neuro-symbolic AI
In plain wordsAI that combines neural networks, which learn patterns from data, with symbolic methods, which follow explicit rules and logic, aiming for systems that both learn flexibly and reason reliably.
The family of approaches that integrate neural networksneural network, n.: A kind of computer program that learns a task from examples by tuning thousands to billions of connections between simple number-crunching units, an arrangement loosely inspired by nerve cells in the brain., which learn statistical patterns from raw data, with symbolicsymbolic AI, n.: An approach to AI that writes knowledge down as human-readable symbols and rules, such as 'every bird has wings', and reasons by manipulating them step by step instead of learning patterns from data. components — logic, rules, programs, knowledge graphsknowledge graph, n.: A huge web of facts stored as connections between things — 'Paris is the capital of France', 'Marie Curie won a Nobel Prize' — that computers can search, combine and reason over., search procedures, solvers — which supply explicit structure, exact reasoning, compositionality and interpretability. Integration takes many forms: neural networks that call symbolic solvers or toolstool use, n.: An AI model's ability to reach beyond its own knowledge by operating outside tools — searching the web, running code, using a calculator or an app — and folding the results into its work.; symbolic search guided by learned heuristics, as in AlphaGoAlphaGo, prop. n.: A DeepMind program that in 2016 beat Lee Sedol, one of the world's best players of the board game Go, a decade or more before many experts expected a computer to manage it.'s Monte Carlo tree searchMonte Carlo tree search, n.: A way for a game-playing program to choose a move by simulating many possible continuations, focusing on the most promising lines, and picking the move that worked out best on average.; neural models that emit programs or logical forms executed symbolically, as in program-aidedprogram-aided language model, n.: A technique in which an AI solves a math or logic problem by writing a small computer program and letting a computer run it, instead of doing the arithmetic in its head. reasoning; differentiable relaxations of logic used as layers or training constraints; and neural modules inside a symbolic architecture. Prominent recent systems pair a language modellarge language model, n.: A computer program trained on enormous amounts of text to predict the words that come next. At vast scale, that simple skill lets it chat, answer questions, translate, summarize and write code. that proposes candidate steps with a symbolic engine that deduces their consequences or a proof checker that verifies them, as in AlphaGeometryAlphaGeometry, prop. n.: A Google DeepMind system from 2024 that solves olympiad geometry problems nearly as well as top human competitors, by pairing a language model that suggests constructions with a logic engine that checks every step. and AlphaProofAlphaProof, prop. n.: A Google DeepMind system that learned to write computer-checked mathematical proofs by trial and error. In 2024 it helped an AI reach silver-medal level at the International Mathematical Olympiad for the first time. respectively. Proponents argue that such hybrids are needed for robust reasoning and generalization; skeptics expect scaled neural networks to acquire these abilities internally.
It's a neuro-symbolic pipeline: the model writes Datalog rules and the engine evaluates them.
Origin
Research on 'neural-symbolic' integration dates from the late 1980s and early 1990s and was consolidated in Artur d'Avila Garcez, Krysia Broda and Dov Gabbay's Neural-Symbolic Learning Systems (2002). The label 'neuro-symbolic' (or 'neurosymbolic') spread widely from about 2019, with the MIT–IBM Neuro-Symbolic Concept Learner and Henry Kautz's 2020 AAAI lecture 'The Third AI Summer', which classified integration types; Gary Marcus has long argued for hybrid models.
Usage note
The label covers very different systems, from a chatbot that calls a calculator to differentiable theorem provers, so general claims about 'neuro-symbolic AI' deserve scrutiny. The symbolic–connectionist debate and the neats and scruffiesneats and scruffies, phr.: A joking split of AI researchers into 'neats', who want clean mathematical theories that can be proved correct, and 'scruffies', who build messy collections of tricks that simply work. dispute are its historical background.
Related
- symbolic AIsymbolic AI, n.: An approach to AI that writes knowledge down as human-readable symbols and rules, such as 'every bird has wings', and reasons by manipulating them step by step instead of learning patterns from data.
- connectionismconnectionism, n.: The view that minds and intelligent machines are best built from huge networks of simple, neuron-like units whose connection strengths are learned, rather than from explicit rules and symbols.
- program synthesisprogram synthesis, n.: Getting a computer to write a program automatically from a description of what the program should do — a logical specification, a few input–output examples or a request in plain language.
- AlphaGeometryAlphaGeometry, prop. n.: A Google DeepMind system from 2024 that solves olympiad geometry problems nearly as well as top human competitors, by pairing a language model that suggests constructions with a logic engine that checks every step.
- AlphaProofAlphaProof, prop. n.: A Google DeepMind system that learned to write computer-checked mathematical proofs by trial and error. In 2024 it helped an AI reach silver-medal level at the International Mathematical Olympiad for the first time.
- Markov logic networkMarkov logic network, n.: A way of combining logical rules with probabilities, so that a rule like 'friends usually like the same music' can sometimes be broken, with each rule given a weight saying how strongly it holds.
- tool usetool use, n.: An AI model's ability to reach beyond its own knowledge by operating outside tools — searching the web, running code, using a calculator or an app — and folding the results into its work.
- Gary MarcusGary Marcus, prop. n.: An American cognitive scientist and leading critic of deep learning, who argues that today's AI cannot reason reliably and needs to be combined with symbolic methods.
- connectionismconnectionism, n.: The view that minds and intelligent machines are best built from huge networks of simple, neuron-like units whose connection strengths are learned, rather than from explicit rules and symbols.
Sources
- d'Avila Garcez, A. S., Broda, K. B. & Gabbay, D. M. (2002). Neural-Symbolic Learning Systems: Foundations and Applications. Springer.
- Mao, J., Gan, C., Kohli, P., Tenenbaum, J. B. & Wu, J. (2019). The Neuro-Symbolic Concept Learner. ICLR 2019.
- Kautz, H. (2022). The Third AI Summer: AAAI Robert S. Engelmore Memorial Lecture. AI Magazine 43(1), 105–125.
Mentioned in
symbolic AIsymbolic AI, n.: An approach to AI that writes knowledge down as human-readable symbols and rules, such as 'every bird has wings', and reasons by manipulating them step by step instead of learning patterns from data., connectionismconnectionism, n.: The view that minds and intelligent machines are best built from huge networks of simple, neuron-like units whose connection strengths are learned, rather than from explicit rules and symbols., Gary MarcusGary Marcus, prop. n.: An American cognitive scientist and leading critic of deep learning, who argues that today's AI cannot reason reliably and needs to be combined with symbolic methods., good old-fashioned AIgood old-fashioned AI, n.: A nickname for the classic style of AI, dominant from the 1950s to the 1980s, that tried to make computers intelligent by writing knowledge and rules down by hand., program synthesisprogram synthesis, n.: Getting a computer to write a program automatically from a description of what the program should do — a logical specification, a few input–output examples or a request in plain language., AlphaGeometryAlphaGeometry, prop. n.: A Google DeepMind system from 2024 that solves olympiad geometry problems nearly as well as top human competitors, by pairing a language model that suggests constructions with a logic engine that checks every step., AlphaProofAlphaProof, prop. n.: A Google DeepMind system that learned to write computer-checked mathematical proofs by trial and error. In 2024 it helped an AI reach silver-medal level at the International Mathematical Olympiad for the first time., DatalogDatalog, prop. n.: A small logic language for asking questions of a database with rules, including rules that refer to themselves, such as 'X is an ancestor of Y if X is a parent of an ancestor of Y'., Douglas LenatDouglas Lenat, prop. n.: An American computer scientist who spent four decades building Cyc, a giant hand-written database of everyday common sense meant to let computers reason like people., inductive logic programminginductive logic programming, n.: A kind of machine learning that discovers general logical rules — such as 'X is a grandparent of Z if X is a parent of a parent of Z' — from examples and background knowledge., knowledge graph embeddingknowledge graph embedding, n.: A way of turning the things and relations in a knowledge graph into lists of numbers, so a computer can guess missing facts from patterns, such as predicting where someone was born., Markov logic networkMarkov logic network, n.: A way of combining logical rules with probabilities, so that a rule like 'friends usually like the same music' can sometimes be broken, with each rule given a weight saying how strongly it holds., MRKL systemMRKL system, n.: An early design, pronounced 'miracle', in which a language model acts as a switchboard, sending each part of a question to the right specialist — a calculator, a database, a search tool — and combining the answers., neats and scruffiesneats and scruffies, phr.: A joking split of AI researchers into 'neats', who want clean mathematical theories that can be proved correct, and 'scruffies', who build messy collections of tricks that simply work., program-aided language modelprogram-aided language model, n.: A technique in which an AI solves a math or logic problem by writing a small computer program and letting a computer run it, instead of doing the arithmetic in its head., systematicitysystematicity, n.: The way thinking and language come in connected packages: if you can understand 'the dog chases the cat', you can also understand 'the cat chases the dog'. Whether neural networks manage this has long been debated.