In November, Sentient Technologies announced a collaboration with, wherein the site uses a 'Visual Filter' built on deep learning - the world's first AI-powered shopping experience. At the other end of the scale, Sentient Technologies is working on an intelligent way to respond to sepsis infections, with an AI nurse - a great example of how deep learning is rapidly transforming and improving areas many industries, from retail and finance, to healthcare and manufacturing.

The new technology being deployed on involves a 'Visual Filter' that has a deep understanding of the what is available in the catalog and rapidly learns and adapts to shoppers' preferences. It also provides instant recommendations of available products, similar to the way a personal shopper in a physical store would help consumers find shoes they love - creating an enjoyable and efficient experience for customers. It's easy to see how this kind of technology will be valuable, as customers are overwhelmed by an ever-growing online market, when introducing AI can cut the time and effort involved in browsing products.

At the RE•WORK Deep Learning Summit in San Francisco this January, Nigel Duffy, CTO of Sentient Technologies, will discuss the significant progress being made in the areas of visual intelligence and deep learning, and visual learning technologies that aid decision-making by observing, interpreting and evaluating users' interactions with visual content.

We spoke with Nigel ahead of the Deep Learning Summit next month to hear more about Sentient Technologies and his thoughts on the deep learning field.

What are the key factors that have enabled recent advancements in deep learning?
There’s a widely held belief that access to more compute, and larger quantities of data have been really important factors. This is clearly the case, first with large distributed systems like Google Brain, and later with fast neural network implementations on GPUs.

I think an under-appreciated factor is the existence of large-scale benchmarks. Large-scale benchmarks such as imagenet provide really important testbeds for the development of new techniques in academia and make it possible to demonstrate progress. Without these benchmarks academics had often been stuck working on smaller scale problems where the benefits of deep learning weren’t apparent.

Context is also important. The content that users generate and consume has become more visual -- there will be 1 trillion photos taken this year. Because of this the ability to analyze and understand images has become much more commercially important.

At Sentient, we have been focused on scaling artificial intelligence (AI) algorithms since our foundation. As the deep learning community moves towards even richer media, e.g., videos, we believe scale will become even more important. Scaling AI can be challenging, especially as you scale past a single GPU, or even a single data center. However, being able to achieve this kind of scale will be essential as we move to more and more complex problems.

What are the main types of problems now being addressed in the deep learning space?
Most deep learning work is focused on what I would call perceptual problems, e.g., understanding images, video, speech, and audio. Intuitively this makes sense as many deep learning approaches implicitly encode modeling biases consistent with perception. For example, these networks often have structures or abstractions that have parallels in human perception, or they have structures that capture biases such as the neighborhood bias, i.e., pixels close to each other are more likely to be related than pixels far from each other.

Increasingly we’re seeing deep learning being used as part of a larger system, e.g., in machine translation. In these cases deep learning is often used to address issues of representation, while language structure is addressed using other machine learning techniques. I believe we will see more and more such hybrid systems. For example, our online shopping experience is backed by a combination of AI approaches including deep learning.

What are the practical applications of your work and what sectors are most likely to be affected?
The online shopping experience hasn’t evolved much since the late 1990’s. As e-commerce competitors expand their catalogs to attract broad audiences, shoppers are presented with a paralyzing array of choices.  When you search for things, you’ll often feel like you’re getting nowhere since items are tagged incorrectly. Visual intelligence is solving this problem.

Recently we partnered with to launch Sentient Aware for e-Commerce. It powers’s Visual Filter experience on The service provides shoppers with instant recommendations of available shoes, similar to the way a personal shopper in a physical store helps people find shoes they love. Our AI works by creating an ongoing, deductive dialogue in real-time with shoppers, which we believe is critical to making the consumer’s journey through online content gratifying. Sentient Aware creates game-like engagement by offering image-based "questions" where every click trains the AI across thousands of attributes (e.g. heel-height, calf length, lace style, shoe shape, textures, etc.) that then informs the shopper’s next choice. As shoppers start to click on shoes, our AI begins to suggest options that it thinks they will like based on their previous clicks. It is as if a private boutique is being curated just for them in real time, out of thousands of items in the warehouse.

We believe this type of deep learning can be applied to many different types of industries, from shoes and apparel to furniture, housewares and automobiles.

What developments can we expect to see in deep learning in the next 5 years?
   To date the Big Data ecosystem has been focused on the collection, management, and curation of large amounts of data. Obviously, there has also been a lot of work on analysis and prediction. Fundamentally though, business users don’t care about any of that. Business users only care about outcomes, i.e., “will this data change the way I behave, will it change the decisions I make”. We believe that these are the key questions to be addressed in the next 5 years. And we believe that AI will be the bridge between data and better decisions.

Obviously, deep learning will play a significant role in that evolution, but it will do so in combination with other AI approaches. Over the next 5 years we will increasingly see hybrid systems where deep learning is used to handle some hard perceptual tasks while other AI and machine learning (ML) techniques are used to address other parts of the problem, e.g., reasoning.

What advancements excite you most in the field?
We’re really excited about the hunger for AI and ML. Increasingly every problem, every project, every business is an AI problem, project or business. The opportunities to use data to make the world more efficient and productive are enormous.

We’re most excited about using AI to make better decisions faster. Exciting new developments like quantified self, the automated home, and smart factories will only be possible with automated decision making.

As individuals, our decision making burden is only increasing. We’re constantly required to make choices and our options have become enormous. However, too many choices can be paralyzing and we believe that AI can make the world simpler by reducing those choices to only the ones we care about most.

At Sentient our overall goal of using AI to make the world simpler, more efficient, and more productive through automated decision making.

Nigel Duffy will be speaking at the RE•WORK Deep Learning Summit in San Francisco, on 28-29 January 2016. Other speakers include Andrew Ng, Baidu; Clement Farabet, Twitter; Naveen Rao, Nervana Systems; Andrej Karpathy, Stanford University; and Oriol Vinyals, Google.

The Deep Learning Summit is taking place alongside the Virtual Assistant Summit.

Early Bird tickets are available until 4 December, for more information visit the event page here.