When I had a drink with Andrej Karpathy a couple of weeks ago, we got to talking about where we thought machine learning was going over the next few years. Andrej threw out the phrase "Software 2.0", and I was instantly jealous because it captured the process I see happening every day across hundreds of projects. I held my tongue until he got his blog post out there, but now I want to expand my thoughts on this too. The pattern is that there's an existing software project doing data processing using explicit programming logic, and the team charged with maintaining it find they can replace it with a deep-learning-based solution. I can only point to examples within Alphabet that we've made public, like upgrading search ranking, data center energy usage, language translation, and solving Go, but these aren't rare exceptions internally.

Summary: This is the first in a series about Chatbots. In this first installment we cover the basics including their brief technological history, uses, basic design choices, and where deep learning comes into play. In subsequent articles we'll describe in more detail about how they are actually programmed and best practice dos and don'ts. According to Chatbot.org there are currently 1,331 active chatbots in the world. That's a lot for a technology that didn't even exist two or three years ago.

In a post-competition interview competition's winners noted the value of focusing on feature generation, also called feature engineering. Data scientists spend a significant portion of their time, effort, and creativity working on engineering good features; in contrast, they spend relatively little time running machine learning algorithms. A simple example of an engineered feature would involve subtracting two columns and including this new number as an additional descriptor of your data. In the case of the whales, the winning team represented each sound clip in its spectrogram form and built features based on how well the spectrogram matched some example templates. After that, they then subsequently iterated new features that would help them correctly classify examples that they got wrong through the use of a previous set of features.

From its humble beginnings in the AMPLab at U.C. Berkeley in 2009, Apache Spark has become one of the key big data distributed processing frameworks in the world. Spark can be deployed in a variety of ways, provides native bindings for the Java, Scala, Python, and R programming languages, and supports SQL, streaming data, machine learning, and graph processing. You'll find it used by banks, telecommunications companies, games companies, governments, and all of the major tech giants such as Apple, Facebook, IBM, and Microsoft. Out of the box, Spark can run in a standalone cluster mode that simply requires the Apache Spark framework and a JVM on each machine in your cluster. However, it's more likely you'll want to take advantage of a resource or cluster management system to take care of allocating workers on demand for you.

Regular articles on Artificial Intelligence (AI), Machine Learning and Deep Learning appear in the media. Some commentators use these terms synonymously. However, although AI, machine learning, and deep learning are often closely intertwined, they are based on completely different technologies and have their unique attributes. Artificial Intelligence – sounds quite futuristic or even science fiction, this is because this topic has been appearing in the media for over 60 years. Until recently, however, we lacked the necessary prerequisites to apply the resources required for complex AI algorithms completely.

To get the deep-learning algorithms compressed enough to fit on the RPi 3 using just a few bits, Ofer and his team employed a technique known as sparsification, a technique that shave's off unneeded redundancies. Doing so allowed them to devise an image detection system that could process 20-times faster on limited hardware without losing any accuracy. Still, the team hasn't yet figured out a way to take ultra-sophisticated AI or a deep-neural network and compress it enough to fit on limited, low-powered hardware. Regardless, this is an unprecedented first step in doing so, and we can certainly expect advancements that will get us there sometime in not too distant future.

As you can see, we have a variety of images. They all are in different resolutions. Cats and dogs are in different shapes, positions, colors. They may be sitting, they may not. They may be happy or sad.

Today, a Chinese manufacturer and a venture backed by the Bill & Melinda Gates Foundation will announce plans to commercialize a microscope that uses deep learning algorithms to automatically identify and count malaria parasites in a blood smear within 20 minutes. AI-powered microscopes could speed up diagnosis and standardize detection of malaria at a time when the mosquito-borne disease kills almost half a million people per year. An experimental version of the AI-powered microscope has already shown that it can detect malaria parasites well enough to meet the highest World Health Organization microscopy standard, known as competence level 1. That rating means that it performs on par with well-trained microscopists, although the researchers note that some expert microscopists can still outperform the automated system. That previous research, presented at the International Conference on Computer Vision [pdf] in October, has inspired the Global Good Fund--a partnership between the company Intellectual Ventures and the Bill & Melinda Gates Foundation--and a Chinese microscope manufacturer called Motic to take the next big commercialization step.

In the figure above, two numbers are supposed to be added. Those numbers are stored in two variables, a and b. The two values are flowing through the graph and arrive at the square node, where they are being added. The result of the addition is stored into another variable, c. Actually, a, b and c can be considered as placeholders. Any numbers that are fed into a and b get added and are stored into c. This is exactly how TensorFlow works. The user defines an abstract representation of the model (neural network) through placeholders and variables. Afterwards, the placeholders get "filled" with real data and the actual computations take place.

On a periodic basis, publicly traded companies are required to report fundamentals: financial data such as revenue, operating income, debt, among others. These data points provide some insight into the financial health of a company. Academic research has identified some factors, i.e. computed features of the reported data, that are known through retrospective analysis to outperform the market average. Two popular factors are the book value normalized by market capitalization (book-to-market) and the operating income normalized by the enterprise value (EBIT/EV). In this paper: we first show through simulation that if we could (clairvoyantly) select stocks using factors calculated on future fundamentals (via oracle), then our portfolios would far outperform a standard factor approach. Motivated by this analysis, we train deep neural networks to forecast future fundamentals based on a trailing 5-years window. Quantitative analysis demonstrates a significant improvement in MSE over a naive strategy. Moreover, in retrospective analysis using an industry-grade stock portfolio simulator (backtester), we show an improvement in compounded annual return to 17.1% (MLP) vs 14.4% for a standard factor model.