Abstract
What is the relationship between artificial intelligence (AI) and artificial general intelligence (AGI)? For the most part, the former is subset of the latter, with AI being used to achieve, or build up to, AGI. AGI is general-purpose AI that can excel at many domains, reaching superhuman cognitive performance on a multitude different tasks or problems. As of now, though, AGI remains a task to be completed, and many top breakthroughs in the past few years of computer science have dealt with AI, specifically deep learning. This project’s STS thesis looked at the ways in which AI in China specifically can bethought of as a social construction, in the sense that they way it is used, viewed, developed, and deployed are completely dependent on the social, economic, and political factors of the country it was developed in (China). Thus, AI research, especially regarding new technology development and mass usage, is vastly different in China than in other countries, and much the same could be said about the United States. This project’s technical portion cataloged the research, development, and instruction behind the novel computer science course CS 1501: Artificial General Intelligence, in which the class combined ideas from AI, cognitive science, and philosophy. The course was taught by the author for 3 semesters to almost 150 UVA students and received rave reviews. The course specifically looked at theoretical questions behind AGI (i.e., What is a mind? What is consciousness?) as well as specific technical questions (i.e., What are cognitive architectures? What is a spiking neural network?), to develop a comprehensive overview and introduction to the nascent field of artificial general intelligence. Thus, the question of connection between the STS thesis and technical project must answered, and one will find that the overlap and relationship between the two are striking. First, many of the conclusions reached in the report on Chinese AI can, and should, be extrapolated to the use of a full scale AGI. The issues that the CCP is encountering, as well as the ways in which China itself is adapting and responding to increased AI usage, will not go away if an AGI is reached and used. This is perhaps the most salient connection, as the problems society is encountering with AI will only be exacerbated by adding a general-purpose component. Next, the AGI itself, hardly in existence currently, will be subject to the same societal restaurants and influences that all inventions are, according to Social Construction of Technology theory. Therefore, an AGI, if it is ever developed, will reflect to some degree the society and time in which it is conceived – it will be different in implementation, design, and internal construction if developed at MIT in 2047 versus Tsinghua University in 2022, just based on the unique atmosphere of each place. Finally, AGI should be thought of not as separate per se of AI, but more of inheriting its flaws and issues. If the problems that AI itself are giving rise to, such as algorithmic bias, state surveillance, and privacy concerns are not absolved, then an AGI will simply intensify them. Again, this means that any time the term “AI” is used in the STS thesis, AGI could be substituted in and for the most part the conclusions reached would still hold. Due to its purely theoretical nature currently, AGI cannot be analyzed in the same way AI can in China, as it does not exist yet for its impact to be measured. In this sense, however, AI can be seen as a proxy to the socio-political effects of AGI, if it is ever created and used.
