Balancing innovation with human rights

“Success in creating effective AI could be the greatest event in the history of our civilization. Or the worst. We just don't know.” In 2017, Stephen Hawking's prophetic warning about artificial intelligence hangs in the air: a potential game changer for humanity or a chiller of doom. According to my colleague Grace Hamilton at Columbia University, the truth lies somewhere in the middle, as with most disruptive technologies.

AI has undoubtedly ushered in a golden age of innovation. From the rapid analysis of big data to the terrifying prescience of predictive analytics, AI algorithms are transforming industries at breakneck speed. It's become ubiquitous, quietly shaping our daily lives—from the familiar face scan at the airport to ChatGPT's uncanny ability to craft a coherent essay in seconds.

This rapid integration has lulled many into a false sense of security. We've grown accustomed to the invisible hand of AI, often seeing it as the inevitable domain of tech giants or simply inevitability. Legislation, lumbering and outdated, struggles to keep pace with this digital cheetah. Here's the wake-up call: AI's human rights implications are neither novel nor inevitable. Remember, technology has a long and distinguished history of protecting and challenging our fundamental rights.

The story of the rise of AI mirrors the explosive growth of the Internet in the 90s. At the time, a laissez-faire approach fueled the creation of tech titans like Amazon and Google. Thrive in the unregulated Wild West, these companies have amassed mountains of user data, the lifeblood of AI development. Today, the result is a landscape dominated by powerful algorithms, some so sophisticated that they can make completely autonomous decisions. While it has revolutionized healthcare, finance and e-commerce, it has also opened a Pandora's box of privacy and discrimination concerns.

After all, AI algorithms are only as good as the data they're trained on. Biased data leads to biased results, perpetuating existing inequalities. Additionally, AI companies' insatiable appetite for personal information raises serious privacy flags. Striking a balance between technological progress and the protection of human rights is a clear challenge of our time.

Consider facial recognition technology. In 1955, the FBI's COINTELPRO program weaponized surveillance against Martin Luther King Jr., a striking example of technology used to silence dissent. Later, in January 2020, Robert Williams answered a knock at his front door. A black man from Detroit, Williams was unprepared for the sight that greeted him — cops on his doorstep, ready to arrest him for a crime he didn't commit. Blame? Stealing a collection of high-end watches from a luxury store. Criminal? A blurry CCTV image matched by poor facial recognition technology.

This was not just a case of mistaken identity. Instead, it was a stark display of how AI, especially facial recognition, can perpetuate racial bias and lead to disastrous consequences. The photo used by the police was of poor quality, and the algorithm, trained on a possibly unbalanced data set, disproportionately misidentified Williams. As a result, Williams spent thirty agonizing hours in jail, away from his family, his reputation tarnished, and his faith in the system shattered.

However, Williams' story became more than just a personal injustice. He publicly expressed the fact that “many black people will not be so lucky” and that “nobody deserves to live with that fear.” With the help of the ACLU and the University of Michigan's Civil Rights Litigation Initiative, he filed a lawsuit against the Detroit Police Department, accusing them of violating his Fourth Amendment rights.

Williams' story is not an isolated incident. It's a reminder of the inherent dangers of relying on biased AI, especially for critical tasks like law enforcement. As of 2016, Williams is one of 117 million people—about half of all American adults—whose images are stored in facial recognition databases used by law enforcement agencies.

In the vast expanse of facial recognition databases, biases are magnified. In fact, studies have shown that facial recognition algorithms have higher error rates when identifying people of color, with the highest error rates for darker-skinned women — compared to lighter-skinned men. Up to 34% higher than the competition.

However, there is a ray of hope. Decentralized autonomous organizations (DAOs) like Decentraland offer a glimpse into the future of transparent, community-driven governance. Leveraging blockchain technology, DAOs empower token holders to participate in decision-making, promoting a more democratic and inclusive approach to technology.

Still, DAOs are not without their flaws. A major security breach in 2022 exposed vulnerabilities in user data, highlighting the privacy risks inherent in decentralized structures. Of course, the absence of centralized oversight can also create breeding grounds for discriminatory practices.

The US Algorithmic Accountability Act (AAA) is a step in the right direction, aiming to shed light on the often confusing world of AI algorithms. The AAA seeks to promote a more transparent and accountable AI ecosystem by requiring companies to review and report potential biases. Technological solutions are also emerging. Diverse datasets and regular ethical audits are being implemented to ensure fairness in AI development.

The road ahead requires a multi-faceted approach. Strong regulations and ethical frameworks are crucial for the protection of human rights. DAOs should incorporate human rights principles into their governance structures and regularly assess AI impacts. Enhancing strict warrant requirements for all data, including Internet activities, is essential to protect intellectual privacy and democratic values.

The legal system needs to address the chilling effects of AI on free speech and intellectual pursuit. Managing the use of differentiated AI is paramount. Facial recognition technology should only be used as additional evidence, with pre-existing safeguards against perpetuating systemic bias.

Finally, slowing the development of runaway AI is critical to allowing regulations to catch up. A national council dedicated to AI legislation can ensure the evolution of human rights frameworks alongside technological progress.

The bottom line? Transparency and accountability are essential. Companies should disclose biases, and governments should set best practices for ethical AI development. We must also ensure fair data sources with diverse datasets trained on individual consent. Only by addressing these challenges can we harness the immense potential of AI while protecting our fundamental rights. The future depends on our ability to walk this narrow path, ensuring that technology serves humanity, not the other way around.