The fact that AI seems to have caught the imagination and become such a hyped phrase, means that the bar has been raised regarding expectations of model accuracy.
“Brexit means Brexit,” so the now infamous quote goes. We hear endlessly about the “back stop”, “red lines” and have all come to be experts in the complicated process of negotiating with the EU.
However, the vast chasms between different interpretations suggest that perhaps our understanding is not as solid as we thought, writes Rich Pugh, co-founder and Chief Data Scientist, Mango Solutions.
In fact, one might say that this lack of clarity around all these turns of phrase has, in fact, driven the entire process to a seemingly impossible impasse. However Brexit is not the first situation where this “emperor’s new clothes” approach to meaning has created more confusion than consensus. In data science, it’s near impossible for business leaders to sort the AI from the analytics.
So what is the jargon of data science, what does it really mean, and why should business leaders care? Let’s take a look at some of the most common buzzwords and clear up the reality behind the hype.
We’ll start with the big one – what even is data science? The term ‘data science’ was first attributed to Professor Jeff Wu – the Coca-Cola Chair in Engineering Statistics at Georgia Institute of Technology – during a talk in 1997. He felt that this title better encapsulated the array of work undertaken by statisticians, which was quickly evolving into a far more multi-facetted role.
Today, ‘data science’ means something very different. The most practical definition we can use is that data science is the proactive use of data and advanced analytics to drive better decision-making. It isn’t some elusive unicorn that will fix all of a business’s problems, but is a practical approach to using information in a smarter, more business-centric way.
From a business perspective, the last part of this definition – “drive better decision-making” – is possibly the most important. If this is ignored, there is a real risk of doing all the costly, sleek stuff and not actually adding much value. As organisations invest more heavily in data science, it’s important that data science delivers – or face a situation in which data science as a phrase becomes associated with expensive initiatives that do not make a meaningful business impact.
Digital transformation; almost everyone seems to be doing it, but current success rates are low. Fewer than a third of organisations have reported that a digital transformation project has improved its performance and sustained these gains. In many ways, digital transformation is the ultimate “Brexit” of the tech world: it means all things to all people, and as a result is proving near impossible to achieve across the board.
Amidst all the misinformation, reports of silver bullet solutions are plentiful, but clarity is in short supply. However, at its core it remains a reasonable strategic priority. Businesses should be looking to invest in the right technology to navigate the era of big data – but cannot expect mythical results off the bat, without sustained careful planning in how to successfully use data and analytics for strategic benefit.
Artificial Intelligence (AI)
Nearly as divisive as the UK’s decision to leave the EU, if you believe everything you read AI is either going to create a utopian vision of the future of society, or it’s going to leave us all unemployed and under the control of a sentient computer army. The truth is actually very different. AI has to do with algorithms that enable the simulation of human intelligence processes by computer systems.
AI has unfortunately attracted negative connotations, largely created by science fiction movie franchises like ‘The Terminator’ and ‘The Matrix, which reinforce the perception that AI presents a threat to humanity. AI is not a robot, nor is it a computer army. What it is, is essentially a set of machine-learning algorithms that are applied to unstructured data that ‘appear’ to have human-like qualities. Back in the 1950s, Minsky and McCarthy, described artificial intelligence as “any task performed by a program or a machine that, if a human carried out the same activity, we would say the human had to apply intelligence to accomplish the task.” These days AI systems exhibit some or all of the behaviours linked with human intelligence, such as planning, learning, reasoning, problem solving, knowledge representation, perception, motion and manipulation.
The fact that AI seems to have caught the imagination and become such a hyped phrase, means that the bar has been raised regarding expectations of model accuracy. Whereas people seem to understand that analytics is an iterative process where we seek to explain outputs and reduce error, the expectation for AI seems to be that it will just work perfectly with little effort and next-to-no error. As such, there’s significant investment in this area often without much thought as to which questions should be asked in order to deliver the type of outcome that will add business value.
Data science helps AI technologies find solutions to problems by connecting similar data for use in the future. Machine learning, the scientific study of algorithms and statistical models used by computer systems to effectively perform a specific task without using explicit instructions, relying on models and inference instead, is the section of AI that works best with data science.
Even with an understanding of each of these terms, business leaders could be forgiven for writing off the entire endeavour of becoming data-driven as something speculative, opaque and without likely results. This is because of a tendency, very much like Brexit, to overgeneralise the importance of the buzzword technology, and underrate the core problem we’re looking to solve.
I would hate to try and suggest what that inherent challenge might be for Brexit, but for applying data science it’s all about narrowing down the focus to a specific question that a business is trying to solve. From this point, business leaders can work towards finding effective, efficient and expedient ways to apply data and analytics to solve a real, tangible problem, and not get caught out by applying buzzwords without basis.