educability
Valiant's proposed defining human capability: the integration of learning from experience, chaining learned knowledge across contexts, and absorbing explicit theories transmitted by others.
As defined in 272 | Leslie Valiant on Learning and Educability in Computers and People
Where it comes up · 1
Most important claims · 8 of 8Search all claims for “educability” →
- The educability model contains no mechanism for checking whether a transmitted theory is correct, so humans are inherently easy prey to false and conspiracy theories; this is a deep human-level weakness, not merely a problem of new technologies like deepfakes, and the modern deluge of information makes it increasingly dangerous.272 | Leslie Valiant on Learning and Educability in Computers and People
- Being educable does not by itself produce useful humans or machines, because someone must decide the content of the education; making a system educable creates even more consequential decisions about what knowledge to provide, since the results differ by content—so educability carries great dangers, as both machines and humans can easily be educated to do undesirable things.272 | Leslie Valiant on Learning and Educability in Computers and People
- Despite talk of a science of education, the field is pursued largely as a matter of best practices rather than science; building a genuine theoretical superstructure would give assurance that techniques which work in one setting will transfer to others, and the educability framework opens new, more rigorous research questions that the concept of intelligence never provided.272 | Leslie Valiant on Learning and Educability in Computers and People
- Attempts since Darwin to identify a single feature unique to humans all fail because for any defined feature some animal can do it; what distinguishes humans is therefore the integration of the three educability components on top of each other, not any one capability in isolation.272 | Leslie Valiant on Learning and Educability in Computers and People
- Educability, the proposed feature distinguishing humans, is the integration of three capabilities: learning from experience (PAC learning), chaining together pieces of knowledge learned in different contexts (a basic form of reasoning), and absorbing explicit theories handed over directly by others without personal experience.272 | Leslie Valiant on Learning and Educability in Computers and People
- The educability capability may have evolved even before modern humans, possibly in a predecessor species around 300,000 years ago, and only became useful slowly via a snowballing effect requiring accumulated shared knowledge; the absence of any mutation spreading through the population in recent tens of thousands of years argues against a recent biological origin tied to agriculture.272 | Leslie Valiant on Learning and Educability in Computers and People
- A proper test of educability must avoid testing for any native skill or pre-existing knowledge and instead assess only the information acquired during the test itself—for example, what is learned within a one-hour window—which is the opposite of how testing is conventionally done.272 | Leslie Valiant on Learning and Educability in Computers and People
- Humans have a great facility to take in other people's theories and process them—drawing implications, planning, imagining future situations—without caring or knowing whether the theories are true, which is why we can engage equally with fact, fiction, or hypothetical future scenarios.272 | Leslie Valiant on Learning and Educability in Computers and People