The UK's Scientific Advisory Group for Emergencies (SAGE) advised the government to introduce mandatory hotel quarantine for travellers arriving into the UK two weeks ago, according to minutes from a meeting on 21 January that were leaked to the Times. On Thursday 21 January, SAGE reportedly warned that "reactive, geographically targeted" travel bans couldn't be relied on to prevent faster-spreading coronavirus variants, such as those identified in South Africa and Brazil, from reaching the UK, adding that: "no intervention, other than a complete, pre-emptive closure of borders, or the mandatory quarantine of all visitors upon arrival in designated facilities, irrespective of testing history, can get close to fully preventing the importation of new cases or new variants." A Downing Street spokesperson said SAGE did not directly advise UK prime minister Boris Johnson to close borders. Universities minister Michelle Donelan told Sky News that the government "always based our decisions on the best medical and scientific advice" and said "the SAGE advice actually said it would probably be ineffective, in fact, to close the borders, which was the same advice that we got at the time from the World Health Organization". Johnson announced geographically targeted hotel quarantine measures for travellers returning from 30 countries, including Brazil and South Africa, last week. UK health minister Matt Hancock urged people living in postcodes in England singled out for enhanced coronavirus testing for the so-called South Africa variant to stay at home unless "absolutely essential". Urgent door-to-door testing for the faster-spreading variant has been deployed after 11 cases with no link to foreign travel were identified in parts of England.

Tensor (multidimensional array) classification problem has become very popular in modern applications such as image recognition and high dimensional spatio-temporal data analysis. Support Tensor Machine (STM) classifier, which is extended from the support vector machine, takes CANDECOMP / Parafac (CP) form of tensor data as input and predicts the data labels. The distribution-free and statistically consistent properties of STM highlight its potential in successfully handling wide varieties of data applications. Training a STM can be computationally expensive with high-dimensional tensors. However, reducing the size of tensor with a random projection technique can reduce the computational time and cost, making it feasible to handle large size tensors on regular machines. We name an STM estimated with randomly projected tensor as Random Projection-based Support Tensor Machine (RPSTM). In this work, we propose a Tensor Ensemble Classifier (TEC), which aggregates multiple RPSTMs for big tensor classification. TEC utilizes the ensemble idea to minimize the excessive classification risk brought by random projection, providing statistically consistent predictions while taking the computational advantage of RPSTM. Since each RPSTM can be estimated independently, TEC can further take advantage of parallel computing techniques and be more computationally efficient. The theoretical and numerical results demonstrate the decent performance of TEC model in high-dimensional tensor classification problems. The model prediction is statistically consistent as its risk is shown to converge to the optimal Bayes risk. Besides, we highlight the trade-off between the computational cost and the prediction risk for TEC model. The method is validated by extensive simulation and a real data example. We prepare a python package for applying TEC, which is available at our GitHub.

We re-examine the topic of machine-learned clause selection guidance in saturation-based theorem provers. The central idea, recently popularized by the ENIGMA system, is to learn a classifier for recognizing clauses that appeared in previously discovered proofs. In subsequent runs, clauses classified positively are prioritized for selection. We propose several improvements to this approach and experimentally confirm their viability. For the demonstration, we use a Recursive Neural Network to classify clauses based on their derivation history and the presence or absence of automatically supplied theory axioms therein. The automatic theorem prover Vampire guided by the network achieves a 41 % improvement on a relevant subset of smt-lib in a real time evaluation.

At CloudFactory we have a pretty intimate seat at the table with well over 100 active tech teams applying AI to a myriad of different use cases and industries. Our clients come in all sizes and from all industries, from small startups to those listed on the Fortune 500, and they work on solutions from cashierless checkout to self-driving cars. With all the AI hype, it sometimes feels like a gold rush, which would make CloudFactory's workforce solutions the picks and shovels. We see no signs of an AI winter. Both AI adoption and business value will increase over the next 12 months.

Fleets of radar satellites are measuring movements on Earth like never before. East Africa has been called the cradle of humanity. But the geologically active region has also given birth to dozens of volcanoes. Few have been monitored for warnings of a potential eruption, and until recently, most were believed to be dormant. Then, Juliet Biggs decided to take a closer look—or rather, a farther look. Biggs, a geophysicist at the University of Bristol, uses a technique called interferometric synthetic aperture radar (InSAR) to detect tiny movements of Earth's surface from space. In a series of studies, she and her co-authors analyzed satellite data on the East African volcanoes. According to their latest results, which were published last month, 14 have been imperceptibly growing or shrinking in the past 5 years—a clue that magma or water is moving underground and that the volcanoes are not completely asleep. “It's really changed the way these volcanoes are viewed, from something that's kind of dormant to really very active systems,” Biggs says. After data showed that the Corbetti volcano, which abuts the fast-growing city of Hawassa, Ethiopia, is inflating steadily at a rate of 6.6 centimeters per year, Biggs's Ethiopian colleagues included it in the country's geological hazard monitoring network. No other technology could produce such a comprehensive survey. Individual GPS stations can track surface movements of less than 1 millimeter, but InSAR can measure changes almost as subtle across a swath hundreds of kilometers wide. That has made it a vital tool for earth scientists studying the heaves and sighs of our restive planet. “We tend to think of the ground as this solid platform,” Biggs says, “and actually, it's really not.” With InSAR, scientists are tracking how ice streams flow, how faults slip in earthquakes, and how the ground moves as fluids are pumped in or out. “Everywhere you look on Earth, you see something new,” says Paul Rosen, an InSAR pioneer at NASA's Jet Propulsion Laboratory (JPL). “It's a little bit like kids in a candy store.” And the flood of InSAR data is growing fast. Since 2018, the number of civil and commercial SAR satellites in orbit has more than doubled. And at least a dozen more are set to launch this year, which would bring the total to more than 60. With the help of computing advances that make data processing easier, the satellite fleets may soon be able to detect daily or even hourly surface changes at just about every patch of ground on Earth. As the technology grows more powerful and ubiquitous, InSAR is spreading beyond the geosciences. With InSAR data, railroads are monitoring the condition of their tracks and cities are monitoring shifts in buildings caused by construction. “It's popping up everywhere,” says Dáire Boyle, who follows trends in the space industry for Evenflow, a consulting firm in Brussels. Analysts value the SAR market at roughly $4 billion, and expect that figure to nearly double over the next 5 years. Many believe InSAR will eventually underpin our daily lives. From measuring the water stored in mountain snowpacks to enabling quick responses to natural disasters, InSAR data will prove invaluable to governments and industries, says Cathleen Jones, a science team leader for NISAR, an upcoming joint SAR mission from NASA and the Indian Space Research Organisation (ISRO). “I want it to become so socially relevant that they can't go back to not having this data.” SYNTHETIC APERTURE RADAR , the “SAR” on which InSAR depends, originated in the 1950s as a tool for airborne military reconnaissance. Like traditional radar, SAR instruments captured images of the planet by sending out microwave pulses and recording the echoes. And like a traditional radar, the instruments could penetrate clouds and worked equally well at night. A key difference was the “synthetic” aspect of SAR. Larger radar antennas, like larger apertures on a camera, collect more of the echoes and enable sharper pictures. But building a single antenna large enough to take a high-resolution image isn't practical. Researchers realized they could instead create an artificially large aperture by combining the signals received on a much smaller antenna as it moved through space. Today, SAR satellites with antennas just a few meters across can produce images with pixel resolutions as sharp as half a meter—better than many satellite-borne cameras. SAR images, on their own, suffice for many types of surveillance, from counterterrorism to tracking oil spills in the ocean. But InSAR goes further, by looking for differences between multiple SAR images. The technique takes advantage of phase information in the returning microwaves—in other words, where a signal is in its sinusoidal path when it hits the antenna. Any phase difference in the signal between SAR images taken from the same position at different times means the round-trip distance has changed, and can reveal surface movements down to a few millimeters. “There's nothing else that compares to it,” says Michelle Sneed, a hydrologist at the U.S. Geological Survey. “I'm still amazed by it after a couple of decades.” The 1978 launch of Seasat, NASA's first ocean-observing satellite, provided data for early InSAR efforts. Seasat operated for just 105 days before a power failure brought the mission to an untimely end. But in that time, it collected repeat images of California's Imperial Valley taken over the course of 12 days. Scientists at JPL later compared those images using InSAR to show the subtle swelling of fields as they soaked up irrigation water. “It is not hard to think of numerous applications for the type of instrument demonstrated,” the authors wrote in a 1989 paper. And they were right. ![Figure][1] CREDITS: (GRAPHIC) N. DESAI/ SCIENCE ; (DATA) ESA; WMO; GUNTER'S SPACE PAGE A classic InSAR study came in 1993, when a team of scientists in France used data from the SAR-enabled European Remote Sensing satellite to study a powerful earthquake that rocked Landers, California, the year before. By analyzing images taken before and after the quake, they calculated that the fault had slipped by up to 6 meters, which agreed with detailed field observations. The InSAR data also revealed how the ground buckled for kilometers around the fault—illustrating the full effects of the temblor at an unprecedented scale. The paper inspired scientists like Sneed, who went on to use InSAR to study how groundwater extraction causes the ground to sink. During a drought in California's San Joaquin Valley in the late 2000s, she and her colleagues discovered that the surface was subsiding as fast as 27 centimeters per year in places where farmers pumped the most groundwater. Irrigation canals were sagging as a result of uneven sinking, impeding water flow. “It's a really expensive problem,” Sneed says. (Another recent InSAR study linked specific water-intensive crops—notably corn, cotton, and soy—to increased subsidence.) Glaciologists adopted the technology, too. As a young researcher at JPL in the 1990s, Ian Joughin used InSAR—which tracks both vertical and horizontal movements—to measure the speed of polar ice streams. Some scientists thought flow rates would be relatively immune to climate change. But, sadly for the world, InSAR studies by Joughin and others proved those predictions wrong. “Especially in the early 2000s, we just saw all kinds of glaciers double their speed,” says Joughin, who now studies the fate of polar ice sheets and their contribution to sea-level rise at the University of Washington, Seattle. By the 2000s, many earth scientists were using InSAR—and grappling with its limitations. There were few SAR satellites in orbit, and they tended to switch between instruments or imaging modes to accommodate different users' needs, making the data hard to use for InSAR. The early missions collected the repeat images needed for InSAR only about once a month, and researchers often had to correct for their wobbly orbits. That meant that although scientists could study an event after it happened, they could rarely watch it unfold in real time. Leaders at the European Space Agency (ESA) were convinced there was a better way. MALCOLM DAVIDSON REMEMBERS the excitement and anxiety he felt on 3 April 2014, the day the first Sentinel-1 satellite launched. “All your life goes into a few minutes,” says Davidson, mission scientist for ESA's flagship SAR program. He also remembers the relief when the satellite safely reached orbit, and the awe that came over him when he saw its first image, of ocean swells. “It was very convincing that the mission was going to do great things,” he says. With Sentinel-1, the plan was simple: “We cut out all the experiments, and we said, ‘Look, this is a mapping machine.’” He and his colleagues chose a primary imaging mode to use over land—surveying a 250-kilometer swath at a resolution of 5 meters by 20 meters—that they hoped would satisfy most researchers, and made sure the orbits would overlap precisely, so all the data would be suitable for InSAR. The first satellite, Sentinel-1a, retraced its path every 12 days. Then, in 2016, ESA launched a clone that made repeat images available about every 6 days for many places on Earth. SAR missions like Italy's COSMO-SkyMed and Germany's TerraSAR-X also support InSAR and can achieve even higher resolutions. But they do not distribute data freely like Sentinel, which many credit for driving a transition from opportunistic experiments to what Davidson sees as “a more operational view of the world.” With Sentinel-1 data, Norway created a national deformation map that has helped identify rockslide hazards and revealed that parts of Oslo's main train station were sinking. Water managers in California rely on the data to track groundwater use and subsidence. And in Belgium, it is used to monitor the structural integrity of bridges. “It can all be done remotely now, saving time, saving money,” Boyle says. The large and growing body of InSAR data has also revealed small surface movements that were previously hidden by noise. As radar signals pass through the atmosphere, they slow down by an amount that depends on the weather, producing variability that can swamp tiny but important displacements. Thanks to long-term records from missions like Sentinel, researchers can now tease information from the noise, for example, helping them track movements of just a few millimeters per year in Earth's crust—enough to strain faults and eventually cause earthquakes. Such efforts would not have been possible without huge gains in computing power. In the 1990s, stacking a single pair of SAR images could take days, Sneed says, and interpreting the results could take much longer. Now, researchers can process hundreds of images overnight, and they increasingly rely on artificial intelligence (AI) algorithms to make sense of the data. In one recent test, an AI algorithm was tasked with identifying small fault movements known as slow earthquakes. It correctly found simulated and historical events, including ones that had eluded human InSAR experts, says Bertrand Rouet-Leduc, a geophysicist at Los Alamos National Laboratory who presented preliminary results in December 2020 at the annual meeting of the American Geophysical Union. Rouet-Leduc and his team now plan to monitor faults around the world using the same approach. He says it's mostly a matter of exploiting the vast quantity of data that “sits on servers without being looked at,” because it's simply too much for scientists to tackle. The researchers hope they will be able to answer questions like when and why slow earthquakes happen, and whether they can trigger big, damaging events by increasing stress on other parts of a fault. Commercial users often lack the expertise to process InSAR data, so hundreds of companies have sprung up to help. One, Dares Technology, monitors the ground for the construction, mining, and oil and gas industries. By tracking surface changes as fluids are injected or extracted from an oil reservoir, for example, Dares can help companies estimate pumping efficiency and prevent dangerous well failures. In the beginning, convincing clients that InSAR data were useful and trustworthy was difficult, says Dares CEO Javier Duro. Now, he says, “Everybody wants to include InSAR in their operations.” Duro is particularly interested in detecting precursors to accidents, for example, by looking for signs of instability in the walls of open-pit mines or in the dams used to store mine tailings. The company usually sends out several alerts per month to clients, who can take actions to avoid disasters. “Typically, InSAR data have been used for back analysis,” Duro says. “Our mission is to focus on the present and the future, and try to predict what could happen.” THE SURGE IN SATELLITES promises to bring yet another InSAR revolution. Italy, Japan, Argentina, and China all plan to launch additional SAR satellites soon, and NISAR, the NASA-ISRO mission, will take flight in late 2022 or early 2023. NISAR will image Earth's full land surface every 6 days, on average, says Rosen, the mission's project scientist. Its two radar sensors will help researchers track many things, including crop growth and changes in woody biomass—crucial for understanding the climate system. With a better view of Antarctica than other missions, NISAR can also monitor changes in ice. Taken together, Sentinel-1, NISAR, and the other civil satellites will image most places on Earth at least every 12 hours, Rosen says. But the temporal resolution of InSAR will remain constrained by the revisit rate of the individual missions, because the technique can't be done with imagery from different missions. However, private companies with large constellations of microsatellites hope to vault the field into yet another realm, by radically increasing revisit frequencies. On 24 January, a SpaceX Falcon 9 rocket blasted off from Cape Canaveral, Florida, carrying three satellites, each about the size of a minifridge and weighing less than 100 kilograms, from Iceye. The Finnish SAR startup has raised more than $150 million toward its audacious goal of imaging every square meter of Earth every hour. The launch brought Iceye's commercial constellation to six, giving it an early lead over rival companies such as Capella Space—which had two satellites on the same rocket—and Umbra, both based in California. Iceye plans to add at least eight more satellites this year, allowing it to revisit most of the globe once a day. “That is groundbreaking,” says Pekka Laurila, who co-founded Iceye as an undergraduate at Aalto University and now serves as the company's chief strategy officer. Ultimately, Iceye hopes to assemble a constellation of as many as 100 satellites as it approaches its hourly monitoring objective. That would open up new applications, like tracking how buildings and dams expand during the heat of the day and contract at night—a clue to their structural integrity. Already, Iceye data have been used to guide ships through Arctic sea ice and to track illegal fishing vessels. “If you can work closer to real time, you can actually do something about it,” Laurila says. So far, though, Iceye has focused on flood monitoring, which can guide disaster response efforts. In fact, the company provided some of the first images of Grand Bahama after Hurricane Dorian devastated the island in 2019, Laurila says. Precise flood data are also valuable to insurers, who can use them to trigger automatic insurance payouts after an event instead of processing claims and sending out inspectors. Until now, Iceye has tracked floods using regular SAR data, but it hopes to start to apply InSAR as it increases its revisit frequencies, because the technique can measure the height and extent of inundation much more precisely. And that's just the beginning of what Laurila hopes Iceye will do. His ultimate goal is to build a “new layer of digital infrastructure” that will provide a “real-time, always-available, objective view on the world,” he says. He believes that, like modern GPS, reliable SAR and InSAR data will support myriad applications, many of which have yet to be imagined. “Nobody thought of your Uber and pizza delivery when they thought of GPS,” Laurila says. If Iceye and its peers succeed, they will expose the shifts and shudders of the planet, day in and day out. They will spy tilting buildings and slumping slopes, and they will witness the growth of crops and the flow of commodities around the world. If space-based imagery often portrays Earth as quiet and still, InSAR reveals the true restlessness of our living planet. [1]: pending:yes

Human-made physical random number generators (RNGs) can be traced back 5000 years or more. Early examples such as knucklebones, two-sided throwsticks, or dice have been found in the Middle East, India, and China. RNGs were used for fortune telling and games of chance, with the oldest known board games of similar age as those of the number generators. Today, RNGs are vital for services and state-of-the-art technologies such as cryptographically secured communication, blockchain technologies, and quantum key distribution. Moreover, RNGs are needed in machine learning and scientific applications such as Monte Carlo numerical methods. On page 948 of this issue, Kim et al. ([ 1 ][1]) demonstrate an ultrafast RNG based on a broad-area laser with a multispot beam that is analogous to generating random numbers by using many dice at once. Random numbers are often generated by using a software algorithm running on a computer, called “pseudo”-random because the sequence eventually repeats. Moreover, relations among the numbers can exist that reveal that the numbers are not uniformly random. Hence, true RNGs (TRNGs) are of great interest, providing random numbers based on physical measurements that involve some noisy or stochastic process. All TRNGs have some nonidealities, such as generating zeroes more frequently than ones for a binary-output device, which must be mitigated by carefully engineering the device and postprocessing the data to improve the randomness quality ([ 2 ][2]). ![Figure][3] Creating bits with laser intensity Ultrafast random bits are generated from a broad-area laser with a bow-tie cavity. In order to generate the random bits, first intensities separated by ∼6 ps are subtracted from each other for the same positions on the detector. This creates bits of either ones or zeroes, which then undergo the exclusive-OR (XOR) logic operation with bits of another spot separated by half the width of the aperture. The XOR operation produces a one if the two inputs are different or a zero if the two inputs are the same. The broad-area laser allows for many different positions on the detector to be used simultaneously, allowing for fast generation of bits. GRAPHIC: N.DESAI/ SCIENCE Some applications require generating random numbers at very high rates, such as encrypting data in cloud-computing data centers, high-speed communication networks, or massive simulations. Photonic devices are a natural fit for these applications because of their potential for high-speed operation, compact size for chip-scale devices, and low power consumption. Recently, Marangon et al. ([ 3 ][4]) developed a TRNG that is based on interfering two different lasers on a beam splitter and detecting the resulting powers that emanate from its two output ports. The randomness comes about from quantum fluctuations in a laser due to a process known as spontaneous emission of photons. This process randomizes the phase of the light emitted by each laser, and this phase variation is converted to an intensity variation through the interference effect. Measuring which output port of the interferometer has the higher or lower intensity can be used to generate a one or a zero, respectively, at random. A compact device can be realized by generating random numbers in real time at a rate of 8 Gb/s for days at a time, passing tests that are used to assess the quality of the bit stream. The bottleneck in reaching higher speeds is that the lasers are single-mode and only generate a Gaussian beam–like spot at a single frequency. Kim et al. overcome this bottleneck by using a broad-area laser that simultaneously emits a plethora of modes, resulting in a multispot beam. The patterns undergo a complex dance, writhing and growing bright and dim because of phase and amplitude variation of the light within the laser (see the figure). For a good TRNG, engineering the broad-area laser cavity is especially necessary so that spatial and temporal correlations are minimized. The authors do so, which is a major achievement. Common broad-area lasers are known to exhibit irregular intensity pulsations in space and time because of the nonlinear interaction of light and the laser medium ([ 4 ][5]). Such instabilities result in correlations of their emission with characteristic spatial and temporal scales and are exceedingly difficult to avoid. This situation has plagued attempts to apply broad-area lasers more widely. Bittner et al. ([ 5 ][6]) showed that they could largely suppress the onset of spatiotemporal instabilities by using a cavity with a D shape, inspired by chaotic billiards; the balls on a D-shaped billiard table follow chaotic trajectories ([ 6 ][7]). Kim et al. introduce another approach based on adapting the shape of the cavity. After performing extensive numerical modeling, the authors chose a bow-tie shape and precisely microfabricated a laser chip. The authors managed to boost the number of modes, avoiding their locking, and thereby substantially reduced the spatial and temporal correlation scales to 1.5 µm and 2.8 ps, respectively. Another advantage of using the spatial degree of freedom of the special laser design is avoiding the two separate lasers and interference on an auxiliary beam splitter ([ 3 ][4]). Random numbers can be “harvested” from the complex emitted pattern by measuring the intensity at 254 spatial positions on ultrafast time scales (on the order of 1 ps) by using a special high-speed camera. This strategy is truly an ultrafast Demeter meeting chance. Through this effort, they achieved a random bit generation rate of 250 Tb/s, which is much more than an order of magnitude greater than previous efforts. A full technical implementation of such an ultrafast TRNG still faces several challenges that need to be overcome. The high-speed camera could only capture data over a limited time (∼2 ns), so they had to collect and concatenate multiple records to generate the more than 109 random numbers needed for the various statistical tests of randomness. Replacing the camera with a multitude of integrated photodetectors is yet to be achieved. Also, the required postprocessing of the measured intensities to ensure randomness is, at such speed, a task for the future. Looking beyond, the innovative approach to tailor the spatial and temporal emission properties of broad-area lasers and manipulating the nonlinear interaction of light with the laser medium opens other applications that require many degrees of freedom. Several machine-learning approaches are based on a random mapping of low-dimensional input data onto a high-dimensional-state space, which might be accomplished by injecting a data-encoded beam into a tailored laser. Hence, broad-area lasers may become attractive photonic integrated circuits for ultrafast information processing ([ 7 ][8], [ 8 ][9]). 1. [↵][10]1. K. Kim et al ., Science 371, 948 (2021). [OpenUrl][11][CrossRef][12] 2. [↵][13]1. J. D. Hart et al ., Appl. Phys. Lett. Photonics 2, 090901 (2017). [OpenUrl][14] 3. [↵][15]1. D. G. Marangon et al ., J. Lightwave Technol. 36, 3778 (2018). [OpenUrl][16] 4. [↵][17]1. I. Fischer, 2. O. Hess, 3. W. Elsäßer, 4. E. Göbel , Europhys. Lett. 35, 579 (1996). [OpenUrl][18][CrossRef][19] 5. [↵][20]1. S. Bittner et al ., Science 361, 1225 (2018). [OpenUrl][21][Abstract/FREE Full Text][22] 6. [↵][23]1. H. Cao, 2. J. Wiersig , Rev. Mod. Phys. 87, 61 (2015). [OpenUrl][24][CrossRef][25][PubMed][26] 7. [↵][27]1. P. R. Prucnal, 2. B. J. Shastri , Neuromorphic Photonics (CRC Press, 2017). 8. [↵][28]1. D. Brunner, 2. M. C. Soriano, 3. G. Van der Sande , Eds., Photonic Reservoir Computing (De Gruyter, 2019). [1]: #ref-1 [2]: #ref-2 [3]: pending:yes [4]: #ref-3 [5]: #ref-4 [6]: #ref-5 [7]: #ref-6 [8]: #ref-7 [9]: #ref-8 [10]: #xref-ref-1-1 "View reference 1 in text" [11]: {openurl}?query=rft.jtitle%253DScience%26rft.stitle%253DScience%26rft.aulast%253DKim%26rft.auinit1%253DK.%26rft.volume%253D371%26rft.issue%253D6532%26rft.spage%253D948%26rft.epage%253D952%26rft.atitle%253DMassively%2Bparallel%2Bultrafast%2Brandom%2Bbit%2Bgeneration%2Bwith%2Ba%2Bchip-scale%2Blaser%26rft_id%253Dinfo%253Adoi%252F10.1126%252Fscience.abc2666%26rft.genre%253Darticle%26rft_val_fmt%253Dinfo%253Aofi%252Ffmt%253Akev%253Amtx%253Ajournal%26ctx_ver%253DZ39.88-2004%26url_ver%253DZ39.88-2004%26url_ctx_fmt%253Dinfo%253Aofi%252Ffmt%253Akev%253Amtx%253Actx [12]: /lookup/external-ref?access_num=10.1126/science.abc2666&link_type=DOI [13]: #xref-ref-2-1 "View reference 2 in text" [14]: {openurl}?query=rft.jtitle%253DAppl.%2BPhys.%2BLett.%2BPhotonics%26rft.volume%253D2%26rft.spage%253D090901%26rft.genre%253Darticle%26rft_val_fmt%253Dinfo%253Aofi%252Ffmt%253Akev%253Amtx%253Ajournal%26ctx_ver%253DZ39.88-2004%26url_ver%253DZ39.88-2004%26url_ctx_fmt%253Dinfo%253Aofi%252Ffmt%253Akev%253Amtx%253Actx [15]: #xref-ref-3-1 "View reference 3 in text" [16]: {openurl}?query=rft.jtitle%253DJ.%2BLightwave%2BTechnol.%26rft.volume%253D36%26rft.spage%253D3778%26rft.genre%253Darticle%26rft_val_fmt%253Dinfo%253Aofi%252Ffmt%253Akev%253Amtx%253Ajournal%26ctx_ver%253DZ39.88-2004%26url_ver%253DZ39.88-2004%26url_ctx_fmt%253Dinfo%253Aofi%252Ffmt%253Akev%253Amtx%253Actx [17]: #xref-ref-4-1 "View reference 4 in text" [18]: {openurl}?query=rft.jtitle%253DEurophys.%2BLett.%26rft.volume%253D35%26rft.spage%253D579%26rft_id%253Dinfo%253Adoi%252F10.1209%252Fepl%252Fi1996-00154-7%26rft.genre%253Darticle%26rft_val_fmt%253Dinfo%253Aofi%252Ffmt%253Akev%253Amtx%253Ajournal%26ctx_ver%253DZ39.88-2004%26url_ver%253DZ39.88-2004%26url_ctx_fmt%253Dinfo%253Aofi%252Ffmt%253Akev%253Amtx%253Actx [19]: /lookup/external-ref?access_num=10.1209/epl/i1996-00154-7&link_type=DOI [20]: #xref-ref-5-1 "View reference 5 in text" [21]: {openurl}?query=rft.jtitle%253DScience%26rft_id%253Dinfo%253Adoi%252F10.1126%252Fscience.aas9437%26rft_id%253Dinfo%253Apmid%252F30115744%26rft.genre%253Darticle%26rft_val_fmt%253Dinfo%253Aofi%252Ffmt%253Akev%253Amtx%253Ajournal%26ctx_ver%253DZ39.88-2004%26url_ver%253DZ39.88-2004%26url_ctx_fmt%253Dinfo%253Aofi%252Ffmt%253Akev%253Amtx%253Actx [22]: /lookup/ijlink/YTozOntzOjQ6InBhdGgiO3M6MTQ6Ii9sb29rdXAvaWpsaW5rIjtzOjU6InF1ZXJ5IjthOjQ6e3M6ODoibGlua1R5cGUiO3M6NDoiQUJTVCI7czoxMToiam91cm5hbENvZGUiO3M6Mzoic2NpIjtzOjU6InJlc2lkIjtzOjEzOiIzNjEvNjQwOC8xMjI1IjtzOjQ6ImF0b20iO3M6MjI6Ii9zY2kvMzcxLzY1MzIvODg5LmF0b20iO31zOjg6ImZyYWdtZW50IjtzOjA6IiI7fQ== [23]: #xref-ref-6-1 "View reference 6 in text" [24]: {openurl}?query=rft.jtitle%253DRev.%2BMod.%2BPhys.%26rft.volume%253D87%26rft.spage%253D61%26rft_id%253Dinfo%253Adoi%252F10.1103%252FRevModPhys.87.61%26rft_id%253Dinfo%253Apmid%252F25739324%26rft.genre%253Darticle%26rft_val_fmt%253Dinfo%253Aofi%252Ffmt%253Akev%253Amtx%253Ajournal%26ctx_ver%253DZ39.88-2004%26url_ver%253DZ39.88-2004%26url_ctx_fmt%253Dinfo%253Aofi%252Ffmt%253Akev%253Amtx%253Actx [25]: /lookup/external-ref?access_num=10.1103/RevModPhys.87.61&link_type=DOI [26]: /lookup/external-ref?access_num=25739324&link_type=MED&atom=%2Fsci%2F371%2F6532%2F889.atom [27]: #xref-ref-7-1 "View reference 7 in text" [28]: #xref-ref-8-1 "View reference 8 in text"

Over the course of 2020, UNICEF hosted a series of global consultations with adolescents to learn about their views and knowledge of the artificial intelligence (AI) systems playing an increasingly important role in their lives. We spoke with 245 young people from Brazil, Chile, South Africa, Sweden and the United States in a series of workshops. These young voices helped shape UNICEF's recent draft Policy Guidance on AI for Children, which includes recommendations for child-centred AI and is aimed at governments and businesses. The workshops were part of a broader AI for Children Project led by the UNICEF Office of Global Insight and Policy.

"Talent and technology are the keys to unlocking our future in this industry-- finding ways for tech to come in and do a better job than people can in roles people have traditionally done," said MDC Partners global president Julia Hammond in explaining AI's value to her holding company. "The challenge with that is it's completely contradictory to the agency model, which has been built around people, so there's been a reluctance to build out AI and machine learning. We're actively pursuing it, in how we resource, how we scale and how we serve clients." Progress is being made elsewhere to find a happy middle ground. Last week, GroupM agency Wavemaker went public with its AI-driven media planning tool, Maximize, which the company claims is generating plans faster and more effectively than human planning teams alone. "It's a question of complexity of the problem solved.

Recently, transformer-based language models have stretched notions of what is possible with the simple self-supervised objective of language modeling, becoming a fixture in state of the art language technologies [Vaswani et al., 2017, Devlin et al., 2019, Brown et al., 2020]. However, the black box nature of these models combined with the complexity of natural language makes it challenging to measure how accurately they represent the world state underlying the text. In order to better measure the extent to which these models can capture the world state underlying the symbolic data they consume, we propose training and studying transformer language models for the game of chess. Chess provides a simple, constrained, and deterministic domain where the exact world state is known. Chess games can also be transcribed exactly and unambiguously using chess notations (Section 2). Most importantly, the form of chess notations allows us to probe our language models for aspects of the board state using simple prompts (Section 3) and without changing the language modeling objective or introducing any new classifiers.

This paper presents our systems for the three Subtasks of SemEval Task4: Reading Comprehension of Abstract Meaning (ReCAM). We explain the algorithms used to learn our models and the process of tuning the algorithms and selecting the best model. Inspired by the similarity of the ReCAM task and the language pre-training, we propose a simple yet effective technology, namely, negative augmentation with language model. Evaluation results demonstrate the effectiveness of our proposed approach. Our models achieve the 4th rank on both official test sets of Subtask 1 and Subtask 2 with an accuracy of 87.9% and an accuracy of 92.8%, respectively. We further conduct comprehensive model analysis and observe interesting error cases, which may promote future researches.