It is surely a commonplace that the solution of difficult intellectual problems requires focus. This is of course, just the opposite of what we see in LLMs, which are trained on massive datasets and are not only costly to train, but also not cheap to apply to simple problems. Problems which require deep thinking are explicitly provided for in many models by getting the LLM to think twice (or more) and reflect upon its results. Away from the generality of LLMs, AI has lately had success in narrower domains by using more focussed methods, notably the deepmind alpha-zero has shown the benefits of focus upon the problem domain at hand.
Alongside the effectiveness of focus in some narrow domains, it sometimes happens that the solution of a relatively broad problem set may be facilitated by focus on certain key subproblems. A special case of this comes with the idea of “the singularity”, the point at which AI becomes capable of redesigning itself leading to progressive acceleration of the AI design cycle and hyperexponential grown in the capabilities of artificial superintelligence. Here we see that focus on the problem of AI design may be expected to advance capability in design across the board, and the potential suggests talking of this as a benefit of singular focus.
It may be that the benefits of focus, and the possibility of singular focus, can be turned into architectural models and strategic plans, and that is the purpose of this note.
There are two kinds of focal thinking which contribute to the proposed architecture. These come as a focal tower and a focal heirchy which are discussed in turn.
To that end, I identify an ultimate purpose, and a fundamental technology, and show how focal methods can be used to construct a tower which leads us as efficiently as possible from the fundamentals to the ends.
The tower consists of a number of “focal layers”. Each later has some notion of competence in addressing a range of problems in a particular context, and a special subdomain competence in which facilitates the delivery of competence in the broader domain. As we advance up the stack, the level and breadth of the relevant capabilities advances, and the context in which those competences are delivered becomes less supportive.
The stack can have any number of layers, so a sketch with a small number is the best place to start. From the bottom up:
Logic and Software Engineering_ In this the broad capability is in logical reasoning in a logical system suitable for all deductive reasoning about declarative knowledge. The context in which that capability is exercised is contemporary accelerated computing. The focus, which is a singular focus, is on reasoning about programs, since the embraces the capability to completely re-engineer this kind of reasoning capability.
The focal tower as described above correlates with the epistemological stack in the following way.