binding problem
In plain wordsThe puzzle of how a brain, or a neural network, keeps track of which features belong together — that the red goes with the apple and the green with the leaf — when they are processed separately.
The problem of how a system that processes features in distributed, partly separate channels combines them into coherent representations of distinct objects and their relations — representing a red square and a blue circle without confusing them with a blue square and a red circle. In neuroscience, proposed solutions include attention, in Anne Treisman's feature-integration theory, and synchronized neural firing, in Christoph von der Malsburg's proposal. In cognitive sciencecognitive science, n.: The scientific study of how minds work — how we reason, use language, remember and see — combining psychology, neuroscience, linguistics, philosophy, anthropology and artificial intelligence. and AI the related problem of variable binding — attaching roles to fillers, as in telling 'dog bites man' from 'man bites dog' — was pressed against 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.. Klaus Greff, Sjoerd van Steenkiste and Jürgen Schmidhuber (2020) argued that neural networks' failures of systematic generalizationsystematicity, 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. stem from unsolved binding, motivating object-centric architectures such as slot attention. Some theorists also tie the unity of conscious experience to binding.
Origin
Discussed in vision science and neuroscience from the 1980s, following Anne Treisman and Garry Gelade's feature-integration theory (1980) and Christoph von der Malsburg's correlation theory of brain function (1981). Greff and colleagues' 'On the Binding Problem in Artificial Neural Networks' (2020) revived the term for deep learning; earliest use of the phrase not established.
Related
- 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.
- 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.
- generalizationgeneralization, n.: How well a model copes with new examples it has never seen, rather than just memorizing its training data, like a student who understands the method, not just the answers.
- consciousnessconsciousness, n.: Having experiences: there being something it feels like to be you, as when you see red or feel pain. Whether any AI is or could become conscious is an open and fiercely debated question.
- attentionattention, n.: A way for an AI model to choose what to focus on. For each word it processes, the model scores how relevant every other word is and draws information mostly from the relevant ones.
Sources
- Treisman, A. M. & Gelade, G. (1980). A Feature-Integration Theory of Attention. Cognitive Psychology 12(1), 97–136.
- Greff, K., van Steenkiste, S. & Schmidhuber, J. (2020). On the Binding Problem in Artificial Neural Networks. arXiv:2012.05208. arxiv.org/abs/2012.05208
Mentioned in
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.