# Artificial Intelligence > [!info] Gateway page > This is the vault's most-linked stub-turned-gateway. It gives the working definition, the distinctions, and the map. The full research cluster — technical grounding plus political economy — lives at [[Artificial Intelligence Hub]]. Artificial intelligence names a moving target: the set of computational tasks that currently look like they require human cognition. The definition slides, because capabilities that get solved stop counting. Chess, optical character recognition, and speech transcription were all AI until they worked, at which point they became software. This is worth holding onto, because it means "AI" in any given sentence is doing one of three jobs, and conflating them produces most of the bad arguments in the field. **AI the technology** is a set of statistical methods. Nearly everything in the current boom is [[Machine Learning]] — specifically [[Deep Learning]], in which multi-layer networks learn patterns from enormous training corpora rather than executing rules a programmer wrote. The rival paradigm, symbolic AI, dominated from the 1950s to the 1980s and lost. That history matters because the field has already had two boom-bust cycles, and the mechanism both times was the same: extravagant promises, institutional overinvestment, correction that destroyed funding for the productive research too. **AI the industry** is an extraordinarily concentrated market. Training a frontier model requires capital expenditure at a scale only a handful of firms can sustain, proprietary data at a scale only incumbents possess, and a talent pool small enough that a few hundred researchers moving between companies reshape the field. [[OpenAI]], [[Google]], [[Anthropic]], [[Meta]], [[Microsoft]] and [[xAI]] sit at the model layer; [[NVIDIA]] holds a near-monopoly one layer down, and the fabrication capacity below that funnels through a handful of facilities. This is the single most load-bearing fact in the vault's AI coverage: the technology's political significance follows from its concentration, not from its capability. **AI the political project** is the part this vault actually tracks. The people building frontier AI overlap heavily with the people funding the political movements documented here — an overlap that is not incidental but constitutive, running through [[Peter Thiel]]'s network, [[Elon Musk]]'s holdings, the defense-tech firms ([[Palantir]], [[Anduril]]) selling perception and targeting systems to the state, and the [[AI Political Spending Ecosystem]] now deploying nine-figure sums against state-level regulation. Two distinctions prevent most category errors here. **Capability is not deployment.** What a model can do in a benchmark and what a system does when embedded in an institution are different questions with different politics. [[Algorithmic Bias]] in a sentencing tool is not a story about model capability; it is a story about a public agency outsourcing a judgment and inheriting a training set. Most of the harm the vault documents lives at the deployment layer. **The safety debate has two camps that are usually mistaken for one.** [[AI Doomerism]] — the existential-risk tradition running through [[Nick Bostrom]], [[MIRI]], and the rationalist community — argues that sufficiently capable systems pose a species-level threat and that the priority is alignment research. The present-harms tradition — [[Timnit Gebru]], [[Shoshana Zuboff]], the algorithmic-accountability literature — argues that the existential framing is itself a political move, drawing scrutiny away from labor displacement, surveillance, and discrimination happening now, and conveniently locating the danger in a future only the labs can navigate. The vault treats both as live positions and neither as the neutral one. [[Effective Accelerationism]] is a third position that rejects the premise of both. One caveat the vault holds to: this is a knowledge base about the authoritarian right, so its AI coverage is scoped to power, money, and control. That is a deliberate framing, not a claim that AI is reducible to those things. ## Where to go from here | Destination | What's there | |---|---| | [[Artificial Intelligence Hub]] | The full cluster — technical grounding, subfields, history, and the political economy in depth | | [[AI Lexicon]] | Defined terms; the vocabulary layer | | [[Machine Learning]] · [[Deep Learning]] · [[Generative AI]] · [[Foundation Model]] | How the technology actually works | | History of Artificial Intelligence · [[Artificial Intelligence Timeline]] | The two prior boom-bust cycles, and the current one | | [[Tech Oligarchy]] · [[AI Sector]] · [[AI Political Spending Ecosystem]] | Who owns it and what they are buying politically | | [[Compute Sovereignty]] · [[Chip Export Controls]] · [[Tech Decoupling]] · [[Stargate]] | The hardware chokepoint and its geopolitics | | [[AI Safety]] · [[AI Doomerism]] · [[AGI]] · [[Superintelligence (Book)]] | The existential-risk tradition | | [[Algorithmic Bias]] · [[AI Ethics]] · [[Facial Recognition]] · [[Surveillance Capitalism]] | The present-harms tradition | | [[AI Action Plan]] · [[AI Arms Race]] | Federal policy and the competition framing | | [[Executive Order 14434]] | The September 2026 order replacing "artificial intelligence" with "Super Intelligence" across the executive branch | ## See also [[Tech Oligarchy]] · [[Surveillance State]] · [[Silicon Valley]] · [[Effective Accelerationism]] · [[Large Language Model]] · [[Palantir]] · [[Anduril]]