If you are looking for a new movie or television series to watch, all you need to do is check out the “Recommended” list on Netflix or iFlix. The same goes for exploring new songs on Spotify, as its “Discover Weekly” playlist, more often than not, hits the right notes to suit your personality.
We are living in a time when artificial intelligence (AI) and machine learning (ML) are enabling companies to deliver personalised experiences and recommendations to their customers. The technologies study customers’ preferences and run them through some complicated algorithms and correlations to come up with tailored recommendations.
Recognising this trend, Amazon Web Services (AWS) head of emerging technologies Olivier Klein says companies that do not make use of their data to venture into ML will simply get left behind. “ML is not optional anymore because your customer experience needs to be good. If you use an app on your phone, you expect it to treat you as an individual and when you are listening to music, you expect the next thing to play to be something you like. All this is only possible if we understand our customers through data analytics, ML and AI.”
ML is a subset of AI which uses data to “train” machines to understand personal preferences.
He was speaking at a media roundtable on emerging technologies at the recent AWS Summit Singapore 2019.
Using the media entertainment industry as an example, Klein says technology can be taken much further to provide a better user experience. Currently, services such as Netflix and iFlix have an initial recommendation engine that asks the user to select shows they have watched so it can initialise the recommendation feature.
This allows prediction based on simple factors such as genres and actors. “What over-the-top platforms can do now is measure when a user starts and stops watching or when they rewind and fast-forward scenes. When this happens, the software can tell where you lost interest and when your interest is piqued. With a computer vision model such as Amazon Rekognition, it can understand specifically what you do not like — for example, gory scenes or those with a lot of explosions,” says Klein.
“Predictions are not just a thing where you have a prediction engine and you are done. It is about how you keep refining, going deeper into the technology and understanding specifically what the user’s preferences are. The better you get at it, the better the customer experience becomes.”
He believes that as technology advances, it will become more human-centric, where computer systems end up adapting to human needs. Not too long ago, the ability to talk to computers with our natural voices sounded like an impossible feat. But now, we have the Amazon Echo device with Alexa voice services, which people take for granted.
“We will see a stronger evolution in this space, where we will build more natural interfaces. I think customers will start to expect quick responses to enquiries that are tailor-made to their interests,” says Klein.
“This is the overall digital transformation exercise that we see happening, even in very large traditional enterprises in Malaysia. We cannot treat IT as that back-end system that we have to power the company. It is something that needs to be at the forefront because it changes the customer experience. You can create things that were not possible before.”
The AI and ML space is not new, but it is extremely complex for someone who is not a data scientist. Klein says AWS wants to change that by asking how it can put these technologies into the hands of every IT professional out there.
To do this, it offers services that seamlessly integrate with a company’s back-end system. This is made up of services fully-trained by AWS, such as Amazon Rekognition, and all businesses need to do is integrate these with their platforms through an application programming interface (API) call or micro-service. “Now, I get capabilities such as image and video material analysis and face and object recognition without the need to understand computer vision models, train them and have data,” says Klein.
An API is a set of functions and procedures that allow applications to access features or data of an operating system, application or service. Microservices, or microservice architecture, is an approach to application development in which a large application is built as a suite of modular components or services.
The next layer of service is for those whose needs are not met with AWS’ fully-trained programmes and would like to build their own models. Klein says an important service in this category is the Amazon Sagemaker, which allows businesses to build their own ML models with limited expertise but using their data.
“With Amazon Sagemaker, you can train your own ML model with built-in algorithms that we provide you. So, we take your data, but you train your ML model with an algorithm that Amazon provides or with one of our partners through the AWS Marketplace,” he says.
“Let’s say you want a fraud detection algorithm. You can go to the AWS Marketplace to look for one and train it with your data in your account. The key point is that it remains your data and your model, but you get the capabilities of training ML models without the need to understand how the models work.”
AI and ML in Asean
Asean has millions of small and medium enterprises (SMEs). AWS senior manager of developer relations Shaun Ray tells Enterprise that SMEs in the region are being disrupted by the technologies around them because of their customers.
“They do not necessarily need to go out and think about technology because their customers are demanding it, whether they are business-to-business or business-to-consumer. For example in the Philippines, people spend 10 hours a day on their phones — the highest usage in the world. And they have 30% mobile user growth per year, which is crazy. That means businesses with any kind of consumer-facing technology have been forced to go mobile and embrace technology because their consumers demanded it,” he says.
“Meanwhile, traditional businesses such as Astro in Malaysia are getting competition from other digital places. So, the company had to rethink its whole business model. It had to change its delivery to mobile and move away from traditional cable television that it is used to. We have been working with Astro for a number of years and saw how it became a platform serving small businesses instead of consumers.”
Klein says his conversations with companies in Malaysia have been on data analytics. “Typically they ask, ‘How can I build a good data analytics platform and drive it effectively?’ This conversation eventually leads to ML. But that is like the endgame for most companies.
“It starts with, ‘How do I do away with having a big data warehouse sitting in my data centre to having something lean and cost-effective that can scale out?’ This leads them to serverless models and the cloud.”
Companies across Malaysia have been gradually adopting technologies. What makes the difference is the level of experimental culture within companies because generally, those that experiment more tend to adopt technologies more quickly, says Klein.
“We provide you with a plethora of services that make ML available to anyone who want to use it. But if you do not encourage your own employees to experiment with some of these services, they are not going to adopt it.
“The adoption barrier to using ML keeps massively dropping and it is getting easier to use it with AWS as we make it accessible to pretty much anyone, which was traditionally very hard to do. But if you do not start experimenting with these technologies, you will get stuck.”
Microservice and serverless solutions
A lot of companies that are looking to innovate have a challenge when it comes to figuring out how to speed up the process. Klein says microservices allow these companies to experiment with technology.
Microservices enable companies to split up their business into smaller, more specific functions so that each can be innovated separately. For example, if you look at a website such as Amazon.com, the shopping cart is one microservice. The login function, registration and fulfilment centre are other microservices.
“The benefit of this strategy is that I can have small and independent teams that iterate on their own with their own software and frameworks, which allow for faster innovation. From an architectural pattern, you will build small pieces of software and tie them together with APIs. This allows the independent teams to work on the different microservices. Later on, we will train them to sync using a communication API layer,” says Klein.
Going serverless is another aspect that SMEs need to consider as it not only helps with the development and innovation of microservices but also allows companies to better focus on software building and core matters of the business.
“In an ideal world, cloud computing allows me to bring things up and shut them down with a few commands automatically and only pay for what I use. But what does that really mean? It is an operational model change. The way we develop software has changed. We build microservices. But the way we run that has changed because we run them serverless,” says Klein.
Serverless is the native architecture of the cloud that enables businesses to shift their operational responsibilities to AWS, increasing agility and innovation. Being serverless allows businesses to build and run applications and services without thinking about servers.
Klein says there should not be any concerns about managing infrastructure when going serverless as companies are allowed to scale and focus on things they require. “If you have a lot of traffic on your online platform, it should automatically scale out. If you have a lot of requests in your fulfilment centre, it should automatically scale out.
“You should only pay for what you use, based on the execution time or storage you use. Hence, you do not need to overprovision. Before this, I had to think what my return of interest would be for the next five years and how much storage I would need — a big headache. In a serverless model, it automatically scales and I only pay for what I use when I use it.”
There is a misconception about integrating with the cloud as some companies think they need to move their entire application into the cloud, which can be quite a heavy undertaking. Klein says what most companies do is gradually move components that would suit the cloud. For example, a company may choose to run its website serverless because it has a lot of traffic while keeping other components on its on-premise infrastructure.
“It is about having seamless integration. So, you need to look at what you currently have, determine where the value of the cloud matters most and start there. Then, you can gradually move bits and pieces into that new model,” says Klein.
So far, most digital native companies comprise larger-scale businesses. Thus, it may be intimidating for some SMEs to take on these technological advancements. So, AWS is trying to level the playing field.
AWS global advisory senior consultant Mario Thomas says SMEs have access to the same tools as the larger enterprises. “I think SMEs have a really exciting time ahead of them in terms of the cloud. I think we enable so much possibility for all of them.
“If you had been an SME 10 years ago, there would have been technologies that only the larger enterprises had access to. So, the democratisation of access to tech is one of the key drivers of cloud computing and a big benefit for SMEs.
“Look at Netflix, Uber and Lyft, all of which are unicorn start-ups. They would not have existed had they not had access to this infrastructure on the basis of paying for what they use when they are using it.”
He says companies that use cloud computing see a significant cost savings and optimisation as the overheads associated with managing and maintaining server infrastructure have been eliminated. Looking at the typical life cycle of an SME, it takes 8 to 18 weeks to procure, order, rack and stack a server, which may be too long for some of them.
“SMEs need to move at a different pace from larger enterprises as they tend to be fast-moving and agile in nature. Elasticity is another key benefit. So, the ability to scale up when you need to and getting rid of what you do not need means you do not have assets sitting on a rack and gathering dust,” says Thomas.