What defines the outermost boundary of the universe? When did the concept of a day come into existence?
These are the inquiries that intrigued computer scientist Fei-Fei Li as a young physicist. As she progressed in her academic journey, she began exploring new realms of inquiry, particularly focusing on human and artificial intelligence.
Li is most notably recognized for her contributions to the field of artificial intelligence. This year marked the release of her memoir, “The Worlds I View: Curiosity, Exploration, and Discovery at the Dawn of AI,” blending her personal narrative with the evolution of AI technologies.
Throughout her career, Li has advocated for a human-centric approach to AI development. This approach emphasizes creating technology that is inspired by human intelligence and biology, leveraging AI to enhance human capabilities rather than replace them, and considering the potential impacts on human society.
Transitioning from Perception to Scientific Inquiry
Li’s scientific journey commenced with a fascination for scientific inquiry itself. She was captivated by the way scientists posed and pursued novel questions.
While delving into the works of prominent physicists, Li was not only intrigued by their inquiries into the molecular world but also by their contemplation of life and intelligence. Her curiosity about the brain was further ignited during an internship at the University of California, Berkeley, where she explored the potential for complex, high-level cognition and perception to emerge from interconnected neural networks.
Of particular interest to Li was the realm of perception.
In her book, she articulates, “Vision has the power to distill our world into discrete concepts that can be articulated,” contrasting the overwhelming sensory input of light, color, and form.
Subsequently, Li ventured into the domain of computer vision within artificial intelligence, which involves training machines to comprehend and interact with visual stimuli such as x-rays and autonomous vehicles. She notes that this process mimics the visual system of animals, albeit utilizing cameras and sensors instead of biological eyes and retinas to capture and interpret visual data.
To achieve this, computer scientists employ neural networks, which Li asserts are inspired by the structure of the human brain. While a single nerve cell constitutes the basic unit of the body, neural networks consist of millions of interconnected nodes arranged in layers. These network components receive and process information akin to neurons in the brain.
Unraveling the Mysteries of Artificial Intelligence
Despite significant advancements in AI technology, Li posits that fundamental questions regarding how AI learns remain unanswered.
While contemporary AI tools like Chat GPT are ubiquitous, Li raises a pertinent query: why do these tools exhibit proficiency in human-like language but often make glaring mathematical errors?
She highlights the enigma of AI decision-making processes, emphasizing that despite leveraging vast datasets from the internet to train AI models, the mechanisms underlying data processing and decision-making remain opaque.
Philosophically, Li probes the essence of computer vision, questioning, “How do computers perceive images, given that a photograph comprises mere light, color, and shading—yet we discern a cat within it?”
As the utilization of AI proliferates and more researchers delve into the field, these unresolved questions are poised to proliferate.
Fostering Socially Responsible AI
While acknowledging the dual potential of AI for both positive and negative outcomes, Li underscores the imperative for scientists to proactively address potential ethical concerns associated with AI, such as algorithmic bias, to ensure its beneficial application.
She advocates for societal engagement in shaping the trajectory of AI development.
Despite the rapid evolution of AI technologies, the concomitant lack of robust governance mechanisms poses a significant challenge. Li contends that this predicament is somewhat inevitable, as governance frameworks typically lag behind technological advancements. She suggests a paradigm shift towards utilizing AI to augment human capabilities rather than supplanting human labor. To this end, she established the Stanford Institute for Human-Centered Artificial Intelligence, underscoring the importance of interdisciplinary collaboration involving experts and non-specialists across various domains in shaping the future of AI.
According to Li, “Human-centric principles should underpin the development, deployment, and governance of AI applications.”
Fei-Fei Li’s memoir, “The Worlds I View: Curiosity, Exploration, and Discovery at the Dawn of AI,” is now accessible to readers. In addition to her role as the Vice President of Google and Chief Scientist of AI and machine learning at Google Cloud from 2017 to 2018, she founded the Human-Centered AI Institute and the volunteer initiative AI4ALL, aimed at empowering underrepresented high school students in STEM fields to engage in artificial intelligence research.
