When we look at the landscape in which AI exists, it’s largely been deployed in the realm of the experts. Those companies with huge data sets of consumer and social behaviour, the Facebooks, the Twitters etc. that have had first mover advantage.
These tech giants had the resources required to build the capabilities and capacity in their teams required to invest in building out the infrastructure to process petabytes of raw data, to stand up server farms to chew through that data to train complex machine learning algorithms and to continually iterate the models as new data arrives.
But how does that scale?
There are only so many people ready, willing or able to be data scientists. I consider myself a reasonably competent software engineer and cloud architect, but to me most of data science is deep mathematical and statistical form of voodoo and/or magic – “Any technology, no matter how primitive, is magic to those who don’t understand it” (perhaps I just don’t understand it, or at least it’s inner workings, yet).
With digital transformation offering the potential to radically enhance almost every organisation on the planet – to reduce costs, improve operational efficiency, to enhance the productivity of their employees and continually improve their products – we’re going to need a way to make its power more accessible.
So maybe it’s OK that I and the legion developers building applications to solve business challenges, streamline processes and seeking to build better products don’t understand the deep inner workings of how the ‘magic’ works, just what’s it for, how to harness it and what it’s limitations are – I don’t need to understand how my car works in order to get from A to B, and so we can seek to package up AI services in a way that makes them more accessible to developers everywhere.
Like much of technology, the answer is in layers of abstraction. Under the hood my computer is doing all manner of complex tasks in binary, allocating blocks on a disk to store individual bits and bytes, but I as a user, or even a developer, have tools and abstractions to make me more productive. I don’t worry about the blocks on the disk, I tell my programme to write these words to disk and it organises them into a file for convenience. Software abstracts away from the details so I can focus on the bigger picture. Sure, it helps if the developer or the user has some knowledge of what’s going on under the hood, but they don’t need to concern themselves (normally) with the nitty gritty details.
And so shall be AI.
In order to drive adoption of AI, to ensure its benefits can be realised cost effectively in a wide spectrum of applications. The ability to abstract away from the details, to reuse pre-built models over your own data, to assemble your own models from pre-built patterns and then consume them as a service will offer a great on-ramp to AI. Advanced scenarios may well require complex, detailed implementations of bespoke models, but as a starter, a way to get initial value, the ability to consume pre-baked models in a turnkey fashion can help unlock the benefit of AI for the masses.
This turnkey consumption model for AI and Machine Learning models is why I love using Azure Cognitive Services – https://azure.microsoft.com/en-gb/services/cognitive-services/ to build demos. With a couple of simple API calls, or even just a few clicks in a browser, I can build a demo that brings facial recognition, speech to text, image classification or sentiment analysis to life. Therefore making proof of concepts easy to achieve and helping to build the business case for investing further in AI to transform your business.
As with so much in technology, the barriers to adoption are often the unknowns or the high cost to try something out, to prove it and build a business case that demonstrates return on investment.
As AI goes mainstream, the tools, patterns and practices to make it available to the masses – to the wider developer community in the form of APIs that can be consumed is custom apps & to end users / power users in an ‘Excel-like’ experience – will be critical to mass adoption, to the democratisation of AI.
 For transparency, I work for Microsoft as a Cloud Solution Architect, so naturally I tend towards Azure Cognitive Services. Take a look at the Vision APIs here – no programming knowledge required with these demos to see powerful AI in action – https://azure.microsoft.com/en-gb/services/cognitive-services/computer-vision/