Imagine you run a bank that lends money to coffee shops. Every day, you have to decide whether a quaint little coffee shop in some town deserves a loan or not. If this task is assigned to a traditional AI system, the machine will probably check the financial statements, the credit rating of the owners, assets and liabilities of the shop, etc. It might also check analytics like the number of repeat customers and the coffee flavours that make the most profit.
But is that all the information required to make that all-important business decision to lend or not to lend? How about checking the commitment and enthusiasm of the owners to scale up their business into a multinational coffee shop chain? How about other factors like the camaraderie between the partners of the coffee shop?
Gauging such factors that come easy to a professional banker, is quite tough for most AI systems. This is one among the several instances where human-centric AI systems would help us in solving our complex business problems.
Human-centric AI never attempts to take the human being out of the business equation. Instead, it empowers humankind to be better at their jobs. You can be a better doctor, teacher, or like in the earlier example - a better banker by using human-centric AI systems.
There are two streams of research currently happening in this field. An ambitious long-term idea is to model deep learning systems by simulating how the human brain works. The other, more promising, stream of research is to include more human input during the entire life cycle of the AI system.
As a child, all of us were incapable of functioning properly in this society. We learned the rules of the world bit by bit from our parents, siblings, teachers, family members, friends, and strangers. We didn't have to see pictures of all breeds of dogs to learn to differentiate a dog from a horse. It's the same with a human-centric AI system. Rather than supplying the AI system with a brute force list of endless data, a curated set of data is provided to human-centric AI systems.
A standard machine-learning algorithm uses large chunks of data in its learning phase. Later, the decisions made by the AI are hugely influenced by the data that is used in the learning phase. So, the quality of data used becomes highly important. But who decides what data is supplied to the AI?
Let's again consider the example of granting loans to coffee shops. In a human-centric AI system, expert bankers will be teaching the system how they decide to approve or reject a loan application. This list of experienced bankers should be diverse enough to represent all genders, ethnicity, and background to prevent the AI system from being biased towards any community.
In the learning phase of the AI system, experts from various fields can be used to improve the accuracy of the system. Fields like sociology, philosophy, economics and game theory would contribute towards improving human-centric AI systems.
Once the initial learning phase is complete, a human-centric AI system begins to arrive at decisions for each business scenario. However, we still need the discretion of a human being to validate the decisions made by the AI system. Any disagreement between the expert and the AI system provides an opportunity for perfecting the system to a better state.
Such validation steps are integral to any human-centric AI system. Remember, the goal is not to remove human beings from the decision-making process. We are only helping everyday decision-makers to be better at their jobs by using human-centric AI.
Human-centric AI makes explainable decisions. This makes it easy for humans to validate the decisions. Coming back to our earlier example of lending to coffee shop owners, consider a scenario in which the new AI system advises not to grant a loan to a small coffee shop owner named Martha. The system will also explain why Martha doesn't deserve a loan. Let's say the AI explains that it's too risky to loan Martha because of her old age. A human being will be able to identify that it is not ethical to deny Martha a loan because of her age. Here comes the most important feature of human-centric AI - being ethically sound.
In addition to solving a business problem, human-centric AI ensures fairness in its operation. In fact, human-centric AI is the answer to all our dystopian concerns about AI taking over our planet.
All industries can benefit from human-centric AI systems. In the medical field, an inexperienced doctor can quickly identify even the smallest hints of pneumonia, which otherwise, could only be identified by senior doctors. This would bridge the gap between the rich and poor in terms of expert medical care. Telecom engineers can detect patterns of problematic networks and take urgent actions without affecting emergency services using a human-centric AI system. Logistics managers can decide on the most efficient and environmentally friendly route to be taken for a particular consignment. The possibilities of human-centric AI to solve business problems ethically are endless.
Even with extensive data collection and analytics, companies struggle to understand the needs of their customers. It's not rare to see multi-million dollar marketing campaigns that ticked all the right boxes fall off the cliff during execution. Often, this happens due to the over-reliance on numbers rather than emotions.
Many products sell due to subjective qualities like beauty, image, and brand perception. It's not easy for a data analytics tool to measure such nonobjective data. Even for a traditional AI, efforts to measure the aesthetic appeal of a prospective product have not been successful in an industry environment. Human-centric AI can solve this prediction problem efficiently and help in determining the market perception of a product or service.
Most business problems are sales problems at the root of them. So if human-centric AI could determine whether what you're doing is going to sell in the grander sense of the word, it could be of help in solving such business problems.
Human-centric AI can solve complex business problems with high accuracy. We can also ensure ethically correct practices by pursuing this technology. All the dangers and concerns of the spiralling growth of AI can be avoided by encouraging human-centric AI methodologies.