As we continue our #DigitalDecember journey, we asked our newly appointed tech-lead, David Walters to delve into the realms of Artificial Intelligence – what it involves, and what it truly means…
It is hard these days to read a business article without a reference to Artificial Intelligence.
It seems that the adoption of AI is everywhere and that unless a business has invested in Ph.D. level mathematicians, you are going to be left behind.
However, as a recent report in the Financial Times suggested only around 5% of start-ups claiming to be an AI company could truly be classified as an Artificial Intelligence business.So how do you define AI? Do you have to have built a robot with cognitive behaviour to be considered ‘true’ AI or is something else happening in the data world? A recommendation engine in Netflix, or a personalised shopping experience in Amazon, is using artificial intelligence to predict purchasing preferences and to serve content driven by machine learning algorithms built by skilled data scientists. This type of AI is often described as “narrow AI” as opposed to “wide AI” in applications such as driverless cars.
Building these types of AI services is becoming increasingly accessible. The advent of cloud computing, open source algorithms and the availability of commercially orientated data scientists has “democratised” AI for commerce and replaced most of the West Coast USA black box engines for a fraction of the price. The challenge is no longer the tech but the people.
Attracting and retaining commercially orientated data scientists is a real art. There are lots of academic data scientists available through the universities but entry into the commercial world can often be a big culture shift – imagine pivoting from a thesis of how to solve world hunger to selling cucumbers on line! Retention is even more of a challenge. Data Scientists like to work in teams; they are constantly collaborating with their peers and their team leaders will need to be able to talk the same language. The rise of the Data Science as a Service (DSaaS) model is a great entry point for trialling AI strategies – effectively outsourcing the talent problem and allowing businesses to get going quick with their AI journey while reducing hiring and tech selection risks.
DSaaS teams typically support narrow AI with common projects including recommendation engines, price optimisation, churn prediction and demand forecasting. However, at this point we are venturing into the confusing world of what’s the difference between AI and Statistics?
To answer this, lets have a quick definition of the 3 main areas that make up big data; Data Science, Machine Learning and Artificial Intelligence – this is how I think about it :-
Data Science: This is the organisation and analysis of large data sets using mathematical formulas and algorithms to develop statistically relevant outcomes. A data scientist will spend most of their time developing and testing different algorithms to prove out a business hypothesis
Machine Learning: Machine Learning is the next step in the process and will take the data models created in the data science phase to create predictions on data sets that are constantly self-learning through iterative results from the applications.
Artificial Intelligence: is the deployment of the data science and machine learning into production. By placing the machine learning algorithms into the cloud and pointing at a website user experience can be predicted and influenced.
It’s important to note that businesses can actually create real value by deploying Data Science and Machine Learning in isolation of AI. These two technologies can discover insights about a business within a few hours that can often take human beings years to discover. This is perhaps where the question around what is true AI has come from? Adopting data science under the AI banner is a common and understandable approach. Artificial Intelligence has become part of our business vocabulary and it is sometimes easier to use AI as a catch all term for clever use of big data.
From my experience the biggest challenge has always been around how to get an organisation to truly recognise the value in their data sets. Data is the oil in smart organisations, can drive IP and ultimately business valuation when put at the heart of an organisation.