How LLMs Are Changing Research

Humans are sophisticated librarians today. At this point in time it is simply more efficient to recall that something is knowable and where to find that knowledge than it is to store a complete record of that knowledge all together locally. For instance, it is probably not nearly as well known that the genealogy of Christ contains a link through Amminadab’s son Nahshon, as it is that the genealogy of Christ is contained at the beginning of the book of Matthew (at least one of them is anyway).

We do not store the information itself unless it is simply so important for us to have fast access to that information as it is for us to look up the information ourselves. So we have this constant battle going on in our minds, a battle which seeks to maximize efficiency overall. The conflict places whether it takes more energy and effort for us to simply memorize the information itself against remembering whether/where the information is attainable.

Back when we had to remember book and paper titles and authors in order to discover where the important information lies, it was often easier to just allow the information to be memorized, especially when to physically put the information before our eyes actually required moving our physical bodies in a significant way (ie. to go to the library, or to even travel to another library to acquire a text). At some point over the past twenty years while the Internet was maturing, it became so ubiquitous to repeat information that almost all of the proprietary information which we tried to guard behind feeble pay walls could simply not be contained, and the majority of what was worth knowing trickled to the top of the human combined social consciousness on the web. Not only this, but our largest companies, with the most financial resources spent all of this time learning how to turn queries into results as quickly as possible for the entire species, which certainly accelerated our ability to acquire information. Suddenly we no longer needed to know the names of the resources that contained the information that we utilized. We could simply query almost all resources at the same time. This was very valuable, but more valuable still is what is happening now.

LLMs are making the acquisition of information faster and more fluid. It is clear that if someone would like the answer to a particular query they could ask an LLM or they could sift through the mountains of information themselves. We have also been busy over the past 100 years developing tens of thousands of statistical details that no statistician can ever hope to have complete command and mastery over. In several ways this is unfortunate as it has precluded anyone from actually becoming a true master of all of the details of the field of Statistics, but on the other hand it is very hopeful, showing us that the philosophy of mathematics in statistics runs so deep that it has yet many new avenues to discover and explore. But all of these details are now handled by a layer of intelligence sitting in between the vast libraries of manuscripts we have created and our own most precious processing units - our brains.