This is the web version of Eye on A.I., Fortune's weekly newsletter covering artificial intelligenceand business. To get it delivered weekly to your in-box, sign up here. Since Bill McDermott became its CEO in November, business software maker ServiceNow has made a couple of notable A.I.-related acquisitions. In January, the company said it would acquire both Passage AI, which specializes in technology that helps computers understand language, and Loom Systems, a startup that uses machine learning to spot errors in corporate infrastructure for IT staff. Those two acquisitions came on top of another acquisition by ServiceNow in October of machine learning and natural language processing startup Attivio, which took place just weeks before McDermott officially started on the job in November.

Major League Baseball is benching Amazon Web Services, with the league picking Google Cloud as its new data and analytics partner. Under a multiyear pact, Google Cloud becomes MLB's official cloud services and cloud data and analytics partner for business operations, including Statcast, the automated service that analyzes player performance and abilities. In addition, for the 2020 season, MLB will use Google Ad Manager and its Dynamic Ad Insertion feature for its digital ads business for the third year in a row. MLB is migrating its cloud and on-premises systems to Google Cloud, using the internet company's machine learning, analytics, application management, and data and video storage capabilities. There's a co-branding aspect to the deal as well: MLB will promote Google Cloud as powering Statcast (as AWS was).

Over the past few weeks, we made several improvements to our transformers and tokenizers libraries, with the goal of making it way easier to train a new language model from scratch. In this post we'll demo how to train a "small" model (84 M parameters 6 layers, 768 hidden size, 12 attention heads) – that's the same number of layers & heads as DistilBERT – on Esperanto. Esperanto is a constructed language with a goal of being easy to learn. You won't need to understand Esperanto to understand this post, but if you do want to learn it, Duolingo has a nice course with 280k active learners. First, let us find a corpus of text in Esperanto.

By the end of 2019, there were more than 400 mass shootings in the U.S. It is the highest number of mass killings in the recent history. Gun violence in the United States needs no introduction. U. S. officials have been trying for years to find a solution to this problem, but the number of casualties is still increasing. But what if you were able to stop a shooter before he even begin to shoot? One way to do that is through artificial intelligence.

If an oil spill were to hit B.C.'s southern coast, threatening the local orca population, the Department of Fisheries and Oceans (DFO) could respond in a way that wasn't technologically possible just two years ago, says Paul Cottrell. For years the marine mammal co-ordinator counted on a network of 18 hydrophones – underwater listening devices lining much of Vancouver Island – to detect calls of the endangered southern resident killer whales and track their movements in the Salish Sea. But what if artificial intelligence could be harnessed to automatically detect the calls of that one particular subgroup of orcas around the clock? That was the pitch Google's (Nasdaq:GOOG) artificial-intelligence division made to the DFO at a 2018 workshop in Victoria. "The opportunity to work with such cutting-edge individuals and technology was amazing," Cottrell said.

Caption Health, a leading medical AI company, announced that its flagship product, Caption AI, the first AI-guided medical imaging acquisition system, is now available for pre-order by healthcare providers. Caption AI is a transformational new technology that enables healthcare practitioners--even those without prior ultrasound experience--with the ability to perform ultrasound exams quickly and accurately, by providing expert guidance, automated quality assessment, and intelligent interpretation capabilities. Caption AI comes equipped with Caption Guidance software, which uses artificial intelligence to provide real-time guidance and feedback on image quality to enable capture of diagnostic quality images. This announcement follows the recent groundbreaking marketing authorization of Caption Guidance software by the U.S. Food and Drug Administration (FDA). The safety and effectiveness of Caption Guidance was clinically validated in a multi-center prospective pivotal trial at Northwestern Medicine and Minneapolis Heart Institute at Allina Health with registered nurses with no prior ultrasound experience.

The new method of criminals in the cyber world is the sound cloning process using artificial intelligence. Audio copies are used for fraud. Especially, the theft of the top executives of the companies has been cloned and theft has increased. The new fraud method of cybercriminals is frauds using artificially cloned sounds. Experts say that along with voice cloning, their voices are no longer safe.

The Defense Department has adopted a series of ethical principles intended to guide the development and use of artificial intelligence on and off the battlefield, including taking "deliberate steps to minimize unintended bias" and ensuring the ability to "deactivate" systems that aren't behaving as expected. The principles are based on the recommendations of the Defense Innovation Board, which spent 15 months consulting with AI experts in industry, government, and academia, according to a Feb. 24 DOD release. The Joint Artificial Intelligence Center will coordinate the implementation of these principles across the department. "The United States, together with our allies and partners, must accelerate the adoption of AI and lead in its national security applications to maintain our strategic position, prevail on future battlefields, and safeguard the rules-based international order," Defense Secretary Mark Esper said in a release. "AI technology will change much about the battlefield of the future, but nothing will change America's steadfast commitment to responsible and lawful behavior."

In January, what had previously only been pixels made a real-world splash on the catwalks of Paris Fashion Week. The models' futuristic-looking clothes, designed in a collaboration between fashion house Acne Studios and artist and programmer Robbie Barrat, were designed by an artificial intelligence (AI). 'When you design a collection, you have an idea of what a jacket looks like, or a pair of trousers,' says Jonny Johansson, creative director of Acne Studios. 'The computer doesn't know what a jacket is. It tries to learn from the images we gave it, and then creates its own idea.

We consider the problem of searching for an unknown target vertex $t$ in a (possibly edge-weighted) graph. Each \emph{vertex-query} points to a vertex $v$ and the response either admits $v$ is the target or provides any neighbor $s\not=v$ that lies on a shortest path from $v$ to $t$. This model has been introduced for trees by Onak and Parys [FOCS 2006] and for general graphs by Emamjomeh-Zadeh et al. [STOC 2016]. In the latter, the authors provide algorithms for the error-less case and for the independent noise model (where each query independently receives an erroneous answer with known probability $p<1/2$ and a correct one with probability $1-p$). We study this problem in both adversarial errors and independent noise models. First, we show an algorithm that needs $\frac{\log_2 n}{1 - H(r)}$ queries against \emph{adversarial} errors, where adversary is bounded with its rate of errors by a known constant $r<1/2$. Our algorithm is in fact a simplification of previous work, and our refinement lies in invoking amortization argument. We then show that our algorithm coupled with Chernoff bound argument leads to an algorithm for independent noise that is simpler and with a query complexity that is both simpler and asymptotically better to one of Emamjomeh-Zadeh et al. [STOC 2016]. Our approach has a wide range of applications. First, it improves and simplifies Robust Interactive Learning framework proposed by Emamjomeh-Zadeh et al. [NIPS 2017]. Secondly, performing analogous analysis for \emph{edge-queries} (where query to edge $e$ returns its endpoint that is closer to target) we actually recover (as a special case) noisy binary search algorithm that is asymptotically optimal, matching the complexity of Feige et al. [SIAM J. Comput. 1994]. Thirdly, we improve and simplify upon existing algorithm for searching of \emph{unbounded} domains due to Aslam and Dhagat [STOC 1991].