Matt Ernst
A general reinforcement learning algorithm that masters chess, shogi, and Go through self-playComputers can beat humans at increasingly complex games, including chess and Go. However, these programs are typically constructed for a particular game, exploiting its properties, such as the symmetries of the board on which it is played. Silver et al. developed a program called AlphaZero, which taught itself to play Go, chess, and shogi (a Japanese version of chess) (see the Editorial, and the Perspective by Campbell). AlphaZero managed to beat state-of-the-art programs specializing in these three games. The ability of AlphaZero to adapt to various game rules is a notable step toward achieving a general game-playing system. Science , this issue p. [1140][1]; see also pp. [1087][2] and [1118][3] The game of chess is the longest-studied domain in the history of artificial intelligence. The strongest programs are based on a combination of sophisticated search techniques, domain-specific adaptations, and handcrafted evaluation functions that have been refined by human experts over several decades. By contrast, the AlphaGo Zero program recently achieved superhuman performance in the game of Go by reinforcement learning from self-play. In this paper, we generalize this approach into a single AlphaZero algorithm that can achieve superhuman performance in many challenging games. Starting from random play and given no domain knowledge except the game rules, AlphaZero convincingly defeated a world champion program in the games of chess and shogi (Japanese chess), as well as Go. [1]: /lookup/doi/10.1126/science.aar6404 [2]: /lookup/doi/10.1126/science.aaw2221 [3]: /lookup/doi/10.1126/science.aav1175
Magnesium-Sodium Hybrid Battery with High Voltage, Capacity and CyclabilityRechargeable magnesium battery has been widely considered as a potential alternative to current Li-ion technology. However, the lack of appropriate cathode with high-energy density and good sustainability hinders the realization of competitive magnesium cells. Recently, a new concept of hybrid battery coupling metal magnesium anode with a cathode undergoing the electrochemical cycling of a secondary ion has received increased attention. Mg-Na hybrid battery, for example, utilizes the dendritic-free deposition of magnesium at the anode and fast Na+-intercalation at the cathode to reversibly store and harvest energy. In the current work, the principles that take the full advantage of metal Mg anode and Na-battery cathode to construct high-performance Mg-Na hybrid battery are described. By rationally applying such design principle, we constructed a Mg-NaCrO2 hybrid battery using metal Mg anode, NaCrO2 cathode and a mixture of all-phenyl complex (PhMgCl-AlCl3, Mg-APC) and sodium carba-closo-dodecaborate (NaCB11H12) as dual-salt electrolyte. The Mg-NaCrO2 cell delivered an energy density of 183 Wh kg-1 at the voltage of 2.3 V averaged in 50 cycles. We found that the amount of electrolyte can be reduced by using solid MgCl2 as additional magnesium reservoir while maintaining comparable electrochemical performance. A hypothetical MgCl2-NaCrO2 hybrid battery is therefore proposed with energy density estimated to be 215 Wh kg-1 and the output voltage over 2 V.