Dataism

In the AI Story

Artificial intelligence is the Dataist sacrament—the ritual bringing the faith nearest to realization. A large language model is pure data processing: ingests billions of words (humanity's textual output), processes through billions of parameters (mathematical operations), produces outputs that are statistically probable continuations. It does not understand the data. It does not value the data. It processes the data. And the quality of processing is measured by plausibility—itself a data-derived metric. This is Dataism operationalized: value determined not by truth, beauty, moral significance, but by statistical consistency with prior data.

Harari acknowledges Dataism's seductive operational effectiveness. The algorithm works. It writes competent prose, generates functional code, produces medical diagnoses rivaling experienced physicians. In domain after domain, the Dataist claim is empirically validated: reduce the problem to data, process with sufficient compute, outputs are good enough—often better. What more proof does faith require? The danger is not that Dataism is false (it may be partially true) but that it is incomplete. The worldview has no vocabulary for experiences that do not register as data: satisfaction of understanding something difficult, pleasure of genuine human connection, value of solitude, importance of boredom. Boredom occupies a paradoxical position. Neuroscientifically, it is the soil in which attention and imagination grow—cognitive under-stimulation forcing the mind to generate its own content. Economically, it is productivity failure—a gap in the data stream, a bottleneck to eliminate.

Segal's Orange Pill amplifier framework is an implicit rejection of Dataist premises. The amplifier model subordinates information flow to purpose. The amplifier is neutral; it amplifies whatever it is fed. Quality of output depends not on data quantity or velocity but on quality of human input—care, judgment, intentionality, moral seriousness. Data is the medium. Purpose is the message. And purpose arises from consciousness—from caring about outcomes, preferring one future over another. The Dataist response: purpose is itself data (preferences as neurochemical outputs). Harari's counter: whether consciousness reduces to data is less important than the practical question of what happens to a society that acts as if it does. What happens is erosion of every value that cannot be quantified.

Origin

Harari coined 'Dataism' in Homo Deus (2015), positioning it as the successor ideology to humanism—the next 'religion' in a sequence running through animism, polytheism, monotheism, humanism. The term synthesizes threads from information theory (Claude Shannon), cybernetics (Norbert Wiener), complexity science, and Silicon Valley culture—the unreflective operationalization of 'more data, more processing' as self-evident good. Harari treats it not as a conspiracy but as an emergence—the uncoordinated convergence of engineers, economists, executives on a shared set of premises they do not recognize as ideological.

Key Ideas

Information flow as supreme value. Dataism evaluates everything—organisms, institutions, ideologies—by contribution to data processing. Maximizing throughput becomes self-justifying.

Operates as common sense, not doctrine. Most adherents do not recognize Dataism as a worldview. It is simply how things are done—'trust the data,' 'follow the metrics,' 'let the algorithm decide.'

AI as operational validation. Every domain where AI succeeds empirically confirms the Dataist premise: reduce to data, process sufficiently, outputs are good enough. The algorithm works.

No vocabulary for the unquantifiable. Rest, boredom, contemplation, beauty for its own sake—these do not register in Dataist frameworks. What cannot be measured is systematically excluded from optimization functions.

The Dataist always wins short-term. Organizations optimizing on Dataist principles outperform (by measurable metrics) those valuing unquantifiables. Long-term, the Dataist burns out—or discovers the metrics were wrong.