Machine learning is the science of getting computers to act without being explicitly programmed. In the past decade, machine learning has given us self-driving cars, practical speech recognition, effective web search, and a vastly improved understanding of the human genome. An alternative to Machine Learning is directly embedding intuitions into computer programs.
In Machine Learning, you begin at the bottom of the DIKW pyramid and walk upwards (from an epistemological point of view). There are no data-allergic machine learners thus far. There simply isn't any ML sans data.
Hence, the perfect bete noire to Machine Learning would be what would be called Kekule-Archimedean learning, which involves a strict downward walk with regard to the DIKW pyramid.
You don't begin with data.
Instead you begin with a brilliant piece of intuition (like a Eureka moment or a Kekulean dream) emanating from deep intellectual thought not necessarily driven by evidences (read data) and use data only in the concept-validation phase.
In that sense, it is very Platonic where 'learning is the development of ideas buried deep in the soul, often under the midwife-like guidance of an interrogator'. Source: Platonic epistemology.
Most hypothetico-deductionist paradigms fall somewhere between the two worlds.
cortical.io presents semantic fingerprints and semantic maps as an alternative to machine learning in How Mimicking Brain Function is Revolutionising NLP.
"There’s a zillion probabilistic techniques (for example, look for the words Bayes or Markov in the name of the algorithm), kernel methods (like SVM, decision trees/random forest, Gaussians, PCA, can-cor…), a zillion kinds of reinforcement learning that has nothing to do with ANNs, artificial reasoning a.k.a. “good old fashioned AI”, many path-planning and intelligent control-systems methods that correspond to “classical AI” (not the same as GOFAI), alife (swarms, cellular automata…), agents and chaos systems (I think those are either the same or overlapping categories depending on your definition of agent). You may or may not include semantic-web technologies in your definition of AI (Berners-Lee doesn’t, but they’re pretty important to question-answering by Google, Siri/WolframAlpha, at least the original version of IBM Watson that blew everyone away in 2011). Also, in my personal opinion AI probably includes large subsets of the field of advanced cryptography, plus of the field of advanced numerical methods. Finally I’m not sure there’s any clear boundary between AI and a lot of topological search methods - sometimes it seems the only boundary is the engineering school versus other across the way in the math department, but they’re both seeking a policy that optimizes the value function (reinforcement learning and linear dynamic programming) or they’re crawling some graph having local and global optima - the number of methods for doing that might exceed everything else I wrote above. "
