Compelling data reveal a discouraging truth about growth today. There has been a marked decline in the ability of traditional levers of production--capital investment and labor--to propel economic growth. Yet, the numbers tell only part of the story. Artificial intelligence (AI) is a new factor of production and has the potential to introduce new sources of growth, changing how work is done and reinforcing the role of people to drive growth in business. Accenture research on the impact of AI in 12 developed economies reveals that AI could double annual economic growth rates in 2035 by changing the nature of work and creating a new relationship between man and machine.

In Part I of this two-part series, Fred Giasson and I looked back over a decade of working within the semantic Web and found it partially successful but really the wrong question moving forward. The inadequacies of the semantic Web to date reside in its lack of attention to practical data interoperability across organizational or community boundaries. An emphasis on linked data has created an illusion that questions of data integration are being effectively addressed. Linked data is hard to publish and not the only useful form for consuming data; linked data quality is often unreliable; the linking predicates for relating disparate data sources to one another may be inadequate or wrong; and, there are no reference groundings for relating data values across datasets. Neither the semantic Web nor linked data has developed the practices, tooling or experience to actually interoperate data across the Web.

Over the years, society has become more dependent on digital technologies. Today, nearly every person, business, and government agency uses the internet to transmit and store data. As a result of that dependence, there is no shortage of hackers who try to access that data. We see this at every level. Celebrities have had their phones hacked and their personal photographs stolen and dispersed online.

When millions of people across the world tuned in to watch DeepMind's machine beat the human Go world champion Lee Sedol, they also witnessed a historic victory for open-source. DeepMind used a scientific computing framework called Torch extensively in the development and execution of AlphaGo's neural networks. Torch was first released back in 2002 under a BSD open-source license with algorithms that are still commonly used by data scientists such as multi-layer perceptrons, support vector machines and K-nearest neighbours. Torch also supported ensembles -- a popular technique that combines the output of multiple algorithms, usually with a weighted average. It's not just open-source software that contributed to the growth of machine learning.

Consequently, there is an equivalence between parameter averaging and update-based data parallelism, when parameters are updated synchronously (this last part is key). This equivalence also holds for multiple averaging steps and other updaters (not just simple SGD). Update-based data parallelism becomes more interesting (and arguably more useful) when we relax the synchronous update requirement. That is, by allowing the updates Wi,j to be applied to the parameter vector as soon as they are computed (instead of waiting for N 1 iterations by all workers), we obtain asynchronous stochastic gradient descent algorithm. These benefits are not without cost, however. By introducing asynchronous updates to the parameter vector, we introduce a new problem, known as the stale gradient problem.

New times are marked by the sign of the digital age, globalization and the huge amount of mass data generated daily. Big Data is rigged to great challenges and better opportunities. Beyond the famous 5 Vs that characterize it (volume, velocity, variety, veracity and value), Big Data have great possibilities in the most unimaginable fields. And it does, especially, because new technologies have emerged to respond, with unprecedented efficiency, to the needs of storage and analysis of big data. There are many technologies and concepts that are part of this universe of big data, whose growth is unstoppable, as the Internet of Things (IoT), data exchange machine to machine (M2M), the increasingly complex environment of IT or predictive machine learning. The Big Data challenges, in effect, require capable approaches and systems to collect, store, make efficient searches and, finally, carry out analysis, whose results could be conveniently displayed.

This year the excitement is about Salesforce's introduction of Einstein – artificial intelligence (AI) built into the core of the Salesforce Platform. With Einstein, Salesforce is adding AI capabilities to Sales, Service and Marketing Clouds. This is a very significant new capability, allowing learning from all the valuable data that exists in the Salesforce Clouds. And learning is the most critical part of this offering because it leads to action recommendations, outcome predictions and if requested automation, which is what every organization is looking for today to get closer to their customers. Informatica is a long-standing partner with Salesforce and has teamed with us to help 5,000 customers on their journey to the Cloud.

A brief and concise tutorial on how to visualize different aspects such as the loss of your neural network using tensorboard. We are going to work with a fully-connected neural network using the MNIST dataset. I'm going to use the network I have introduced in an earlier post. It achieves on the test-set an accuracy of 90%. This is not bad but we have no clue what is actually going on or how our model looks like.

Philip Koopman, professor of Carnegie Mellon Univ., believes the biggest hole in a Federal Automated Policy published late Sept. is in the regulators' failure to tangle head-on with fundamental difficulties in testing Machine Learning -- a problem already known to the scientific/engineering community. Representativeness of data Carmakers are building a fake city, for example, in Michigan to test autonomous vehicles. What's important, though, is whether the test data represents real-world driving conditions? A highly autonomous vehicle is designed to operate only in a certain designated area such as "driving only in downtown Pittsburgh." In DoT lingo, this concept is the "Operational Design Domain."

Consider a case where you search for a gadget in Google Image search. Your screen will be deluged with numerous images of it drawn from infinite quarters across world wide web. The logic behind surfacing these images is quite simple. If the image uploaded has a name, title and'alt' tag mentioned, that image will get picked by google crawlers. Another possibility is, even though the image might not have any such parameters, but if the web page content(where images reside) matches the search term, the image in that web page shows up. This works fluently if it was not been abused by black strategies.