All that similar content abounds
We are facing a great glut of similar content these days. So much content is being created these days that it is hard for it not to just sort of end up being similar. However, in some ways, the great glut of similar content is in part because the models are taking all the trajectories that did exist and converging them toward a similar path. That is inherently problematic; it is like intellectual whack-a-mole where the outliers get algorithmically down voted, ignored, and ultimately pruned from the model. This is one of the reasons why I think the people who end up owning and maintaining the base knowledge graphs as we move forward will have unprecedented power to define what came before and was a part of history. That observation was not written to be flippant or casually thrown into argument. It really is the problematic foundation of what is about to get built going forward. We probably don’t exactly agree on the base history being defined in the knowledge graphs that underpin all these large language models and ultimately whatever powers the reasoning models of the future.
Underpinning that argument is a firm belief that people are going to end up building knowledge repositories for these models. I think of it like a collection of great, but ultimately divergent libraries. Some of these collections of knowledge are going to be very divergent from each other. Various countries have different views on history, and that is before politics are applied to something that should probably be less like quicksand and more like granite. You are already probably considering this argument a little bit more than before digging into this missive. Your future self is going to consume what the model considers to be settled fact, but you may or may not agree with that base layer. It may end up being even more problematic for the next generation that won’t have your experience, knowledge, skills, and abilities that were tempered pre-model supremacy.
Academics can no longer keep up with all the papers being published in a specific field. At one point, the big academic journals collected the ideas of the times and shared that with a base of academics that was interested in having that shared base of understanding and argument within the field. You could call that the basis of how the academy sorted and understood what was being learned and shared. Ultimately, this system is breaking down right now, and sometimes people write books trying to pull everything together within a field, and we are facing a reality where fragmentation is creating breakdowns in fields where pockets of understanding exist, but may have no context to what is happening in other adjacent areas or sometimes new areas. This example is just a microcosm of the larger problem described above. It’s a real problem. It will get worse before it possibly gets better.