I've just finished reading This Idea Must Die - Scientific Theories That Are Blocking Progress, the 2014 Edge question. What scientific idea is ready for retirement (so that science can advance)?
About 180 answers of visionary thinkers are collected in the book. The Big Bang was the first moment of time. Entropy. Infinity. Cause and Effect. Essentialism. Free will. Cognitive agency. Robot companions. Nature = Objects...to select a few. NN Taleb takes down the standard deviation (again) and Emanuel Derman questions the power of statistics (I've read the books of both motivated by my engagement in quant finance).
Ideas that block progress
But what I'm interested in more general: few radical innovations are developed without first abandoning old ideas and systems. And some ideas do really block progress:
One size fits all - in-search-of-perfection seduces us to select frameworks, models, methods, programming languages...that fit "everything". But what we need is multi-model and -method approaches, hybrid, multi-paradigm programming…down to platform-agnostic systems.
Data Science replaces everything - the opposite is true…it's hard to work in the data salt mines and the danger of being fooled by noise is ubiquitous. Intelligently combining modeling and data driven methods is the right way to go.
Tightly coupled complex systems - A complex system can have untended consequences and tightly coupled means there is not enough time to react to them. Diversify and decentralize your system, make it agile by increasing its adaptability and optionality. Make intelligent independent system components that are accurate and robust, but flexible. Let them co-exist and co-evolute.
Product = technology - programming languages, engines that implement them, data frameworks, deployment systems, clouds…are technologies enabling many transformative developments of products,
AI = neural nets - first, the expert system thinking of AI died, then the idea of "Artificial Life"…now we think: because our brains have neurons AIs are well represented by neurons too? IMO, AIs are built of a set of techniques of mathematics, engineering, science…not a post human "species". Artificial neural nets will be of great help in this multi-factor approaches. Behavior must ve quantified and knowledge must be computational.
Symbolic (exclusive) or numerical computation - both…the mathematical discipline is asymptotic mathematics: us symbolic constructors and builders to get models of good nature for numerical solvers and use symbolic computation for dynamic result interpretation.
In short, forget the (exclusive)OR, believe in the AND effect.