Successful failure: Growing Artificial and Human Intelligence

I made a mistake. As a leader, I'm no stranger to errors, but this one was mine personally. I failed to communicate some information to a team member properly and navigate through the right channels. While such relatively minor lapses can seem trivial - after all, to err is human - they exemplify a broader, more profound lesson about growth and understanding, both personally and technologically.

Making mistakes is fundamentally human. They remind us that imperfection is a part of life, serving as opportunities to learn, to enhance our efficiency, communication, compassion, and mutual respect. In essence, our errors are not merely setbacks; they are setups for future success, encouraging us to 'fail forward' - to not only recover from our falls but to rise informed and improved.

"The universe does not allow perfection. One of the basic rules of the universe is that nothing is perfect. Perfection simply doesn't exist... Without imperfection, neither you nor I would exist." - Dr. Stephen Hawking

Failure drives innovation. A compelling example of embracing mistakes to innovate comes from SpaceX . In their ambitious quest to not only launch rockets but also achieve the seemingly impossible feat of landing them back on Earth, SpaceX leveraged failures as stepping stones. Initially, the mathematics of of landing a rocket posed massive challenges. Some said that it was mathematically impossible. In response, SpaceX adopted a strategy of learning directly from errors: they would launch rockets, attempt landings, and, more often than not, watch them fail spectacularly. Each failed attempt, each explosion, was meticulously analyzed to understand what went wrong. This cycle of iterative testing - launching, failing, analyzing, and adjusting - eventually culminated in success. On December 21, 2015, SpaceX made history when their Falcon 9 rocket not only delivered its payload of 11 communication satellites to orbit but also returned to land successfully in reverse on a landing pad in Cape Canaveral - the first-ever orbital class rocket landing - marking a monumental achievement in aerospace technology. As it turned out, failure is also success!

This concept of learning from mistakes bridges us from human fallibility to artificial intelligence. Like humans, AI systems are not immune to errors. However, the nature of mistakes made by AI - and the learning processes that follow - differs significantly from our own experiences. In the realm of artificial intelligence, failures are critical learning opportunities. Take, for instance, the early iterations of facial recognition technology. Initially, these systems exhibited significant biases, particularly in accurately identifying individuals from diverse ethnic backgrounds. This was notably highlighted in a study by MIT researchers, who found that facial analysis algorithms were far less accurate at identifying gender for women and people of color (Source ). These errors sparked widespread critique and led to intensified scrutiny and improvements in AI training datasets and algorithm design to enhance fairness and accuracy.

Another example is autonomous vehicle technology. Early prototypes often struggled with unexpected scenarios, like recognizing stop signs obscured by graffiti or interpreting the movements of pedestrians. Incidents involving autonomous vehicles, including a few high-profile accidents, have underscored the importance of extensive real-world testing. Each mishap has provided invaluable data, driving advancements in sensor technology and machine learning models to better interpret complex, unpredictable environments.

The instances of AI failing are not just setbacks but can be invaluable for the maturation of the technology. They particularly highlight the need for robust, diverse datasets and rigorous testing environments to train AI systems, ensuring they are safe and effective for real-world application. By learning from these failures, we enhance AI's reliability and trustworthiness, ultimately paving the way for more sophisticated and equitable technologies. The real-life impact of AI errors can be profound. When AI systems falter - say, by misinterpreting data or exhibiting bias - the consequences can range from minor inconveniences to significant misunderstandings or even injustices. Regrettably, the human cost of autonomous cars making errors has been high and has resulted in the loss of several human lives (Source ). Such mistakes, unlike most human errors, stem from the quality of the data fed into them and the design of their learning algorithms. These are not just technical glitches but reflections of deeper systemic issues, often rooted in datasets that are incomplete or unrepresentative of the diverse world we live in.

In our interactions with AI, these technological missteps create a different dynamic than the errors made between humans. While a human error may lead to a breach of trust or a moment of empathy, which often can be positively affected by an apology and learning, an error in AI challenges us to reconsider our reliance on and the governance of technology. It raises questions about the biases embedded within AI systems and the responsibilities of tech companies to create algorithms that do not perpetuate these biases. This conversation around AI biases has not only permeated academic and technological circles but has also ignited a broader societal movement demanding more ethical AI practices.

Influential books such as "Weapons of Math Destruction" by Cathy O'Neil and "Unmaking AI" by Joy Dr. Joy Buolamwini and documentaries such as Netflix ′s "The Social Dilemma", featuring Tristan Harris , co-founder at the Center for Humane Technology have played pivotal roles in this discourse. Their research reveal how opaque, unregulated algorithms can reinforce discrimination, leading to devastating impacts on people's lives, particularly women, minorities, the poor, and underserved. Their words have galvanized activists, policy makers, and concerned citizens alike, pushing for regulations that ensure AI technologies are developed and deployed with a keen awareness of their social implications.

These growing concerns and the call for action have prompted governments around the globe to take more proactive steps in regulating AI technologies. The European Union's AI Act is a landmark proposal aiming to set comprehensive rules for AI use, with specific provisions to prevent discrimination by AI systems. Similarly, Italy's recent bill on AI legislation explicitly forbids discriminatory practices, setting a legal standard for AI fairness. In the United States, the federal government has been integrating AI governance into its structures , appointing Chief AI Officers and mandatory bias testing across all federal agencies to ensure ethical AI deployment. I am proud to be affiliated with the incredible Center for AI and Digital Policy in Washington DC, which plays a crucial role in shaping US national and global AI strategies. Other significant initiatives include the Australian Safety Commissioner's efforts to oversee AI in digital platforms and the African coalition for AI ethics, which aims to safeguard rights and promote ethical standards across the continent, as well as many other regions and nations. Together, these efforts represent a comprehensive global movement towards responsible AI that respects human rights and addresses societal challenges.

This momentum has encouraged a more critical view of AI's role in society, urging tech companies to commit to transparency and fairness in algorithmic decision-making. By advocating for 'algorithmic accountability ', the AI ethics movements stress the importance of ethical standards that go beyond technical efficiency, addressing potential social harms. As a result, some companies have begun implementing ethical AI guidelines and conducting more rigorous bias audits before deploying AI systems in real-world applications.

Recognizing the distinction between human and artificial intelligence, and how each learns from mistakes, is crucial in our increasingly digital world. It underpins our need for stringent oversight of AI development and a thorough understanding of AI's capabilities and limitations. As educators and leaders, it is our responsibility to ensure that these technologies are leveraged to enhance educational outcomes, and reinforce equitable learning, not undermine them.

As we continue to integrate AI into our educational systems and daily lives, we must remain vigilant about the errors these systems make. We must advocate for transparency, accountability, and continuous improvement in AI technologies to safeguard our values and ensure these tools benefit all students equitably.

Ultimately, our journey with AI, much like our personal paths, is about learning - from every mistake, every glitch, every unexpected outcome. As we teach our machines to learn, we must also learn from them, ensuring that our technological advancements reflect our deepest human values: understanding, respect, and the perpetual pursuit of knowledge.

#edtech #aiethics #shapeai #shapingtheworld

Click here for a reading list of books on Bias and Equity in AI.

Click here to watch a #CogXfestival2023 talk by Tristan Harris.

Follow the Center for AI and Digital Policy on Linkedin .

Follow Globeducate on Linkedin .

To follow me on Linkedin, please click here .