We spoke to Amazon Web Services’ Chief Evangelist, Ian Massingham, and General Manager AI, Matt Wood, and discussed what they are currently doing in the AI space and what they plan to do in the future.
Amazon Web Services (AWS for short), has only been around since April 2006 – 11 years to be exact – but they’ve already made quite a splash in the AI scene. In those 11 years they’ve developed one of the world’s most popular cloud services, which allows customers and businesses to store data securely among many other things, but they’ve also been working on some very advanced artificial intelligence. They’ve only recently ventured into machine learning, yet they’re already developing advanced AI software such as Amazon Lex (which builds the voices for AI assistants, as well as the text chatbots use), Polly (which turns text into lifelike speech) and Rekognition (recognises and analyses images).
When we asked Ian Massingham, AWS’ Chief Evangelist, how they were involved in the AI space, he spoke about how they were looking at what he termed the “foundation level”. He said that their AI was, “intended for people who want to self-serve,” people who wanted to, “carry heavy-lifting themselves, making use of low cost by the hour to build resources to do ML [machine learning]”.
Massinhgam went on to mention how open-source deep learning frameworks like MXNet, “makes it simple for customers to use DL [deep learning] framework on their technology.” This philosophy was very much echoed by AWS’ General Manager, Matt Wood, who said that the rate of adoption of AI will evolve in 2017 through the broader applications of deep learning. He said that using these open-sourced AI platforms, “will be some of the ways to solve the biggest problems in computer science.”
Massingham then went on to talk more specifically about Amazon’s ML service, which was launched two years ago. Much like all of the services Amazon offers, their machine learning platform is very much geared towards the customer. They want as many people as possible to use their software, naturally, but they also want to make that experience as easy as possible for developers. You don’t necessarily need to be an expert coder or know how to develop complicated algorithms to be able to use the platform.
The conversation then moved onto AWS’ other AI-based software, namely Amazon Lex, Rekognition and Polly. Massingham explained how they were, “all powered by DL neural networks.” Polly for instance, can, “turn text into lifelike speech, stream audio. This is where neural networks come into play,” he said. He went on to explain how their AI – which is incorporated into Amazon’s virtual assistant, Alexa – can actually have a conversation with you. This is something that Apple’s Siri cannot do yet. However, Google Assistant can do something similar, and Samsung is promising that their virtual assistant, Bixby, will be the most revolutionary yet.
But back to Amazon Lex. This is the software behind Alexa’s speech and gives it the ability to have a conversation with the user. Massingham explained how you can ask it a question. He suggested, “I’d like to book a flight from London to Las Vegas.” The AI would then answer, “When do you want to travel?” This is really quite incredible because Lex will, as Massingham put it, “know what to ask”. It can then pass this information onto an API with meaning, the API can then interpret the information accordingly, and before you know it, you’ve successfully booked a ticket from London to Las Vegas. Bon voyage.
This is known as NLU (natural language understanding), which is an area that many companies have been battling with getting it right, and thanks to the work done on Lex by AWS, Alexa is now one of the most advanced AI assistants on the market.
We then asked both Massingham and Wood which areas of industry is AWS gaining the most traction in relation to AI. Interestingly, their responses differed slightly. Massingham spoke of customer services, finance, and even education, whereas Wood felt their AI was being used most innovatively in healthcare. Massingham spoke of how one of their customers, Capital One, are, “big users of AWS Lambda,” which is, according to their website, “a compute service that lets you run code without provisioning or managing servers.”
However, one of the most interesting areas that Massingham mentioned was education. He mentioned how they were seeing AI being used in language tuition. He spoke of language tutoring apps such as DuoLingo and another app called Whizz-kidz, which teaches children with learning difficulties how to read and write. He even spoke of how the Royal Institute for the Blind used Amazon Polly to design software that can transcribe text into audio for blind people. Then he discussed how NASA was using Lex to talk to ROV-E on Mars!
With regards to healthcare, Wood discussed how companies like GE Healthcare were using AI to create “medical imaging reports”. He said that AI was, “driving outcomes for patients in care,” and that we are heading ever quicker to what he referred to as, “digitised health”. There are other companies like AliveCor developing AI-based tools to help doctors better detect symptoms that could potentially lead to strokes.
But what sets AWS apart from other competitors in the market? When we asked Massingham and Wood this question they both singled out their long track record of providing these services. “The customer’s needs and requirements are the only things that we look at,” said Massingham. He went on to discuss how they offer a, “broad range of different services.” Amazon Lex, Rekognition and Polly would be fine examples of this. He then spoke of ASW Lambda which he said, “enables customers to build elastic ways to build conversational interfaces, rich and full featured applications using services in conjunction with one another.” He finished off by telling us that AWS was, “lucky to have open and honest customers.”
Naturally, such a big company like AWS will be planning for the future, and will probably have a five year plan in mind. So, we asked Massingham and Wood where they see AWS in five years time. “It’s difficult to project a roadmap for this length of time,” admitted Massingham. He continued, “We try to resist the urge to do big roadmaps. Priorities change.” He went on to talk about how AWS is “constantly reprioritising.” However, he did reveal that AWS will offer, “additional services in the future with DL at the core.”
Wood was a lot more to the point when we asked him the same question. “AWS will become the centre of gravity for AI.” This just goes to prove how seriously AWS take artificial intelligence and how confident they are that their software will be at the heart of the fourth industrial revolution.