AI is a powerful tool with incredible potential, particularly in the medical field, where it has shown a remarkable ability to accurately diagnose diseases. Recently, I embarked on a project in Korea. Despite my lack of knowledge of the Korean language, I discovered that the AI translation app on my phone frequently outsources better translations than my native Korean colleague. This just goes to show that AI can enhance our capabilities. The recent events surrounding top judges in the U.S. have ignited conversations about whether AI could eventually step in to help deliver fairer, more impartial rulings. We are on the brink of a transformative era, and embracing AI is crucial.
As we approach the U.S. election, AI-generated ads are making waves. Just recently, the former president shared AI-created images that misleadingly suggested Taylor Swift’s endorsement. This situation highlights the importance of using AI responsibly to mitigate manipulation. While deepfakes and their moral implications have garnered attention, it’s essential for us to shift our focus towards understanding the implications of flawed AI systems and how we can make improvements.
The effectiveness of AI relies heavily on data models that necessitate vast amounts of information; its accuracy hinges on the quality of that data. We must acknowledge that AI algorithms are crafted by humans, often reflecting the backgrounds of their creators, predominantly male and highly educated. This reality emphasizes the importance of integrating diverse perspectives to build fairer AI systems, so that even with the best intentions, we can avoid unintended biases.
Consider a scenario with an autonomous car facing an imminent accident. With a large truck heading directly toward it, the car must make a critical choice. On one side of the road, there are elderly people; on the other, a young child, alongside lamp posts in both directions. The car’s AI identifies four possible actions:
- a) Crash into the lamp post on the left, sacrificing the driver but saving the passengers and pedestrians.
- b) Crash into the lamp post on the right, similarly sacrificing the driver but preserving everyone else.
- c) Crash into the group of elderly individuals, thus sparing the child and those inside the vehicle.
- d) Crash into the child, rescuing the elderly group and the occupants of the car.
From an engineering perspective, devising a scoring system based on life circumstances could guide us toward making decisions aimed at minimizing harm. For instance, elderly individuals, who may have experienced most of their lives, could receive lower scores due to their shorter life expectancy, regardless of their collective value in numbers. Conversely, the child, filled with potential, would likely be assigned a higher score. The safety of the car’s occupants, especially the driver—often the owner—might also merit prioritization, leading to high scores for those within the car.
However, we must ask ourselves: is this ethical? Can the value of human life truly be captured by a numerical system?
Another possibility could involve the use of big data to examine the outcomes of similar incidents, allowing us to create models that guide us toward the most ethical actions. Yet, the key issue remains: the quality of that data forms the foundation of any decision. In high-pressure situations, human instinct often sways toward self-preservation, which may unintentionally influence the AI’s choices. Should this tendency serve as a guide for AI actions in these scenarios?
It is important to note that numerous tech companies employ AI to evaluate job candidates, sometimes introducing bias that favors younger male applicants. Similarly, in healthcare, AI-powered diagnostic tools have displayed lower accuracy rates for Black patients compared to their white counterparts.
Data imbalance significantly contributes to AI bias. One simplistic solution often proposed is to develop open algorithms subject to community input and revisions, akin to Wikipedia’s model. However, the algorithms and databases tend to be closely guarded trade secrets of tech companies. Furthermore, the technical complexity of these algorithms restricts the ability of the general public to review or modify them, potentially perpetuating biases.
In my personal devotion, I’ve been exploring the Old Testament—specifically the Book of the Prophets—where many prophets decried the injustices faced by the Israelites. AI bias, too, represents a form of injustice. As AI technology continues to evolve, it is vital for individuals of faith to be advocates for the marginalized and to raise awareness about the injustices present in the digital landscape. Get actively involved in shaping the narrative. For those embedded within the industry, take on the mantle of your company’s ethical voice and strive to develop fairness-centered algorithms.