In his 2017 forecast, Scott Zoldi, analyst for big data analytics, artificial intelligence(AI) and cyber security company, FICO (Fair Isaac Corporation), suggested that we were “just at the beginning of the golden age of analytics, in which the value and contributions of AI, machine learning (ML) and deep learning will only grow as we accept and incorporate these tools into our businesses.”
As it happens, AI, ML and deep learning did not just grow up during 2017… they blew up.
And now as 2018 swiftly kicks off, Zoldi says that the development and use of these technologies will continue to expand and flourish at an incredible pace.
He adds that in so many ways, it’s hard to quantify the expected results of such innovations as there can be many different interpretations of each, but this is where he believes the most significant impact will be felt in terms of artificial intelligence in 2018:
AI will learn how to fight back
In 2018, Defensive AI will be front and centre. We’ve long talked about how machine learning and AI will help companies to drive differentiation in competitive markets. This applies in the criminal world too, where attackers will use malicious AI and ML to circumvent the ones companies have in place. This arms race, in which criminals arm themselves with “adversarial machine learning,” tops American global computer security software company McAfee’s 2018 security forecast.
Earlier this year, Zoldi wrote about FICO’s patent-pending Defensive AI technology, which will usher in systems that will seed their outputs with “faint signatures” to mislead, confuse or identify the attackers learning the AI system’s response.
AI will have to explain itself
Explainable AI (XAI) will begin to answer the need for explanations for decisions based on scores, including those produced by AI and ML systems. We’ve long dealt with this challenge in fraud and credit risk decisioning, but in using ML across multiple industries, there are entire sets of proposed explanatory algorithms which are either right, ineffective or flat wrong. Zoldi looks forward to these advancements as it is imperative for companies to be able to correctly explain the decisioning processes of their AI and ML systems.
AI will augment us
The idea that Artificial Intelligence and humans will play nice together, rather than battling it out in a Terminator scenario, is not new.
In 2018 however, AI will augment much more of the work place. Not just through better software, but through the facilitation of increasingly better versions of ourselves.
Whether it’s drawing the information together for us to be superhuman at investigation, data recall, or improving how we learn new topics, AI will augment our ability to process new information. The question will be as to whether our human brains will atrophy, improve, or simply evolve to the rate and frequency of data.
AI gets operationalised
More than 25 years after Artificial Intelligence/Machine Learning first gained widespread use in fighting payment card fraud, the idea of using AI properly (extensively), still poses a tremendous challenge for many organisations. Those that focus first on how to use, care, and feed their AI systems and machine learning, have been sufficiently rewarded. Far more than those that are enamored with new algorithms or increases in computation complexity, which often require more Cloud and more GPU.
Zoldi believes that in 2018, companies will focus on operationalising AI, particularly in the cloud, to more easily build, refine, deploy and enhance machine learning environments.
Chatbots will get better at understanding and manipulating us
In many aspects of society, it is becoming difficult to determine what is human versus a robot; organically occurring versus automated; and real versus fake. (A reminder of the old advertising adage: “It’s what’s inside that counts”).
In 2018 chatbots will rapidly become more sophisticated, dramatically reducing costs of routine customer care activities while also improving the customer experience. In the coming year, chatbots will quickly understand the tone, content and predicted highest-value conversational paths to meet various objectives. On the dark side, this subtle “engagement” can turn to manipulation through AI that learns the magic words to sway our attitude, actions and possibly elicit en masse reactions.
AI will combine with blockchain
Beyond its association with cryptocurrencies, blockchain technology will soon record ‘time chains of events’, as applied to contracts, interactions and occurrences. In these ‘time chains’, people and the items they interact with will have encrypted identities. The blockchain distributed will be the single source of truth, allowing audit trails of data usage in models – particularly in data permission rights.
Think of the latest vehicle-leasing innovation that is set to hit South Africa in the very near future: You walk up to a car, parked in a secured public bay, which you have been pre-approved to lease for, say, the afternoon. Insurance contracts are attached to this car’s blockchain. The car itself also has a history of past drivers, events, and maintenance that is codified. And as you drive through the city, you interact with toll roads and parking spaces, all automatically recorded and monitored on the blockchain. When you leave the car at your desired destination, your lease is completed and becomes auditable on the chain – with all the necessary information in place, including whether you locked the car’s door or not.
Data event chains will create new opportunities for graph analytics, and novel new AI algorithms for the consumption of relationship data at scale.
In 2018, we will see new analytics around relationship epochs. Think of your daily interactions at work for example. Most days are relatively routine, but sometimes chains of events occur that have meaning, such as AML activity, bust out fraud, suicide prevention opportunities, and many others.
Understanding these webs of relationships of events will certainly add more insight. Analytics like these will feature scoring based on shifting chains and graphs. Their webs of interaction will thus hold tremendous predictive power.