Nick Bostrom is considered one of the leading voices of today’s tech world. An Oxford philosopher, Bostrom wrote the bestseller “Superintelligence: Paths, Dangers, and Strategies,” published by Oxford University Press and recommended by industry giants such as Bill Gates, Elon Musk, and Max Tegmark. Bostrom is one of the most influential people in the tech industry, and his book outlines the existential risks of artificial intelligence. The book puts AI safety at the forefront of societal concerns about AI’s rapid growth.
One of his key ideas is to offset AI’s potential problems by aligning it with human values, giving it the ability to go beyond explicit instructions and achieve its goals without harming humans in the process. This formed the basis of AI research disciplines. The AI industry supports this effort. The only problem is, whose values are we designing it to be?
The following is an email Bostrom sent to one of his colleagues in 2023: “I have always liked the uncompromising objective way of thinking and speaking. Take, for example, the following sentence: Blacks are more stupid than whites. I like that sentence and think it is true.”
Bostrom’s key idea is that, to combat the potential problems AI may pose, it must be made compatible with human interests. It just happens that the humans in this case- the people on the leading edge of technology, the same people who are leading and shaping the future of AI- are racist.
Another prominent figure in the AI industry is Timnit Gebru. Gebru was born in Ethiopia, migrated to the United States as a child, and pursued a career in electrical engineering and artificial intelligence. In Karen Hoe’s book, “Empire of AI,” Gebru’s path through the world of AI is filled with roadblocks. Even though the book is not about her, she is sprinkled throughout this story of AI from the second page all the way to the last chapter. When we trace her path, we get a clear picture of where AI may be heading in the wrong direction.
Timnit Gebru worked in a world dominated by white men. In the book Empire of AI, Karen Hoa explained Gebru’s thoughts on the lack of Black people in AI. “She imagined starting a YouTube channel to lampoon the situation, changing her hairstyle and acting out different personalities to dramatize speaking to herself.”
At conferences, she would often be the only Black person in attendance. At one conference, the Neural Information Processing Systems, she was even physically harassed by a group of young white Google employees who grabbed her and forced a kiss on her cheek. One of them snapped a humiliating photo of the incident. I believe that if there had been even a smattering of Black people there, that incident would not have happened. But since there are so few Black people in that environment, the dominant group takes liberties. It feels empowered, through their look-alike peers and supervisors, to behave and act unprofessionally without impunity.
Gebru did not passively walk away from this incident. She wrote a critique in an anonymous open letter, saying the current homogeneity of the people building AI was discouraging talented people from participating and leading to a narrow conception of AI. She made it clear when she added, “We don’t have to project into the future to see AI’s potential adverse effects; it’s already happening.” Gebru, like many other people of color in the AI world, believes that the dominant direction of AI needs to change. All of this was taking place during the murder of George Floyd and the hate that was going around and has yet to go away peacefully.
As a Google researcher, she led Google’s ethical AI team. She produced the groundbreaking paper called “Gender Shades”, which found that facial analysis applications were designed without Black people in mind. This led Amazon, Microsoft, and IBM to ban their sales of facial recognition to police. More on that later. She also started “Black in AI,” a nonprofit dedicated, as written on their website,”… to shift the power dynamic across the AI ecosystem to broaden the range of voices contributing to the development, deployment, and regulation of AI technology.”
Gebru would eventually get fired from Google after her research paper, “On the Structure of Stochastic Parrots,” warned about the problems of large language models and the environmental dangers of building data centers. Google rejected the paper and wanted its employees’ names removed. Gebru refused and was fired. They said she resigned.
Years ago, I wrote about the dangers of those so-called gunshot detectors that were beginning to be installed throughout the United States. My concern then was that these systems were not designed for what they claim to do. They are designed for surveillance and data collection of all types, from facial recognition to license plates and conversations. At that time, I wrote a column (Technicalities) for the Colorado Springs Independent. The editor got angry with me and asked what proof I had and how I could predict the future. Today, those gunshot detection systems have added cameras to the mics, and real-time surveillance is taking place. Guess which neighborhoods.
As an educator, I encourage everyone, whether they are going into cybersecurity or tech in general, to get a core understanding of technology. Certifications and professional certifications in technology should be given out just as diplomas are in high school. I am in the AI environment, and even here in Colorado, the number of Black people is minimal in this environment. Just as in the rest of the country, Black people are not participating in the AI ecosystems. With the rapid development of AI and its economic impact, this is a problem that needs to be seriously addressed. Why is this important?
The only way to ensure AI is fair is to have a diverse population at all levels of research and development. But that is not the direction it is heading in. The risk is that communities of color and other underrepresented populations are harmed because of the embedding of sex and racial stereotypes into AI development by a small group of people who do not represent the majority of the population. We do not need racist views replicated in AI. Good grief, don’t we have enough problems?
Editor’s note: Thomas Holt Russell is the founder of SEMtech (Student Engagement & Mentoring in Tech), writer, educator, photographer, modern-day Luddite, Existentialist, and secular humanist.
