Integrated Information Theory
A theory of consciousness proposed by neuroscientist Giulio Tononi, which uses the value Φ (phi) to measure the degree of information integration within a system. The theory posits that the intensity of consciousness correlates positively with a system's integrated information quantity (Φ value), and is one of the mainstream theories in consciousness science.
Core Facts
Integrated Information Theory (IIT) was first systematically proposed by neuroscientist Giulio Tononi in 2004, using the quantitative measure Φ to gauge a system's capacity to integrate information.
Integrated Information Theory was proposed by neuroscientist Giulio Tononi, asserting that the degree of consciousness can be quantified by the Φ value, which measures the degree of integrated information in a system.
Timeline (last 90 days)
The Φ value of standard feedforward neural networks approaches zero, because information flows only in one direction and lacks recurrent integration.
A 2023 'adversarial collaboration' study involving multiple independent laboratories directly compared IIT and Global Workspace Theory, and the results provided decisive support for neither theory.
The precise calculation of Φ faces a combinatorial explosion problem that already exceeds the capacity of current computers even for networks of tens of nodes, while the human brain has 86 billion neurons.
Integrated Information Theory (IIT) was first systematically proposed by neuroscientist Giulio Tononi in 2004, using the quantitative measure Φ to gauge a system's capacity to integrate information.
By IIT standards, feedforward neural networks (including the standard Transformer inference process) may have very low Φ values because information flows primarily in a single direction and lacks recurrent causal interaction.
In September 2023, more than 120 consciousness researchers jointly signed an open letter criticizing IIT's scientific validation methods, arguing that its core predictions are difficult to rigorously falsify.
IIT is built on five fundamental axioms: intrinsic existence, compositionality, information, integration, and exclusion.
According to IIT, in feedforward neural networks (including the standard Transformer inference process), information flows primarily in a single direction, lacking rich recurrent causal interactions, resulting in a degree of information integration far below that of the biological brain, and thus current AI architectures are fundamentally unsuited to generating consciousness.
GWT holds that consciousness is a functional state that can arise in any system implementing that function, while IIT holds that consciousness is an intrinsic property of the physical substrate, meaning certain computational architectures may lack consciousness even if they are functionally equivalent.
Integrated Information Theory was proposed by neuroscientist Giulio Tononi, asserting that the degree of consciousness can be quantified by the Φ value, which measures the degree of integrated information in a system.
1 more timeline events
All Facts (11)
Integrated Information Theory (IIT) was first systematically proposed by neuroscientist Giulio Tononi in 2004, using the quantitative measure Φ to gauge a system's capacity to integrate information.
75%VerifiedIntegrated Information Theory was proposed by neuroscientist Giulio Tononi, asserting that the degree of consciousness can be quantified by the Φ value, which measures the degree of integrated information in a system.
75%UnverifiedThe precise calculation of Φ faces a combinatorial explosion problem that already exceeds the capacity of current computers even for networks of tens of nodes, while the human brain has 86 billion neurons.
60%UnverifiedGWT holds that consciousness is a functional state that can arise in any system implementing that function, while IIT holds that consciousness is an intrinsic property of the physical substrate, meaning certain computational architectures may lack consciousness even if they are functionally equivalent.
60%UnverifiedThe Φ value of standard feedforward neural networks approaches zero, because information flows only in one direction and lacks recurrent integration.
50%UnverifiedA 2023 'adversarial collaboration' study involving multiple independent laboratories directly compared IIT and Global Workspace Theory, and the results provided decisive support for neither theory.
50%UnverifiedBy IIT standards, feedforward neural networks (including the standard Transformer inference process) may have very low Φ values because information flows primarily in a single direction and lacks recurrent causal interaction.
50%UnverifiedAccording to IIT, in feedforward neural networks (including the standard Transformer inference process), information flows primarily in a single direction, lacking rich recurrent causal interactions, resulting in a degree of information integration far below that of the biological brain, and thus current AI architectures are fundamentally unsuited to generating consciousness.
50%UnverifiedIIT is built on five fundamental axioms: intrinsic existence, compositionality, information, integration, and exclusion.
50%Unverified全局工作空间理论和整合信息理论（IIT）是当前主流意识理论，但彼此在意识的必要条件上无法达成一致
50%UnverifiedIn September 2023, more than 120 consciousness researchers jointly signed an open letter criticizing IIT's scientific validation methods, arguing that its core predictions are difficult to rigorously falsify.
50%