arXiv subjects

Robert Worden

Publications and source records attributed to Robert Worden.

15 recordsLinked to original sources

arXiv2026

AI and World Models

While large neural nets perform impressively on specific tasks, they are unreliable and unsafe, as is shown by the persistent hallucinations of large language models. This paper shows that large neural nets are intrinsically unreliable, because it is not possible to make or validate a tractable theory of how a neural net works. There is no reliable way to extrapolate its performance from a limited number of test cases to an unlimited set of use cases. To have confidence in the performance of a neural net, it is necessary to enclose it in a guardrail which is provably safe, so that whatever the neural net does, there cannot be harmful consequences. World models have been proposed as a way to do this. This paper discusses the scope and architecture required of world models. World models are often conceived as models of the physical and natural world, using established theories of natural science, or learned regularities, to predict the physical consequences of AI actions. However, unforeseen consequences of AI actions impact the human social world as much as the physical world. To predict and control the consequences of AI, a world model needs to include a model of the human social world. I explore the challenges that this entails. Human language is based on a Common Ground of mutual understanding of the world, shared by the people conversing. The common ground is an overlapping subset of each persons world model, including their models of the physical, social and mental worlds. LLMs have no stable representation of a common ground. To be reliable, AI systems will need to represent a common ground with their users, including physical, mental and social domains.

q-bio.NC↗

arXiv2025

Parallel Neuron Groups in the Drosophila Brain

The full connectome of an adult Drosophila enables a search for novel neural structures in the insect brain. I describe a new neural structure, called a Parallel Neuron Group (PNG). Two neurons are called parallel if they share a significant number of input neurons and output neurons. Most pairs of neurons in the Drosophila brain have very small parallel match. There are about twenty larger groups of neurons for which any pair of neurons in the group has a high match. These are the parallel groups. Parallel groups contain only about 1000 out of the 65,000 neurons in the brain, and have distinctive properties. There are groups in the right mushroom bodies, the antennal lobes, the lobula, and in two central neuropils (GNG and EB). Most parallel groups do not have lateral symmetry. A group usually has one major input neuron, which inputs to all the neurons in the group, and a small number of major output neurons. The major input and output neurons are laterally asymmetric. Parallel neuron groups present puzzles, such as: what does a group do, that could not be done by one larger neuron? Do all neurons in a group fire in synchrony, or do they perform different functions? Why are they laterally asymmetric? These may merit further investigation.

q-bio.NC↗

arXiv2025

Multi state neurons

Neurons, as eukaryotic cells, have powerful internal computation capabilities. One neuron can have many distinct states, and brains can use this capability. Processes of neuron growth and maintenance use chemical signalling between cell bodies and synapses, ferrying chemical messengers over microtubules and actin fibres within cells. These processes are computations which, while slower than neural electrical signalling, could allow any neuron to change its state over intervals of seconds or minutes. Based on its state, a single neuron can selectively de-activate some of its synapses, sculpting a dynamic neural net from the static neural connections of the brain. Without this dynamic selection, the static neural networks in brains are too amorphous and dilute to do the computations of neural cognitive models. The use of multi-state neurons in animal brains is illustrated in hierarchical Bayesian object recognition. Multi-state neurons may support a design which is more efficient than two-state neurons, and scales better as object complexity increases. Brains could have evolved to use multi-state neurons. Multi-state neurons could be used in artificial neural networks, to use a kind of non-Hebbian learning which is faster and more focused and controllable than traditional neural net learning. This possibility has not yet been explored in computational models.

q-bio.NC↗

arXiv2025

Consciousness Self and Language

Theories of consciousness depend on data, and it needs to be appropriate data, without overwhelming confounding factors. The reports of Minimal Phenomenal Experience (MPE) in [Metzinger 2024] relate to consciousness in a state purer than everyday consciousness, which may have fewer confounding factors. This essay suggests that the confounding factors, which are absent or diminished in MPE states, are related to language. The self which is absent in mindful states is a product of language. The link between language and MPE states is demonstrated by reference to the phenomenal reports in [Metzinger 2024]. Language, emotion, and mindfulness are analysed in terms of Bayesian pattern matching, or equivalently minimisation of Free Energy, using three types of pattern which are specific to humans. These types are the word patterns of language, self-patterns which drive our emotions and which are also a part of language, and mindful patterns. The practice of mindfulness involves learning mindful patterns, which compete with self-patterns and displace them, allowing mindful states to occur. Consequences of this picture for theories of consciousness, and their relation to MPE states, are explored.

q-bio.NC↗

arXiv2025

A Unified Theory of Language

A unified theory of language combines a Bayesian cognitive linguistic model of language processing, with the proposal that language evolved by sexual selection for the display of intelligence. The theory accounts for the major facts of language, including its speed and expressivity, and data on language diversity, pragmatics, syntax and semantics. The computational element of the theory is based on Construction Grammars. These give an account of the syntax and semantics of the worlds languages, using constructions and unification. Two novel elements are added to construction grammars: an account of language pragmatics, and an account of fast, precise language learning. Constructions are represented in the mind as graph like feature structures. People use slow general inference to understand the first few examples they hear of any construction. After that it is learned as a feature structure, and is rapidly applied by unification. All aspects of language (phonology, syntax, semantics, and pragmatics) are seamlessly computed by fast unification; there is no boundary between semantics and pragmatics. This accounts for the major puzzles of pragmatics, and for detailed pragmatic phenomena. Unification is Bayesian maximum likelihood pattern matching. This gives evolutionary continuity between language processing in the human brain, and Bayesian cognition in animal brains. Language is the basis of our mind reading abilities, our cooperation, self esteem and emotions; the foundations of human culture and society.

q-bio.NC↗

arXiv2024

Assessing the Brain Wave Hypothesis: Call for Commentary

It has been proposed that there is a wave excitation in animal brains, whose role is to represent three dimensional local space in a working memory. Evidence for the wave comes from the mammalian thalamus, the central body of the insect brain, and from computational models of spatial cognition. This is described in related papers. I assess the Bayesian probability that the wave exists, from this evidence. The probability of the wave in the brain is robustly greater than 0.4. If there is such a wave, we may need to re-think our whole understanding of the brain, in a break from classical neuroscience. I ask other researchers to comment on the wave hypothesis and on this assessment. In a companion paper, I outline possible ways to test it.

q-bio.NC↗

arXiv2024

Testing the Brain Wave Hypothesis

It has been proposed that there is a wave excitation in animal brains, whose function is to represent three-dimensional space around the animal as a working spatial memory. After surveying the evidence supporting the hypothesis, I discuss ways in which it can be tested. There are many ways to investigate it, theoretically and experimentally. They include connectome studies, computational modelling, experimental neuroscience, genomics and proteomics, studies of animal behaviour, and biophysics. If the wave exists, there is a compelling case to identify it as the source of consciousness. This would advance our understanding of one of the greatest scientific challenges of all time, while changing our view of the human mind.

q-bio.NC↗

arXiv2024

The Projective Wave Theory of Consciousness

Neural theories of consciousness face three difficulties: (1) The selection problem: how are those neurons which cause consciousness selected, from all the other neurons which do not? (2) the precision problem: how do neurons hold a detailed internal model of 3D space, as the origin of our spatial conscious experience? and (3) the decoding problem: how are the many distorted neural representations of space in the brain decoded, to give our largely undistorted conscious experience of space? These problems can all be addressed if the brains internal model of local 3D space is held not in neurons, but in a wave excitation (holding a projective transform of Euclidean space), and if the wave is the source of spatial consciousness. Such a wave has not yet been detected in the brain, but there are good reasons why it has not been detected; and there is indirect evidence for a wave, in the mammalian thalamus, and in the central body of the insect brain. The resulting projective wave theory of consciousness gives good agreement with the spatial form of our consciousness. It has a positive Bayesian balance between the complexity of its assumptions and the data it accounts for; this gives a basis to believe it.

q-bio.NC↗

arXiv2024

Spatial Cognition: a Wave Hypothesis

Animals build Bayesian 3D models of their surroundings, to control their movements. There is strong selection pressure to make these models as precise as possible, given their sense data. A previous paper has described how a precise 3D model of space can be built by object tracking. This only works if 3D locations are stored with high spatial precision. Neural models of 3D spatial memory have large random errors; too large to support the tracking model. An alternative is described, in which neurons couple to a wave excitation in the brain, representing 3-D space. This can give high spatial precision, fast response, and other benefits. Three lines of evidence support the wave hypothesis: (1) it has better precision and speed than neural spatial memory, and is good enough to support object tracking; (2) the central body of the insect brain, whose form is highly conserved across all insect species, is well suited to hold a wave; and (3) the thalamus, whose round shape is conserved across all mammal species, is well suited to hold a wave. These lines of evidence strongly support the wave hypothesis.

q-bio.NC↗

arXiv2024

Three Dimensional Spatial Cognition: Bees and Bats

The paper describes a program which computes the best possible Bayesian model of 3D space from vision (in bees) or echo location (in bats), at Marrs [1982] Level 2. The model exploits the strong Bayesian prior probability that most other things do not move, as the animal moves. 3D locations of things are computed from successive sightings or echoes, computing structure from the animals motion (SFM). The program can be downloaded and run. It also computes a tracking approximate model, which is more tractable for animal brains than the full Bayesian computation. The tracking model is nearly as good as the full Bayesian model, but only if spatial memory storage errors are small. Neural storage of spatial positions gives too high error levels, and is too slow. Alternatively, a 3D model of space could be stored in a wave excitation, as a Fourier transform of real space. This could give high memory capacity and precision, with low spatial distortion, fast response, and simpler computation. Evidence is summarized from related papers that a wave excitation holds spatial memory in the mammalian thalamus, and in the central body of the insect brain.

q-bio.NC↗

arXiv2024

The Requirement for Cognition, in an Equation

A model of the evolution of cognition is used to derive a Requirement Equation (RE), which defines what computations the fittest possible brain must make, or must choose actions as if it had made those computations. The terms in the RE depend on factors outside an animals brain, which can be modelled without making assumptions about how the brain works, from knowledge of the animals habitat and biology. In simple domains where the choices of actions have small information content, it may not be necessary to build internal models of reality; short cut computations may be just as good at choosing actions. In complex domains such as 3D spatial cognition, which underpins many complex choices of action, the RE implies that brains build Bayesian internal models of the animals surroundings; and that the models are constrained to be true to external reality.

q-bio.NC↗

arXiv2024

The Evolution of Language and Human Rationality

If language evolved by sexual selection to display superior intelligence, then we require conversational skills, to impress other people, gain high social status, and get a mate. Conversational skills include a Theory of Mind, a sense of self, self esteem and social emotions. To be impressive, we must converse fluently and fast. The syntax of an utterance is defined by fast unification of feature structures. The pragmatic skills of conversation are also learned and deployed as feature structures; we rehearse conversations as verbal thoughts. Many aspects of our mental lives (such as our Theory of Mind, and our social emotions) work by fast, pre conscious unification of learned feature structures, rather than rational deliberation. As we think, we use the Fast Theory of Mind to infer (unreliably) how a Shadow Audience will regard what we think, say, and do. These forces, which determine our motivations and actions, are less rational and deliberate than we like to suppose

q-bio.NC↗

arXiv2022

A Speed Limit for Evolution: Postscript

In 1995 I wrote a paper: "A Speed Limit for Evolution" whose main result was that evolution must proceed rather slowly, in accordance with the earlier views and intuitions of many authors. The abstract of the paper said: "The genetic information expressed in some part of the phenotype of a species cannot increase faster than a given rate, determined by the selection pressure on that part. This rate is typically a small fraction of a bit per generation". This result was derived in the presence of sexual reproduction and other effects such as temporarily isolated sub-populations. In 1999 David Mackay published a paper which apparently contradicted this result. In the abstract, he wrote "We find striking differences between populations that have recombination and populations that do not. If variation is produced by mutation alone, then the entire population gains up to roughly 1 bit per generation. If variation is created by recombination, the population can gain of the order of sqrt(G) bits per generation." Mackay proposed that there were outstanding evolutionary benefits to sexual reproduction, and that my result was too low by a very large factor. He later repeated this result in a textbook he wrote in 2003. The purpose of this note is to show that the key assumption of Mackays model, that "fitness is a strictly additive trait" is so unrealistic as to render his results irrelevant to any actual life form. In consequence, the speed limit I derived is still valid, and has important consequences for human cognitive evolution.

q-bio.PE↗

arXiv2021

A Theory of Language Learning

A theory of language learning is described, which uses Bayesian induction of feature structures (scripts) and script functions. Each word sense in a language is mentally represented by an m-script, a script function which embodies all the syntax and semantics of the word. M-scripts form a fully-lexicalised unification grammar, which can support adult language. Each word m-script can be learnt robustly from about six learning examples. The theory has been implemented as a computer model, which can bootstrap-learn a language from zero vocabulary. The Bayesian learning mechanism is (1) Capable: to learn arbitrarily complex meanings and syntactic structures; (2) Fast: learning these structures from a few examples each; (3) Robust: learning in the presence of much irrelevant noise, and (4) Self-repairing: able to acquire implicit negative evidence, using it to learn exceptions. Children learning language are clearly all of (1) - (4), whereas connectionist theories fail on (1) and (2), and symbolic theories fail on (3) and (4). The theory is in good agreement with many key facts of language acquisition, including facts which are problematic for other theories. It is compared with over 100 key cross-linguistic findings about acquisition of the lexicon, phrase structure, morphology, complementation and control, auxiliaries, verb argument structures, gaps and movement - in nearly all cases giving unforced agreement without extra assumptions.

cs.CL↗

arXiv2020

The Aggregator Model of Spatial Cognition

Tracking the positions of objects in local space is a core function of animal brains. We do not yet understand how it is done with limited neural resources. The challenges of spatial cognition are discussed under the criteria: (a) scaling of computational costs; (b) feature binding; (c) precise calculation of spatial displacements; (d) fast learning of invariant patterns; and (e) exploiting the strong Bayesian prior of object constancy. The leading current models of spatial cognition are Hierarchical Bayesian models of vision, and Deep Neural Nets. These are typically fully distributed models, which compute using direct communication links between a set of modular knowledge sources, and no other essential components. Their distributed nature leads to difficulties with the criteria (a) - (e). I discuss an alternative model of spatial cognition, which uses a single central position aggregator to store estimated locations of each object or feature, and applies constraints on locations in an iterative cycle between the aggregator and the knowledge sources. This model has advantages in addressing the criteria (a) - (e). If there is an aggregator in mammalian brains, there are reasons to believe that it is in the thalamus. I outline a possible neural realisation of the aggregator function in the thalamus.

q-bio.NC↗