# Addressing AI Bias Through Diverse Data in Online Dating Profiles
Written on
Chapter 1: The Prevalence of Online Dating Profiles
Online dating profiles are being traded in vast numbers, presenting a unique opportunity to address the persistent issue of racial bias in artificial intelligence.
Photo by Charles ?? on Unsplash
We are approaching a transformative era where intelligent automation will permeate nearly every aspect of our lives. However, many algorithms are tainted by deeply rooted racial biases that have affected society for generations. Alexandria Ocasio-Cortez highlighted this concern regarding an IBM facial recognition initiative, which could enable law enforcement to identify suspects based on skin color and other prejudiced criteria.
Coincidentally, just a day before her remarks, I had recorded a video discussing the issue of racial bias in AI:
It's encouraging to see emerging political figures championing this critical issue, and it's likely that others will join this important conversation. Nevertheless, historical lessons remind us that politics and technology operate on vastly different timelines.
Consider the past instance when the U.S. government believed that Microsoft monopolized the future of consumer technology. Today, Microsoft is not even among the top five consumer tech companies, illustrating that we cannot depend on governmental intervention to manage these complexities. They function in different realms of comprehension. Fortunately, we are still in the nascent stages of AI development, giving us a chance to address these biases proactively.
Chapter 2: The Impact of AI in Our Lives
Realistically, we can enumerate the few platforms that currently leverage AI, which may be influenced by racial bias—YouTube, Facebook, and content services like Spotify and Netflix, along with select applications in healthcare and the justice system. We are still in the early adopter phase.
Kevin Kelly famously stated, “The business plans of the next 10,000 startups are straightforward: take X and add AI.” This prediction suggests that we will see a significant rise in AI companies over the next decade. Thus, addressing the issue of bias sooner rather than later will mitigate future complications. One potential solution lies in standardizing the datasets utilized to train AI systems.
Section 2.1: The Need for Diverse Data
AI systems are only as competent as the data they learn from. Unfortunately, the existing datasets often harbor biases, leading to similarly biased AI outcomes. There is a significant market opportunity for businesses to create diverse datasets for developers to utilize in training their algorithms. While data brokering has been attempted previously, it lacked the refinement and ethical standards we aspire to achieve.
For example, USDate is a business that sells extensive online dating profile data to entrepreneurs looking to launch their own dating platforms or those simply seeking a wealth of information. Other public marketplaces, like Factual and BDEX, provide consumer data ranging from location insights to purchase histories. It's alarming that one can acquire hundreds of thousands of profiles for less than a hundred dollars, often without any verification of the authenticity of the information.
This scenario exemplifies the current unregulated landscape of public data brokering, which resembles a dubious underground operation in desperate need of reform.
Section 2.2: The Value of Proprietary Data
Proprietary data is poised to become one of the most valuable assets for companies, whether they are small local businesses or expansive marketing firms. Consequently, there is an urgent need for the establishment of data brokering platforms that are more credible and reliable, offering superior datasets.
Mattermark was a notable company on this path until its acquisition by FullContact led to the privatization of its data. Mattermark provided comprehensive reports and trends on the private startup sector, quickly becoming an invaluable resource, often compared to the Bloomberg Terminal for startups.
What sets Mattermark apart is its ability to generate data in areas where none existed. They engaged in extensive research, piecing together resources to create their own datasets.
This is the model I envision for the next generation of data brokering entities as they tackle the issue of biased datasets. They will need to actively seek out and construct their own inclusive datasets to ensure that future AI developments are founded on unbiased information.
While this is a monumental task, the organization that successfully navigates this challenge will not only achieve substantial financial success but also help alleviate the fears of human racial bias seeping into AI technologies.