Artificial intelligence, along with the debates around its merits and applications, has long remained top of mind for technologists and other intellectuals.
However, media reports around everything from TV (seen in Westworld) to this spring’s Met Gala, have brought it to the national spotlight. Tabloid media reporting on Elon Musk and his new paramour’s Met Gala debut focused on how Musk contacted her about a shared appreciation for her “Rococo’s Basilisk” joke. The pun referenced the AI premise Roko’s basilisk, a thought experience that looks at the risks tied to the development of AI and the potential of the intelligent technology going awry.
Which leads us to ask, is this prediction terribly far-fetched? Much has been made of AI and its potential, with its ability to self-learn and predict outcomes through the translation of large pools of data and repeat activity and inputs. But it’s not infallible, and its ultimate success assumes that the world works in a linear manner, which we know it does not.
Many industries are successfully leveraging AI and machine learning, from healthcare to cybersecurity to automotive to financial services. Of these, financial services – where AI is applied across a number of disciplines, from trading and investments to loan/risk assessment, to fraud protection – is most certainly one of the most natural contenders to demonstrate the ultimate power of AI and its ability to replace human processing. With a high volume of mathematical data and transactions from which to extract patterns, extend a myriad of potential scenarios and drive intelligent solutions, AI + financial services seem like a perfect marriage.
But, regardless of industry, human thought and interaction must be inserted in the process to provide this input and ultimately understand what this ultimately means for any market – short term or long term. The truth is, computers are simply not infallible and that’s unlikely to change anytime soon. Let’s look at commodities futures. In order to inform a position, AI can be leveraged to take in millions of data points and tens of thousands of potential outcomes, and then predict long-term outcomes with a fair degree of accuracy. But AI could not have possibly known that this year would see a juxtaposition of both unprecedented draught AND heavy rains in the Midwest and southern plains which would, in turn, impact crop production. Human thought and interaction must be inserted in the process to provide this input and ultimately understand out what that means for commodities, both in the long and the short term.
This is no truer when detecting and preventing fraud, where AI is extraordinarily important, but far from the panacea to solve all of the fraud woes facing the world today. For over a decade, AI and machine learning have been used to improve banks and payment processors’ ability to identify and mitigate all types of fraudulent activity. But we at Kount understand that machines alone are not enough and that not all AI is created equal. This is where unsupervised and supervised machine learning come in to combine the strength of computers with human intuition to help companies make better decisions. This human intuition is known as feature engineering and is where the domain expertise of the unique business is applied to the data. Where the models and algorithms are similar from one vendor to another, feature extraction is where the most exciting work happens within the machine learning process. Feature engineering differentiates the solution and allows the machine learning lifecycle to become more intuitive and continually evolve. This requires insights from fraud experts, best practices and a level of domain expertise that can be incorporated into the data analysis.
Kount’s combined approach enables our systems to instantaneously leverage both current environmental conditions (akin to the aforementioned draught + the flooding) as well as historical insights to help companies assess the relative risk and safety of transactions, fight fraud and, ultimately, boost sales.
The advances made by AI cannot be ignored – in finance, fraud, or even in Page Six – but AI alone is not omnipotent, and it’s important to understand where AI is valuable and where human intervention is necessary. Where artificial intelligence can be the guiding principle, human thought should be the deciding principal.