibly bright,” says Chatterjee. Experts have come up with dozens of theories, such as the cataclysmic collision of neutron stars or a black hole tearing itself apart. But no one has agreed on a single explanation. Then the discovery of FRB 121102 changed everything. Because of its repeating nature, astronomers know that its source can’t be anything explosive or an object being destroyed. “Something like that could not repeat again at the same place at the same distance,” says Chatterjee. “So that basically put the end to a huge swath of models.” Maybe more than one thing is capable of creating FRBs — and that’s why there hasn’t been a single explanation. But the only way to know for sure was to find the host galaxy. Zeroing in on the home address To pinpoint the source, Chatterjee and his team turned to the Karl G. Jansky Very Large Array (VLA). It’s a collection of 27 radio telescopes in New Mexico, spread out in a Y shape. Until now, FRBs have only ever been detected with large, single dish radio telescopes that observe a wide patch of sky at at time. These types of telescopes are more sensitive to picking up FRBs, but they aren’t as adept at location. The VLA, however, acts like a telescope that is miles wide, allowing astronomers to get the high-resolution images they needed of FRB 121102’s home address. The dishes of the Karl G. Jansky Very Large Array. Danielle Futselaar At the spot they found, the researchers discovered two enticing details. There was indeed a persistent radio source at the location — possibly the origin of the repeating FRB. And there was a faint smudge of light that turned out to be a tiny galaxy. The researchers were then able to measure how fast the galaxy was moving away from Earth, using its spectrum of light. By figuring out the movement, researchers were able to place the galaxy 3 billion light years away. “That’s extraordinary, because that immediately tells you how much energy there must be when the pulse is emitted for it to travel for 3 billion years spreading out through the Universe and to get to us and be detectable,” says Chatterjee. Unsolved mysteries Chatterjee’s team still isn’t certain as to what’s causing the repetitive radio waves, though they have a few potential explanations. One idea is that the black hole at the center of the galaxy is active, spewing out jets of particles at light speed. Every now and then, a blob of plasma may vaporize one of the jets, creating a bright flash. “We don’t really like this model very much, but it’s possible,” says Chatterjee. Another possibility is that the FRB is coming from a type of dense neutron star with an incredibly strong magnetic field, called a magnetar. Astronomers have discovered magnetars in our galaxy that produce bright radio pulses, but nothing as bright as FRB 121101. So something would have to be amplifying the pulses, like the way a magnifying glass focuses a beam of light on ants. That may mean blobs of plasma are lining up just right to focus the radio waves on Earth, making them extra bright, says Chatterjee. “This is very plausible,” he says. “We’re not invoking any radical new physics.” "The next priority is to find another fast radio burst that repeats" FRB 121101 is just one sample, though, and scientists like multiple samples. So, the next priority is to find another FRB that repeats, says Chatterjee. Astronomers want to know if all FRBs repeat — a question that hasn’t been fully answered yet. It’s possible that every FRB has multiple bursts, and we simply haven’t detected them. Or perhaps there’s a whole spectrum of FRBs. “We know from gamma rays that there are multiple classes of them; the same could be true of fast radio bursts,” Duncan Lorimer, an astrophysicist at West Virginia University who discovered the first FRB, tells The Verge. “That’s in the back of people’s minds, that this could be one end of the population.” Finding another repeating FRB would make it easier to pinpoint another host galaxy. And once there is a big enough population of FRB sources, then scientists can really start to use them to measure the matter that’s between galaxies. Astronomers were able to probe intergalactic dust last year with a newly discovered FRB, but since they didn’t know the signal’s exact origin, their measurements aren’t ironclad. That’s not the case with today’s discovery, so it may still yield important data on the gas and plasma the signal ran through. “There’s no hand waving about this,” says Chatterjee. “We know it’s coming from this dwarf galaxy.” theverge.com

Intel spending $50 million to develop silicon qubits that they would scale to quantum computers with millions of qubit

Intel has a team of quantum hardware engineers in Portland, Oregon, who collaborate with researchers in the Netherlands, at TU Delft’sQuTech quantum research institute, under a $50 million grant established last year. Earlier this month Intel’s group reported that they can now layer the ultra-pure silicon needed for a quantum computer onto the standard wafers used in chip factories. This strategy makes Intel an outlier among industry and academic groups working on qubits, as the basic components needed for quantum computers are known. Other companies can run code on prototype chips with several qubits made from superconducting circuits A quantum computer would need to have thousands or millions of qubits to be broadly useful, though. And Jim Clarke, who leads Intel’s project as director of quantum hardware, argues that silicon qubits are more likely to get to that point (although Intel is also doing some research on superconducting qubits). One thing in silicon’s favor, he says: the expertise and equipment used to make conventional chips with billions of identical transistors should allow work on perfecting and scaling up silicon qubits to progress quickly. Intel’s silicon qubits represent data in a quantum property called the “spin” of a single electron trapped inside a modified version of the transistors in its existing commercial chips. “The hope is that if we make the best transistors, then with a few material and design changes we can make the best qubits,” says Clarke. Another reason to work on silicon qubits is that they should be more reliable than the superconducting equivalents. Still, all qubits are error prone because they work on data using very weak quantum effects.