We hear a lot about deep learning algorithms and their applications on very large data sets. This interview delves into a company and its customer base that works in that area. We provide a specific segment of artificial intelligence called machine super intelligence, which is the intersection of advanced deep learning and high-performance computing. Our customers are some of the largest customers in the world. We sit on top of their big data stack and we track patterns.

Artificial Intelligence is still a nascent technology; much of the groundbreaking work moving the industry forward is done inside AI research labs. It's often from those labs that open source projects are started. Institutes like Open AI, NASA's JPL, Google Deepmind, MIT CSAIL, BAIR, The Turing Institute, and Max Planck -- to name just a handful -- are presenting at ODSC in 2019, helping us bring our community to the leading edge of AI. To learn more about the labs' role at ODSC, visit ODSC West. Since our next conference is in San Francisco, we're looking west at a few exciting research labs in the area that are participating in ODSC this year.

You might ask: why 3 convolutional layers? Or why 2 convolutional blocks?We included these numbers as hyperparameters in a hyperparameter search. Together with parameters such as: number of dense layers, dropout level, batch normalization and the number of convolutional filters we did a randomized search to find the optimal combination of hyperparameters.

Raise your hand if you've been caught in the confusion of differentiating artificial intelligence (AI) vs machine learning (ML) vs deep learning (DL)… Bring down your hand, buddy, we can't see it! Although the three terminologies are usually used interchangeably, they do not quite refer to the same things. Andrey Bulezyuk, who is a German-based computer expert and has more than five years of experience in teaching people how artificial intelligence systems work, says that "practitioners in this field can clearly articulate the differences between the three closely-related terms." Therefore, is there a difference between artificial intelligence, machine learning, and deep learning? As you can see on the above image of three concentric circles, DL is a subset of ML, which is also a subset of AI.

There are several outstanding training sessions and tutorials that cover PyTorch, TensorFlow, and other important topics at the O'Reilly Artificial Intelligence conference in San Jose, September 9-12, 2019. Early price ends August 2. In a recent survey--AI Adoption in the Enterprise, which drew more than 1,300 respondents--we found significant usage of several machine learning (ML) libraries and frameworks. About half indicated they used TensorFlow or scikit-learn, and a third reported they were using PyTorch or Keras. I recently attended an interesting RISELab presentation delivered by Caroline Lemieux describing recent work on AutoPandas and automation tools that rely on program synthesis.

As machine learning continues to be employed in business processes, AI has found its way into the heart of businesses by enabling machines to perform tasks the way humans do. This is apart from the fact that AI-enabled solutions are able to adjust to situations while learning from experience. Artificial intelligence software employs advanced technologies such as natural language processing and deep learning, making them trainable in carrying out tasks. Such systems accomplish these by being able to recognize patterns in high volumes of data.

In the last few years, deep learning has solved seemingly intractable problems, boosting the hope to find approximate solutions to problems that now are considered unsolvable. Earthquake prediction, the Grail of Seismology, is, in this context of continuous exciting discoveries, an obvious choice for deep learning exploration. The artificial neural network (ANN) (shallow or deep) is rapidly rising as one of the most powerful go-to techniques not only in data science [LeCun et al., 2015; Jordan and Mitchell, 2016] but also for solving hard and intractable problems of Physics (e.g., many-body problem [Carleo and Troyer, 2017], chaotic systems [Pathak et al., 2018], high-dimensional partial differential equations [Han et al., 2018]). This is justified by the superior performance of ANNs in discovering complex patterns in very large datasets with the advantage of not requiring feature extraction or engineering, as data can be used directly to train the network with potentially great results. It comes as no surprise that machine learning at large -- including ANNs -- has become popular in Statistical Seismology [Kong et al., 2019] and gives fresh hope for earthquake prediction [Rouet-Leduc et al., 2017; DeVries et al., 2018].

If you are, like me, confused by why deep neural networks can generalize to out-of-sample data points without drastic overfitting, keep on reading. If you are like me, entering into the field of deep learning with experience in traditional machine learning, you may often ponder over this question: Since a typical deep neural network has so many parameters and training error can easily be perfect, it should surely suffer from substantial overfitting. How could it be ever generalized to out-of-sample data points? The effort in understanding why deep neural networks can generalize somehow reminds me of this interesting paper on System Biology -- "Can a biologist fix a radio?" (Lazebnik, 2002). If a biologist intends to fix a radio machine like how she works on a biological system, life could be hard. Because the full mechanism of the radio system is not revealed, poking small local functionalities might give some hints but it can hardly present all the interactions within the system, let alone the entire working flow. No matter whether you think it is relevant to DL, it is a very fun read.

Therefore, through learning the Deep Learning tools, we will be able to invent new products and applications. Even if we are not concerned with the development of the Computer Vision systems, creative neural network architectures from this research field could inspire us to create many other methods in the fields of speech recognition, text processing or audio processing.

In 2012, Geoffrey Hinton's research team used only two NVIDIA GPUs to train AlexNet, the revolutionary network architecture that handily won the ImageNet Large Scale Visual Recognition Challenge. It probably never occurred to these groundbreaking researchers that just seven years later, a new team of researchers would use almost 10,000 times more GPUs to train their AI model. A research team from NVIDIA, Oak Ridge National Laboratory (ORNL), and Uber has introduced new techniques that enabled them to train a fully convolutional neural network on the world's fastest supercomputer, Summit, with up to 27,600 NVIDIA GPUs. They managed to achieve an impressive, near-linear scaling of 0.93 on distributed training and produce a model capable of atomically-accurate reconstruction of materials -- a longstanding scientific problem involving materials imaging. In June 2018 the US Department of Energy's Oak Ridge National Laboratory in Tennessee unveiled the world's fastest supercomputer Summit, whosecomputing power reaches 200 petaflops.