Amazon will continue to sell its controversial facial recognition software to law enforcement and other entities after its shareholders shot down a proposal to reel the technology in. The vote effectively kills two initiatives brought before Amazon's board. One proposal would have required board approval to sell the software to governments, with approval only being given if the client meets certain standards of civil liberties. Another proposal called for a study on the technology's implications on rights and privacy. The exact breakdown of the vote is unclear and according to an Amazon representative it will only be made available via SEC filings later this week.

Generative Adversarial Networks (GANs) enjoy great success at image generation, but have proven difficult to train in the domain of natural language. Challenges with gradient estimation, optimization instability, and mode collapse have lead practitioners to resort to maximum likelihood pre-training, followed by small amounts of adversarial fine-tuning. The benefits of GAN fine-tuning for language generation are unclear, as the resulting models produce comparable or worse samples than traditional language models. We show it is in fact possible to train a language GAN from scratch -- without maximum likelihood pre-training. We combine existing techniques such as large batch sizes, dense rewards and discriminator regularization to stabilize and improve language GANs. The resulting model, ScratchGAN, performs comparably to maximum likelihood training on EMNLP2017 News and WikiText-103 corpora according to quality and diversity metrics.

Spiking neural networks (SNNs) offer a promising alternative to current artificial neural networks to enable low-power event-driven neuromorphic hardware. Spike-based neuromorphic applications require processing and extracting meaningful information from spatio-temporal data, represented as series of spike trains over time. In this paper, we propose a method to synthesize images from multiple modalities in a spike-based environment. We use spiking auto-encoders to convert image and audio inputs into compact spatio-temporal representations that is then decoded for image synthesis. For this, we use a direct training algorithm that computes loss on the membrane potential of the output layer and back-propagates it by using a sigmoid approximation of the neuron's activation function to enable differentiability. The spiking autoencoders are benchmarked on MNIST and Fashion-MNIST and achieve very low reconstruction loss, comparable to ANNs. Then, spiking autoencoders are trained to learn meaningful spatio-temporal representations of the data, across the two modalities - audio and visual. We synthesize images from audio in a spike-based environment by first generating, and then utilizing such shared multi-modal spatio-temporal representations. Our audio to image synthesis model is tested on the task of converting TI-46 digits audio samples to MNIST images. We are able to synthesize images with high fidelity and the model achieves competitive performance against ANNs.

Knowledge distillation is effective for producing small high-performance neural networks for classification, but these small networks are vulnerable to adversarial attacks. We first study how robustness transfers from robust teacher to student network during knowledge distillation. We find that a large amount of robustness may be inherited by the student even when distilled on only clean images. Second, we introduce Adversarially Robust Distillation (ARD) for distilling robustness onto small student networks. ARD is an analogue of adversarial training but for distillation. In addition to producing small models with high test accuracy like conventional distillation, ARD also passes the superior robustness of large networks onto the student. In our experiments, we find that ARD student models decisively outperform adversarially trained networks of identical architecture on robust accuracy. Finally, we adapt recent fast adversarial training methods to ARD for accelerated robust distillation.

May 17, 2019 -- Microsoft Corp. and global education provider General Assembly (GA) on Friday announced a partnership to close skills gaps in the rapidly growing fields of artificial intelligence (AI), cloud and data engineering, machine learning, data science, and more. This initiative will create standards and credentials for AI skills, upskill and reskill 15,000 workers by 2022, and create a pool of AI talent for the global workforce. Technologies like AI are creating demand for new worker skills and competencies: According to the World Economic Forum, up to 133 million new roles could be created by 2022 as a result of the new division of labor between humans, machines and algorithms. To address this challenge, Microsoft and GA will power 2,000 job transitions for workers into AI and machine learning roles in year one and will train an additional 13,000 workers with AI-related skills across sectors in the next three years. "Artificial intelligence is driving the greatest disruption to our global economy since industrialization, and Microsoft is an amazing partner as we develop solutions to empower companies and workers to meet that disruption head on," said Jake Schwartz, CEO and co-founder of GA. "At its core, GA has always been laser-focused on connecting what companies need to the skills that workers obtain, and we are excited to team up with Microsoft to tackle the AI skills gap."

We are in the midst of an exciting time for data policy in the United States. There are few points in history when government's policymakers have been so enthused by the topic of data – and in a promising way. As the Data Coalition's new CEO, I'm excited to lead our organizations and members into this new era. Whether you come from the open data, evidence, science, evaluation, statistics, or privacy community, there are many encouraging activities underway inside government to make data more accessible and useful. For those interested in an effective and efficient government that actually meets the needs of the American public, accessibility of information about policies and programs is essential.

The Organisation for Economic Co-operation and Development unveiled the first intergovernmental standard for artificial intelligence policies Wednesday--and the organization's 36 member countries including America have initially signed on along with Argentina, Brazil, Colombia, Costa Rica, Peru and Romania. OECD, an international forum that unites stakeholders from many nations to work together to address challenges of globalization, released "Recommendations of the Council on Artificial Intelligence" to help foster a global policy ecosystem that leverages the evolving technology's benefits, while also protecting human rights and democratic values. OECD's Director of the Science, Technology and Innovation Directorate Andrew Wyckoff told reporters that the principles' creators hope they'll help shape a stable regulatory environment that promotes the tech's positive uses, while withstanding unethical abuses. "AI is what we would call a'general purpose technology.' It's going to change the way we do things in nearly every single sector of the economy--that's part of the reason we give so much importance to its development," he said.

The Defense Advanced Research Projects Agency (DARPA) is funding research that could give a future generation of soldiers the power to control machines and weapons with their minds. The agency said it will fund six organizations through the Next-Generation Nonsurgical Neurotechnology (N3) program who will work to design and build interfaces for application in the U.S. military, that could be worn be soldiers and translate their brain signals into instructions. Those instructions could be used to control swarms of unmanned aerial vehicles, wield cyber defense systems, or facilitate military communications. Soldiers may be able to control vehicles and more by using only their minds under a new initiative from the U.S. Department of Defense. While the feat may sound firmly in the realm of science fiction, according to DARPA it is setting a completion date within four years.

Foreign Minister Taro Kono plans to ask overseas media outlets to write the names of Japanese people with the family name first, as is customary in the Japanese language. If realized, the new policy would mark a major shift in the country's long-running practice for handling Japanese names in foreign languages -- which began in the 19th to early 20th centuries amid the growing influence of Western culture. At a news conference Tuesday, Kono said that Prime Minister Shinzo Abe's name should be written as "Abe Shinzo," in line with other Asian leaders such as Chinese President Xi Jinping and South Korean President Moon Jae-in. Now is the right time to make the change, given that the Reiwa Era has just begun and several major events -- including next month's Group of 20 summit and the 2020 Tokyo Olympics -- are approaching, Kono said. "I plan to ask international media organizations to do this. Domestic media outlets that have English services should consider it, too," he said, citing a report released in 2000 by the education ministry's National Language Council that said it was desirable to write Japanese names with the family name first in all instances.

U.S. Senators Rob Portman (R-OH), Martin Heinrich (D-NM), and Brian Schatz (D-HI) today proposed the Artificial Intelligence Initiative Act, legislation to pump $2.2 billion into federal research and development and create a national AI strategy. The $2.2 billion would be doled out over the course of the next 5 years to federal agencies like the Department of Energy, Department of Commerce's National Institute of Standards and Technology (NIST), and others. The legislation would establish a National AI Coordination Office to lead federal AI efforts, require the National Science Foundation (NSF) to study the effects of AI on society and education, and allocate $40 million a year to NIST to create AI evaluation standards. The bill would also include $20 million a year from 2020-2024 to fund the creation of 5 multidisciplinary AI research centers, with one focused solely on K-12 education. Plans to open national AI centers in the bill closely resembles plans from the 20-year AI research program proposed by the Computing Consortium.