I am a business integration architect who is fascinated with the near term and practical applications of Artificial Intelligence (AI) for financial services. In order to identify practical applications, I have found that it is first necessary to understand the different paradigms of AI in order to get a better sense of what type of AI fits where.
The two paradigms referred to in this post are statistical and deterministic. Each is useful and provides solutions for very different scenarios.
What is the Difference?
When I think of statistical, I think of generating insight and offering predictions from vast amounts of data. Statistical, such as machine learning, helps in the discovery process for an expert. It is great for identifying patterns that in turn inspire new ideas.
Deterministic, in my mind, is on smaller data sets that require an understanding of context. To do this, the machine and the person need to have a dialogue. In this case, the machine is the expert.
For example, when I go to the doctor, I really don’t provide him with much information at the outset. He asks me questions and I provide answers, most of which I didn’t think were important. He processes those answers against his experience with patients in a similar demographic, knowledge of my family history (which I gave him the year before), and medical journal articles he recently read.
Then, he asks me more questions. From the new answers, he determines that I really don’t need the stress test that I alluded to when I came in. Instead, I should really be concerned with getting blood tests for reasons that never occurred to me. This is not the same conversation he had with any other patient.
Statistical AI will help a doctor identify the most effective treatments based on the statistical analysis of huge amounts of historical and current data. However, that doctor shouldn’t apply those treatments until she or he determines relevance based on the dialogue.
It is the deterministic AI that I find really exciting.
Deterministic AI is directly applicable to the end user who is not a data scientist or expert. Scenarios include qualifying the suitability of an individual for a product, generating hyper-targeted messages to consumers, providing compliance insight before a trade, or displaying information that is helpful and relevant to the individual.
In these examples, the narrative (or speech) that is provided by the machine needs to be descriptive. However, it also needs to be inquisitive and provide reasoning and explanation.
Expertise Brought at Scale
In the end, insurance or financial decisions should be an individual decision that reflects quantitative data. The decisions should also be informed by an individual's quality of life and vision for the future. Deterministic AI can provide this expert dialogue at scale.