Use the future to build the present
Neural Network Algorithms
Comment
Stakeholder Type
,
1Quantum Revolution& Advanced AI2HumanAugmentation3Eco-Regeneration& Geo-Engineering4Science& Diplomacy1.11.21.31.42.12.22.32.43.13.23.33.43.54.14.24.34.44.5HIGHEST ANTICIPATIONPOTENTIALAdvancedArtificial IntelligenceQuantumTechnologiesBrain-inspiredComputingBiologicalComputingCognitiveEnhancementHuman Applications of Genetic EngineeringRadical HealthExtensionConsciousnessAugmentation DecarbonisationWorldSimulationFuture FoodSystemsSpaceResourcesOceanStewardshipComplex Systems forSocial EnhancementScience-basedDiplomacyInnovationsin EducationSustainableEconomicsCollaborativeScience Diplomacy
1Quantum Revolution& Advanced AI2HumanAugmentation3Eco-Regeneration& Geo-Engineering4Science& Diplomacy1.11.21.31.42.12.22.32.43.13.23.33.43.54.14.24.34.44.5HIGHEST ANTICIPATIONPOTENTIALAdvancedArtificial IntelligenceQuantumTechnologiesBrain-inspiredComputingBiologicalComputingCognitiveEnhancementHuman Applications of Genetic EngineeringRadical HealthExtensionConsciousnessAugmentation DecarbonisationWorldSimulationFuture FoodSystemsSpaceResourcesOceanStewardshipComplex Systems forSocial EnhancementScience-basedDiplomacyInnovationsin EducationSustainableEconomicsCollaborativeScience Diplomacy

Sub-Field:

1.3.3Neural Network Algorithms

The architecture of the brain is inextricably intertwined with the algorithms that it performs; in many ways, the architecture is the algorithm. Furthermore, the brain is autonomous and uses distributed computing: it contains many densely interconnected local hubs that build a modular, hierarchical, but interconnected system. It also exhibits extraordinary plasticity: memory and experience (in performing tasks, for example) actually change its physical structure. This makes it even more difficult to separate its architecture from its function: the connections and topology of its neural network create its functionality, and these properties are fluid and flexible.

This has two consequences for brain-inspired computing. First, truly neuromorphic computing will be fundamentally different from the familiar Turing machines, where a range of programs can run on a single machine. With the algorithm physically implemented in the network structure of a neuromorphic computer, sequential programming ideas simply do not apply. Although this means we will have to compute in a new and different paradigm, there are clear upsides: brain-inspired computing may well open up avenues of information processing that are impossible with traditional machines.12

Second, architecture (hardware) choices affect the range of algorithms that can be run on each implementation. At the most basic level, the closer to normal silicon computing, the more flexible and reprogrammable the machine will be; the more analogue and physical, the more the algorithms are fixed by the architecture choice. The hardware-specificity of neuromorphic computing has limiting effects on both innovation and progress, and there is a need for standardisation in the way algorithms can be implemented. There is progress here: in October 2020, for instance, researchers laid out a conceptual foundation for designing algorithms and hardware separately.13

Future Horizons:

×××

5-yearhorizon

Multi-sense processing is consolidated

Engineers develop stretchable, smart, large-scale electronic skin; low-power and low-latency 3D vision, motion detectors; olfactory sensors and chemical sensors; sensors for electric fields and air currents. These are augmented with smart signal processing that enables efficient extraction of task-relevant information and its integration in multimodal concepts.

10-yearhorizon

Neuromorphic computing takes the AI crown in niche applications

Neuromorphic computing becomes the dominant computing framework for embodied AI — AI that works with sensory signals and motion control — as well as for human-machine interaction and computing on the interface to the physical world, including large-scale simulations. Conventional computers will only be used for storing and processing “vintage” digital data, with computing distributed between ultra-edge (the smart device), edge (computer in the room) and cloud (server), supported by ultra-fast and high-throughput wireless connectivity.

25-yearhorizon

The rules of thinking emerge

We have neuro-physics on the level of today’s physics, with models and explanations across different levels — from molecules to societies.

Neural Network Algorithms - Anticipation Scores

How the experts see this field in terms of the expected time to maturity, transformational effect across science and industries, current state of awareness among stakeholders and its possible impact on people, society and the planet. See methodology for more information.

GESDA Best Reads and Key Resources