Neural-Net-Driven Abstract Science Research
Progress in theoretical physics and pure mathematics is driven by two modes of development: leaps of insight and detailed analysis. By this I mean in the latter existing work is analyzed by a person seeking to advance it. The analysis continues until a leap of insight is attained, at which point the work shifts to detailing the insight and analyzing its correctness.
Technological development has somewhat aided the human side with both efforts. Data visualization has helped researchers find patterns more easily and inspire bursts of insight, and theorem-proving software has aided in details of validation. However, in these cases the human element is still required.
Most technological research on aiding scientific research is focused on streamlining the detail portions in order that the human element may focus on forming bursts of insight. With modern sophistication of neural net software, more focus should be placed on determining the exact nature of this insight so that it may be duplicated artificially.
Naively, these bursts are merely sophisticated pattern recognition and prediction, something computers can already do quite well. The only real stumbling block is that neural nets work "deductively" (from a base of infallible information and tweaking its results to match those) while humans work more "inductively" (from a base of fallible information that is tweaked more towards the unknown but correct results). More examination of these differences would be required before a system could be designed to duplicate the results of human insight.
Coupled with existing theorem-proving techniques, the human element of scientific progress in pure mathematics and theoretical physics could, in theory, be supplanted entirely by machine progress. If this pattern-recognition technique of the human mind could be formalized, the results of such digital research could even be "translated into human language" so that even though humans did not create the advances in those fields, they could be "taught by machines" the automated advances.
Whiteboard on Your Face
The technology already exists to put a clear (though low resolution) display on glasses. The main problems stopping widespread adoption of this device are price and resolution. The technology also already exists to couple this with a low-quality, tiny webcam equipped with line-recognition to move and orient the display in real-time according to how the wearer orients his head. Next, couple this with the technology I showed you using a wiimote to capture pointed IR radiation, and you get a display on which you can draw digital lines in real space. The lines you drew will move depending on how you orient your head, so any blank wall becomes a whiteboard. Next, couple this with Google Video-type internet technology, and share your display with other wearers of the technology.
New Millennium Higher Education
The existing university-based higher educational structure was first laid down centuries ago by various philosophers. At that time, the body of knowledge that was kept, added to, and passed down by each generation was extremely small by today's standards. Due to hundreds of years of research and study, and through multiple generations of information technology that is improving at an exponential rate, the previous paradigm of higher education should be reexamined and restructured for improved efficiency.
Problems with Existing Structure
- A wide array of courses are required, from literature to physics to philosophy and psychology, that do little to legitimately introduce a student to the topic discussed. Few students walk away learning anything significant about the topic, when the purpose of requiring those classes is to expose students to multiple different ways of thinking about the world around them.
Existing models of economics are poor representations of real-world activity. Particular models have subsections of the economy in which the can form predictions of events, assuming axioms of constancy that cannot realistically be assumed. Furthermore, an extremely basic economic model of industrial interaction, or even a cross-model method of calculation, has not been accomplished.
The formalism and success of quantum mechanics in modeling complex interactions efficiently and correctly has spurred a century of fruitful research in theoretical physics. Thus, the formalism of quantum mechanics could be employed in an economical model. Interactions could be modeled with commutation relations, reducing complex interactions to simple pair-wise defined terms. The role of the reduced Planck's constant would then act as a measure from which the model deviates from non-interacting models, such as those used today.