Amazon wants to put AI, ML in hands of every developer, IT professionals: Navdeep Manaktala
Here’s how Amazon is playing a role in democratizing the Artificial Intelligence in India
Amazon is a big player in the cloud. Its Amazon Web Services (AWS) cloud computing platform is used by a number of small and big enterprises around the world. The company is now looking to push its AWS with a newer approach that leverages Artificial Intelligence (AI) and Machine Learning.
We spoke to Navdeep Manaktala, head of business development for Amazon Internet Services to dive deeper on the company’s efforts to utilise the futuristic technologies for its cloud computing platform. Here are the edited excerpts of the interview.
From “Cloud is the new normal” to “Machine Learning is the new normal”. How has the AWS journey evolved?
Every company in the world has to keep evolving and improving its customer experience and business overall to remain competitive.
In the digital age, businesses should be available through whatever channel is most comfortable and convenient for their customers. In other words, the customer should be able to easily connect via disparate channels, such as online, mobile, call center, Facebook Messenger, WeChat or a chatbot. Machine learning (ML) is one of the ways companies can approach this. It helps create seamless experiences for the end consumers through voice, augmented reality, virtual reality, and connectivity. ML can also create a frictionless customer experience by providing a 360-degree view of your customer, no matter what the channel is. That’s why we say, adopting machine learning is not a question of ‘If’, it’s a question of ‘Why not yet’?
Amazon has a long history of over 20 years in Artificial Intelligence (AI) and machine learning, with thousands of experts working on machine learning, across the company. In fact, many of the capabilities of Amazon are driven by machine learning. The Amazon retail recommendations engine is driven by ML, the pick pass that optimizes robotic picking routes in our fulfillment centers are driven by ML, our supply chain, forecasting, and capacity planning are informed by ML algorithms. Amazon Alexa is fueled by natural language understanding and short-form automatic speech recognition. Or, if you look at what we’re doing with the revolutionary concept of being able to walk out of a store without having to go through a checkout line with Amazon Go, those are also informed and fueled by machine learning and deep learning.
The AWS mission is to put machine learning into the hands of every developer, data scientist, and IT professional. Our approach is to create multiple layers of ML services. The top layer provides ML models for a developer, that are fully managed, fully trained, and continuously improving, without them having the need to actually build any of that. They can simply use SDKs or make an API call, and get access to natural language understanding, automatic speech recognition, services that allow them to build chatbot functionalities, facial recognition, and more. The second layer offers platform-specific services. This is ideal if a company wants to offer something specific to their own company or user base. This layer introduces services such as Amazon SageMaker, with the ability to train, build and deploy machine learning models, using your own, or inbuilt algorithms provided in the service. The third layer is about enabling those data scientists, using any of the most popular frameworks such as PyTorch, Scikit, TensorFlow, MXNet, Microsoft CNTK, or more; made available through the platform or through services such as Amazon SageMaker. You can submit your model, have it in any way and train it on your data. You don’t need to worry about how the training works, or how the infrastructure gets optimized.
Share a few use case scenarios where companies have adopted the AWS AI/ML solutions instead of the traditional methods.
Some AWS customers, like Netflix, Pinterest, Airbnb, GE, and Wolfram Alpha, have developed amazing experiences for their customers using AWS’ AI/ML solutions.
Here in India, Haptik, India’s first personal-assistant app, which is a chatbot service, has been using the AWS solutions. One popular feature of Haptik is its reminder service. In the early days, reminders were sent out as notifications, which their users could miss by error, or if they had too many notifications to read through. Realizing this, Haptik chose AWS Polly to convert text to speech, which means their customers now get a call to alert them about the reminder that they’ve set on the app.
Another AWS customer is Butterfleye, a cordless security camera for businesses and homes combines activity-based recording, facial recognition, and military-grade technology to decide when to record and when to arm and disarm the system based on the user’s GPS location. By combining Butterfleye’s on-camera facial functionality and Amazon Rekognition’s API, the company can now identify and tag millions of faces accurately.
FM Wakayama is a community FM radio station in Japan’s Wakayama prefecture. Finding it difficult to hire announcers to deliver news and weather reports in the early morning and midnight shifts, they started experimenting with Amazon Polly, using it to deliver text to speech broadcasts five times a day. The station believes that Amazon Polly is getting smarter each day. Wakayama is located in a tsunami zone, so the radio station built a new application called “Da Capo” to deliver information in Japanese and English, reaching out to both the locals and the tourists during an emergency.
AWS has been talking about the democratization of AI. How do you plan on achieving this in India?
AWS has always worked towards democratizing AI and ML and making it available to the entire developer community, not just data scientists and specialists. Our fully-managed services, frameworks, and tools are now being made available to just about everyone, irrespective of their level of technical skills and abilities, including enterprises, startups, developers and data scientists.
If you look at our offerings, they’re broadly classified into three layers in which, companies and organizations can ‘dock’ their business. The topmost layer consists of the API-driven ML services that facilitate developers to use a diverse set of pre-trained services to easily add intelligence to any application. Developers do not have to create and train models to achieve many generic capabilities offered by these services. The middle layer is for customers with differentiated data sets, who want to build custom models. We provide them with managed capabilities that enable them to effortlessly build, train, and deploy machine learning models with high-performance ML algorithms. The last layer consists of the AI engines, a collection of open source and deep learning frameworks for data scientists who want to build cutting-edge intelligent systems.
For AI, you need volumes of data to be able to train your models. This becomes a challenge for regular developers as they typically don’t have access to many data sets. They also won’t have access to massive computational power. That’s where AWS comes in. We offer the required computational power on our cloud platform, which includes several kinds of GPUs. We also provide a repository of public data sets that can be seamlessly integrated into most of the AWS Cloud-based applications. These datasets will not be charged to the community and, as with all AWS services, you pay only for the computing and storage you use for your own applications.
If you look at all that we’re doing here, we’re trying to democratize AI by making sure that these services are readily available to every data scientist and developer.
What are the AI/ML adoption trends w.r.t. startups and enterprises? What are their expectations from AWS?
While the incentive of lowering the costs of infrastructure has often been the catalyst for enterprises to take action and deploy new applications to the cloud, or to migrate as many existing applications as possible, it’s typically not the most important driver. The main drivers are actually agility and innovation, and the opportunity for a business to adopt new services such as Machine Learning, and to keep iterating - and the cloud enables this in a very significant way.
Cerner, a leading global supplier of health information technology solutions, uses AWS and big data to gain actionable, real-time insights, simplifying healthcare delivery while reducing costs for payers, providers, and patients. They chose AWS for its global reach and breadth of services, including machine learning and artificial intelligence.
PolicyBazaar uses Amazon Polly to convert text to speech in their interactive voice response system (IVR). It made a significant difference to their processes and efficiency by enabling them to shift from sharing a ‘one-size-fits-all’, pre-defined response to customer queries, to generating a voice call that resolves queries from customers in a personalized way.
Another startup, Shaadi.com, has become one of the world’s biggest matrimonial services. The company uses Amazon Rekognition to recognize photos uploaded by its users. This service makes it easy to add image and video analysis to applications that can identify people, objects, text, scenes, and activities, as well as detect any inappropriate content. Shaadi.com is now able to detect, analyze, and compare faces for a wide variety of user verification, as part of its business.
Ixigo, India’s leading travel mobile marketplace, has reduced their research infrastructure cost by 30% while almost halving their time to market, by building and testing cutting-edge AI/ML models. These are all examples of how startups are innovating with AI/ML services right now. There’s no doubt, that we’ll see more exciting ways that both enterprises and startups innovate.
To help both our enterprise and startup customers become more agile, AWS is continually expanding its services to support all kinds of cloud workloads, and we now offer over 125 services that range from compute, storage, networking, database, analytics, application services, developer, mobile, security, hybrid, and enterprise applications; and of course IoT, AI/ML, and augmented or virtual reality.
Most applications that are being used today, will have some form of AI/ML infused in them. If you look at AI usage, it’s still in its early days. But, nearly every enterprise and startup is interested in it. The vast majority of machine learning done in the cloud today is running on AWS.
- The role of digitization & integration across industrial IoT
- How emerging technologies are creating a race for insights
- The rise of Artificial Intelligence and impending takeover
- Opinion | Beware the dark side of Artificial Intelligence
- Cognitive systems and their role in enabling value creation within an enterprise