AI and PhD papers: a shortcut to nothing

Transcript

The research process is about learning and discovery, but when the machine is in that mix, mimicking the learning and manufacturing language constructs that look like discovery, then no actual learning or discovery happens. Here's the question. What is the purpose of scientific papers? I ask because in the discourse around AI being used in science and AI systems eventually doing their own research and producing their own papers, it seems as if a lot of people think the purpose. Of science and research is to produce papers. That's not the purpose of science though, and it's also not what research is. That's almost like saying the purpose of learning math is to be able to do a math test. The paper is the end result of a process that involves many different components, and in the end the researcher or researchers synthesize everything they've done into a paper to document the work they did. What? They went in thinking would happen the hypothesis what they found in their data and then make some form of conclusion to share with others. And the paper serves multiple purposes. It is a way of conveying all that information to other researchers. It's also a way of proving that work was actually done to justify it in a capitalist sense. I spent time and money. This is the end result. And it's a means for academics to show their academic prowess and be able to build their academic credentials. Over time. So the paper isn't the point. The point is the actual research. Maybe doubly so for PhD researchers because they are building their skills of doing the research part, and then they produce a paper at the end to prove that they did this work, consider that, and then put the AI into the mix. If the AI starts doing a bunch of that research work, or starts writing the papers, or does all of it, then no research is happening and no one is learning anything. The research process is about learning and discovery, but when the machine is in that mix, mimicking the learning and manufacturing language constructs that look like discovery, then no actual learning or discovery happens. The entire meaningful part of research is offloaded to a machine that draws no meaning from that research and is not able to come up with novel ideas or do anything about them. So while AI can be helpful as a tool as part of a research toolkit. The very real risk here is that we offload the actual research to an AI, therefore the researchers aren't learning how to do the proper research, and we lose all of this great research that would otherwise have been created if people were doing it. And if the AI of rights the papers themselves, they're just doing math exams instead of anyone learning math.