JULY COVERAGE SCHEDULE
- Morgan Wright, Chief Security Advisor at SentinelOne: "Digital Defenses for Medical Databases"
- Doug Milburn, Founder of 45Drives: "Demicrosoftification: A New Paradigm in Enterprise Computing"
- Emanuel Pastreich, President of the Asia Institute: "Miracle in Ulaanbaatar"
- Jin Hyung Kim, Emeritus Professor at KAIST: "Democratizing AI: Innovation for Everyone”
- Layne Hartsell, PhD, and Alexander Krabbe, MD: The essay "War in Eurasia"
- Eric Jaremalm, CEO of Midsummer: "The virgin market for industrial thin-film solar roofs"
By Jin Hyung Kim, Emeritus Professor at KAIST
It has been nearly 70 years since the advent of computers and the beginning of artificial intelligence (AI) research. AI technology has undergone significant advancements, resulting in revolutionary outcomes. AI enables computers to perform intelligent tasks, a pursuit that truly began after the invention of computers. It was initially assumed that AI would be a straightforward process to develop, but this proved to be a significant miscalculation. From emulating human cognition to optimizing it, AI is now a central component of automation technology.
AI has made significant advancements in scientific research and is demonstrating remarkable outcomes in industrial applications. Daily, new AI services are launched, frequently resulting in millionaire startup founders. It is reasonable to conclude that AI has permeated every aspect of our daily lives. Some experts even claim that artificial general intelligence (AGI) is just around the corner. AGI differs from narrow AI in that it is capable of performing a range of tasks at a level that rivals or exceeds that of a human. This is the ultimate objective of research in this field. I admire the Silicon Valley ecosystem, where AGI research and development is privately funded.
Along with the excitement, there are also growing concerns. AI is not a simple technology. There is a risk that this powerful technology will be misused. Some dictatorships are using AI systems to oppress and control their people. Additionally, the complex nature of AI could lead to mismanagement and harm to humanity, even with the best intentions. Furthermore, there is a risk that some people may create AI with bad intentions from the start.
As technology continues to advance rapidly and transform the world in profound ways, many individuals are left behind, unable to keep pace with the changes. There is a concern that AI could undo thousands of years of civilization and pose a threat to humanity's survival. As new AI technologies emerge, there is a growing consensus that they are too powerful and require regulation. The Vatican has issued warnings, and the European Union has enacted legislation to regulate it.
What are the key drivers behind the development of this powerful and potentially disruptive technology? What research philosophies have informed the development of AI over the past 70 years? In hindsight, the most significant driving force was the forward-thinking objective and catchphrase of "building computers that think." Furthermore, the democratization of technology to disseminate knowledge on a broad scale has been of significant importance. Intellectuals are by nature inclined to disseminate their knowledge. They convene at conferences, publish research papers, and disseminate and discuss their findings with their colleagues. This culture of knowledge communities has been a driving force behind the development of humanity’s scientific and technological civilization. The ecosystem of open and shared knowledge created by Western science and technology researchers during the Reformation and Industrial Revolution is an inspiring model for modern business.
The open software movement is focused on opening and sharing software in a way that allows anyone to use and improve it without restrictions. It extends beyond the publication of knowledge in papers to include the full disclosure of ideas and source code, enabling anyone to test and replicate the software. It is a movement to share everything, including source code, which is developed with great effort and has significant economic value, and training data, which requires considerable effort to collect. It incorporates elements of socialism as well as democratization.
The digital civilization we currently enjoy has been shaped significantly by open software. All software utilized in the construction and operation of the Internet is open source, and smartphones and personal computers run on open source operating system software. The World Wide Web, the Internet's primary information-sharing system, is also standardized, with its source code being open for public access. It is accessible to all, offering information and the ability to engage in e-commerce. The Internet economy has benefited significantly from the use of open software.
Seventy years ago, at the outset of computer software technology and the early stages of AI development, this philosophy of open software gained significant traction in research, largely driven by concerns about the potential implications of AI. From its modest origins, the ambitions were considerable, and the concept of AI research could be daunting to the general public, who were not conversant with the cutting edge of the field. This resulted in the philosophy of "full" disclosure of software in all research conducted under the auspices of AI. It has become standard practice for researchers working in the field of AI to publish their ideas, regardless of how unconventional they may be. Richard Stallman, one of the pioneers of the open software movement, was a researcher at MIT’s AI Lab, which is illustrative of the early AI research scene.
I benefited personally from open software while I was studying in the United States. I reviewed the source code, gained an understanding of the technology, and modified it to test the feasibility of my ideas. This approach enabled me to accelerate the pace and depth of my research. Upon returning home, I continued to recognize the advantages of open source, particularly in the context of the Internet’s renewed growth. It is advantageous to be able to exchange scientific and technological ideas at high speed.
As the focus of AI development has shifted to machine learning, the importance of training data has grown in comparison to training algorithms. The initiative to collect and disseminate training data among researchers has been a significant driver of AI development. It has become standard practice in the machine learning community to publish and share trained neural networks or AI models. This is when the term "democratization of AI" became more commonly used than "open software" to describe the sharing of trained AI models.
Several proposals have been put forward to promote the democratization of AI. The concept of a data-sharing platform called federated learning is an excellent one. The issue of data ownership is of significant interest, given the considerable economic value of data and the importance of data privacy. Personal health information is of a particularly sensitive nature. There is a desire to utilise data belonging to others, but a reluctance to share one's own. Federated learning enables the sharing of data without making it public, thereby significantly advancing AI while preserving property rights. The learning algorithm visits the data owner's site, learns from it, and then leaves without replicating the data.
The impact of democratizing AI on society and the economy is significant and far-reaching. Making the technology public and open reduces the fear of AI and encourages younger people to embrace new challenges. The democratization of AI drives technological advancement and spurs innovation across numerous sectors. Furthermore, it makes the technology accessible to a wide range of users, including the general public and small and medium-sized enterprises, rather than limiting its use to a select few experts or large companies. Young people in remote parts of the world can leverage open AI technologies to develop new services and launch their own businesses. The global Internet will provide these startups with the opportunity to expand into global companies.
By democratizing AI, these changes will benefit us all. It’s not just about technology; it’s about social justice and equality. It will make our future a more equitable and innovative society. Democratizing AI is something we all need to work towards.
About the Author
Dr. Jin Hyung Kim is an emeritus professor at KAIST, with 50 years of expertise in AI and software computing. He earned his Bachelor's degree from Seoul National University and a Ph.D. in computer science from UCLA. Dr. Kim has held prominent positions, including senior researcher at the Hughes Institute for Artificial Intelligence, director of the Software Policy Institute, and president of the Artificial Intelligence Research Institute. He has also served as a professor at KAIST, president of several institutions, and acted as chair professor at Chung-Ang University. A life member of both the National Academy of Sciences and the National Academy of Engineering, Dr. Kim has received numerous accolades, including the 2010 Korea Internet Awards' Chairman’s Award for Individual Achievement, the Green Bird Distinguished Service Medal in 2015, and the Internet Award in 2001. He has authored several books, including "The Best Class in AI with Professor Jin Hyung Kim of KAIST."
