Science Snippets: AI is Destroying the Living Planet
Draft script:
Artificial Intelligence is all the rage. The various platforms supporting AI contain abundant information. However, AI is a product of human knowledge and intellect. As a result, AI is constrained by human knowledge and human intelligence. Now you know why I have been referring to AI as arrogant idiot for many years.
I’m sure AI has its place. For those seeking the ability to collate abundant information, I suspect AI will be valuable. Organizing the information might also prove to be a good task for AI. Because most of my work during the last decade-and-a-half has depended on my ability to collate and organize the work of other scholars, it seems AI could be useful. On the other hand, I have become quite adept at collating and organizing information myself, a process that began when I was a graduate student.
My grad-school days were characterized by considerably less modern technology than we have today. There was no AI, at least not in the public domain. I typed my master’s thesis on the campus’s mainframe computer. Each draft would require time to print, and the printer was a five-minute walk from my shared office. As a result, I carefully reviewed each draft after it laboriously took time to print.
Fast forward a few years, and I was writing my Ph.D. dissertation on a personal computer. It was not a laptop, but I only had to walk a few hundred meters to gain access to a computer and printer at the nearby library. I retained the information on a 5-1/4” floppy disk. Within a year or so, the disk was replaced with a sturdier, 3-1/2” version.
One of my graduate advisers told me several times that I needed to retain my data on cards and on film. These were the primary sources of data retention back in the 1980s. I took his advice and hauled thousands of punched cards and an 18” reel of film from Lubbock, Texas to Athens, Georgia and then to College Station, Texas. Not long after I hauled these valuable stacks of cards and the reel of film to Tucson, Arizona, I asked the campus tech dude how I could transfer them to more contemporary forms of data collection. After a long conversation filled with him asking me for clarification, the campus tech dude concluded I was trapped in a previous century. This Research-One university had no way to read data from punched cards or 18” reels of film. So much for my graduate adviser and his advice.
Fast-forward a few decades after I studiously kept track of that seemingly valuable data from my work as a graduate student. Retaining data has become a much bigger job. Not only are scientists dealing with much larger datasets, but the ability to store those data has become expensive. In addition, we now realize that the electrical demand produced by AI is enormous. The difficulty is described in a peer-reviewed paper.
The peer-reviewed paper is not open-access. My Emeritus status allows me to read it. I can assure you that the Abstract contains all the relevant information in this paper, which is barely more than four pages long.
“The boosting development of artificial intelligence … is contributing to rapid exponential surge of computing power demand, which results in the concerns on the increased energy consumption and carbon emission. To highlight the environmental impact of AI, a quantified analysis on the carbon emission associated with AI systems was conducted in this study, with the hope of offering guidelines for police [policy] maker to setup emission limits or studies interested in this issue and beyond. It has been discovered that both industry and academia play pivotal roles in driving AI development forward. The carbon emissions from 79 prominent AI systems released between 2020 and 2024 were quantified. The projected total carbon footprint from the AI systems in the top 20 of carbon emissions could reach up to 102.6 Mt of CO2 equivalent per year. This could potentially have a substantial impact on the environmental market, exceeding $10 billion annually, especially considering potential carbon penalties in the near future. Hence, it is appealed to take proactive measures to develop quantitative analysis methodologies and establish appropriate standards for measuring carbon emissions associated with AI systems. Emission cap is also crucial to drive the industry to adopt more environmentally friendly practices and technologies, in order to build a more sustainable future for AI.”
AI seems to be a classic case of tradeoffs between individuals and society. The vast majority of us would like somebody else to do the work on our behalf. AI seems to offer that potential. As with many other advances in technology, I suspect AI offers more than it can deliver. Scholarship demands that we do the work of critically gather, collating, and organizing information. We recently saw how AI was blindly used to indiscriminately assign trade tariffs on islands occupied only by penguins. Critical thinking must be part of the process of reaching conclusions. Counting on AI to complete this step is not a good idea.
I question this statement: "The vast majority of us would like somebody else to do the work on our behalf." Not that I don't believe it's true, because I believe it is. But I question if we really do want someone else to do the "work", whatever that might be. Because doing the work ourselves so often leads to personal satisfaction and growth. Where will our personal satisfaction/growth come from when we don't have to do the work, or the thinking?
The new technology will not save us from the old technology.