Building AI: “It’s simpler than you think”, says AWS Chief Evangelist & Worldwide Lead
Posted: 6 October 2017 | By Charlie Moloney
Access AI met Ian Massingham, Amazon Web Services (AWS) Worldwide Lead and Chief Evangelist, yesterday at IP Expo Europe and we talked all things Artificial Intelligence (AI).
Amazon claims that “more AI is built on AWS than anywhere else”. AWS offers services, powered by APIs, which can help users achieve natural language understanding (NLU), text-to-speech, and image recognition.
But who is building AI on AWS? Could anyone do it? Should anyone do it? We asked Ian, and he told us:
AAI: Should everybody, in every industry, be building AI technologies?
Ian Massingham: “I wouldn’t say that they’re essential for every business, but I think every business is starting to see applications now. Situations where they can apply machine learning for competitive advantage, or apply machine learning to improve the customer experience, or even build entirely new products, and I think one of the reasons for that is accessibility.”
AAI: So, do you disagree with those who say that this is a Kodak Moment, and that every company should at least be looking into bringing in AI technology?
IM: “Every company is a very broad spectrum. If I’m a builder, my job is putting up houses, I don’t need a machine learning system.”
AAI: Perhaps a builder could optimise their workflow with an AI system on their phone?
IM: “Perhaps, but as a trader in that environment – am I gonna build the system or am I gonna buy the system? That’s the nuance.”
“I think it’s likely that over time pretty much everybody will end up using systems that have AI and Neural Network features within them, but not everybody will end up building them.”
“It’s going to be incorporated into applications that we use everyday. If you use Pinterest on your phone, and you use their image search, you’re using a neural network, but you never built it as a Pinterest user. You’re using something that’s in TensorFlow, on top of the AWS platform.”
“So, I think everybody will use features in their everyday life that are based upon neural network and AI technology, but not everybody will build them themselves.”
AAI: What would you say to people who think that building this technology would be too complicated for them?
IM: “It’s simpler than you think. You can look at these abstracted services that provide pre-packaged, functionality for you to directly consume like natural language generation, or image recognition, or chatbot engines.”
“Beyond that there are frameworks, and frameworks over time like all software will become more and more accessible as more and more abstractions are built on top of it.”
Experiment with AI and neural networks and see what you can achieve. It’s not as difficult as developers might think it is
“I think over time that will make them even more accessible for developers to use, and there’s already some excellent resources out there for developers who want to get started with MXnet and want to get started with Tensorflow, and if you use AWS to do that then you can do it at low cost without the big up-front investment as well.”
“So, experiment with AI and neural networks and see what you can achieve. It’s not as difficult as developers might think it is.”
AAI: And what would you tell people who are afraid of trying and failing to build an AI solution on AWS?
IM: “We have a culture of innovation within AWS. If you try something and fail it is seen not seen as a negative. It is seen as a positive, because you have tried something. It is really encouraged within the organisation.”
“One of the reasons that AWS was created was to enable a culture where failure is acceptable to be created within Amazon itself at a higher scale, by industrialising the components that we use, reducing the cost, and making them available for media developers.”
“Developers can run experiments, and if those experiments are unsuccessful they can return the resources without capital investment and then they can be reused for the next project.”
“And today an awful lot of our customers put AWS to use in the same way: to run speculative experiments for new services, for new business models, even new derivatives of products that already exist. Test those, and if they’re successful then scale them, but if they’re unsuccessful then minimize the impact of failure and move on to the next idea.”