Archive for February, 2020

Artificial Intelligence Transubstantiation

A future beckons where typing may be redundant. In that future, what if your brain could type? Beginning with a sufficiently intelligent algorithm (henceforth AL. Gorithm) we have argued that AL. needs a conscious and wisdom. For the purposes of this Blog essay, a conscious and wisdom are presented as images, allowing us to generalize within a mathematical morphology. A mirror metaphor is used simply to highlight the challenged faced at that moment in time when geometric neuron patterns of a conscious brain are transubstantiated onto a physical object. This Blog essay builds on earlier research[1]. In the game transition, we have chosen to label AL. Gorithm’s coherent  understanding of the word ‘mirror’ as a process of artificial intelligence transubstantiation. Work in progress. There are ppt slides and documentation: http://www.patrickmcnutt.com/news/the-tao-of-robot-ethics-artificial-moral-agents/.

 

Introduction

 

A future beckons with augmented reality, a future where smartphones are replaced with smart glasses, where typing is redundant. In that future, what if your brain could type? A not dissimilar question was posed by Facebook in 2017. By using sensor technology, for example, researchers have enabled a brain-computer speech-to-text interface. Artificial intelligence is on a long path to human intelligence. Albeit, reasoning, common sense, conceptual thinking, cross-domain thinking, creativity, self-awareness remain, for the moment, beyond the reach of algorithms. Many argue that artificial intelligence systems, however mathematically robust, are unlikely to acquire real person wisdom. It’s a fantasy, argues Koch (2020). Until a master algorithm[2] is conscious of itself, Mitchell (2020), who delivers an excellent review of the literature admits that ‘human-like’ intelligence is really far away. We explore some options. Beginning with a sufficiently intelligent algorithm (henceforth AL. Gorithm) we have argued that AL. needs a conscious and wisdom. For the purposes of this Blog essay, a conscious and wisdom are presented as images, allowing us to generalise within a mathematical morphology. We introduce a mirror metaphor to explain. Framing a robotic conscious within the rigors of a Prisoners’ dilemma requires humans to be less selfish in their behaviour that is captured and processed into data packages. AL. Gorithm ‘thinking’ introduces both a philosophical and a mathematical dimension to a debate that can be housed within the dualism of mind v brain, and thus, complementing an underlying game topology. Borrowing a phrase from Mayor (2018), the algorithms are biotechne that is, they are ‘made, not born’. Consequently, there are deep ethical issues on existence, thought, creativity, perception and reality to be unlocked before we can answer the question: what if your brain could type? But what if Facebook’s goal is realised: what if the data patterns of a conscious brain can be translated onto a physical object – isn’t your brain typing?

 

Mathematical Morphology

 

If the data patterns are geometric then the artificial machine intelligence (machine learning, deep learning) ultimately relies on the geometry of human patterns and inference. Mathematical morphology is a non-linear approach to image processing that relies on the complete lattice structure. What if the brain-computer interface had an abstract structure that can be studied as a lattice structure? Deep learning neural networks are trained on bits and bytes of labeled data packages; then they use what they’ve learned to mathematically select and identify patterns from images, sounds, or payments. As the robots adjust to changes in their programmed environment the biotechne algorithms ‘think’ in real time. Researchers[3] in the enhancement of brain MRIs have explored the brain-computer interface as a lattice structure. But unlike the human brain, AlphaGo, for example, which beat the world’s master in the ancient game of Go, does not play chess; pigeons, however, do play chess[4].

 

Mirror Metaphor

 

What if there is a game dimension underpinning the concept of cognitive connections between brain neurons and a physical object like a keyboard. You can visualize a sentence or word in your mind’s eye but your brain has to project it onto a visual. This is a challenge. Try this simple experiment: write the word ‘mirror’ on a piece of paper. Hold up the word ‘mirror’ against a mirror. Now, look at the image. It is difficult to decipher. The brain is consciously reading or ‘opening’ an image from the mirror but you cannot read the word ‘mirror’. This is analogous to the challenge faced by a thinking biotechne algorithm. The translations of the structuring of the visual image of the word ‘mirror’ create a memory bank of images.

 

The challenge for AL. Gorithm is to read or[5] ‘open’ the visual images of the word ‘mirror’ so that an image can be unscrambled by a robot and understood as the word ‘mirror’ as if the robot had a thinking, emotional brain. A thinking algorithm has an advantage provided a memory of the image is retained in a memory bank, thus providing the bits and bytes of data that intelligent algorithms require to function.  And in a process not unlike predictive text on a smartphone, AL. Gorithm, by employing morphological filters, will decipher that image of the word ‘mirror’ coherent with a real person’s vision of the word ‘mirror’ at a moment in time.

 

 

Artificial Intelligence Transubstantiation

 

The word ‘mirror’ cannot be understood without reference[6] to the morphology of the mirror image. A circular interpretation of the word or sentence is a technical translation of the substance of the brain and the conscious mind onto a physical object like a mirror, a keyboard, or a robot or a computer. Significant research on brain MRIs and the challenges addressed in mathematical morphology are noted. Early thoughts on the mathematics of the process as a game on manifolds are located in the ppt slideshows referred to above. In the game transition, we have chosen to label the player AL. Gorithm’s coherent  understanding of the word ‘mirror’ as a process of artificial intelligence transubstantiation that is, that a move by AL. Gorithm, as a player, is diffeomorphic to a move by a real person[7]. At that moment in time in the game phase when the biotechne algorithm translates the image of the word ‘mirror’ before a human a transubstantiated robotic ‘bluetooth brain’ is at a point where it is ready to type. The mirror metaphor is used simply to highlight the challenged faced by the mathematicians at that moment in time when geometric neuron patterns of a conscious brain are transubstantiated onto a physical object.

 

No Meaning No Conscious

 

Our thoughts, our memory, our emotions – in essence, our mind – and, our search history have become data packages that can be ascribed a geometry. They can be arranged into geometric ordered patterns. But data packages have no meaning for AL. No meaning, no conscious. The master biotechne algorithms – like the ones deployed by AlphaGo or by Facebook’s F8 laboratory or by Neuralink in their plans for brain-reading[8] – will require an ethical boundary. The boundary will only evolve from a coded gesture of human understanding, knowledge, reasoning and wisdom that can be wholly independently ascribed to AL, consciously aware and processed, coded and programmed.  We call this ‘the big equation’. In the Masterclass, we did explore, briefly, ‘the big equation’ as a cognitive awareness (visual) experiment by looking at the Winograd semantics and perception in terms of Kanizsa patterns and illusions. The purpose of that exercise was to identify a link between meaning and conscious and to demonstrate – with cognitive simple visual aids – an axiom[9] of no meaning no conscious.

 

Transubstantiated ‘Bluetooth Brain’

 

Facial recognition software, for example, builds on the architecture of the neural networks that specialize in processing data by translating an image into a binary set of visual data. So, is there a Gestalt switch, a diffeomorphism, wherein AL’s ethical perspective can be mapped? If so, at what node in the complex coded pattern of digital behavior does it occur. Our research agenda is to find that switch, analogous to an opening move in a game: because at that switch our memory has (already) been outsourced to AL as we rely exclusively on smart devices. And de facto at that switch our brain ≈ our mind. Our memory is entangled with smart algorithms that have intruded our everyday life. We rely on AL. Gorithm. The AGI dream of an artificial human brain that can run on cloud computing systems requires an artificial intelligence transubstantiation.

 

Zimmer’s conjecture

 

The mirror metaphor earlier, is but a first principles approach to understanding a transubstantiated ‘bluetooth brain’. A conscious brain, ordered and reflective, has a geometry. Measured and programmed data patterns are a lower dimension in the geometry. They are programmed in order to inform the feedback loop of the master algorithm. At a higher dimension, arguably, data has meaning only when AL. Gorithm thinks and is consciously aware. Then the data patterns as measured are at a higher dimension in the geometry, programmed to inform the feedback loop of a master algorithm and, coded into an abstract ‘thinking’ biotechne algorithm. At that diffeomorphic Gestalt switch within the network architecture, a transubstantiated robotic ‘bluetooth brain’ is ready to type.

 

Prognosis

 

If the research on the  brains’ neurons as an algebraic dual space can be ascribed a game space then could the brain be measured as a smooth manifold with bits and bytes of structured data? The associated algebra and the topology is already embedded in the memorized data patterns in the brain. So in order to recover the manifold structure AL. Gorithm could be programmed as a Player Unknown in the asymmetric game space. In the AGI[10] search for a single piece of software capable of learning almost any human task mathematical logic has been applied. When the rank[11] of the lattice is larger than the dimension of the manifold our brain ≈ our mind ≈ conscious AL. Gorithm. At that point of balance, artificial intelligence transubstantiation would be complete. So let’s solve the big equation.

 

Our transubstantiated ‘bluetooth brain’ imbued with a deep learning programming code[12] with ethical values, would be ready to type. If, and only if, the Winograd sentence ‘time flies like an arrow’ could be interpreted by AL. Gorithm as a statement of fact and not an expression of love and emotion would the transhumanism fears about ‘the big equation’ be assuaged. This would be a small step in the right direction. The transhumanist dream of downloading one’s mind into a computer circuity processed by biotechne algorithms would have been realized. It is happening; no doubt within Facebook’s F8 laboratory, or at Neuralink, or OpenAI or DeepMind Google Brain significant advances have been made. Was it a car or a cat I saw?

 

 

 

Bibliography

Christian, B & T. Griffiths (2017): Algorithms to Live By Picador Publishers

Koch, C (2020): The Feeling of Life Itself: Why Consciousness is Everywhere but Can’t Be Computed MIT Press

Lee, Kai-Fu (2018): AI Superpowers: China, Silicon Valley and the New World Houghton Mifflin Harcourt

Mayor, A (2018): Gods and Robots Princeton University Press

Mitchell, M (2020): Artificial Intelligence: A Guide for Thinking Humans Pelican Press

Citations

Philip Ball ‘Android Dreams’ Article in Prospect Magazine pp57-59 March 2020

Fortune Magazine Europe edition February 24 2020 on Artificial Intelligence.

[1] This short essay builds on earlier work in progress with neotenic data patterns where algorithms are emotionally connected to code. Further arguments in support of interpreting data ≈ data with emotions involved a game between onsumers (online consumers) and a personalized ‘act-like’ a human Al. Gorithm. The tao of robot ethics and related materials are available on my webpage.

 

[2] Term employed by Pedro Domingos (2015): The Master Algorithm, published by Penguin books.

[3] There are articles on brain MRI in The Journal of Ambient Intelligence & Humanized Computing [read 2019 article by Vikrant Bhateja et al ‘Human Visual System Based Mathematical Morphology’] and early research available at Semantic Scholar by Olivier Lezoray et al (2014): ‘Learning Complete Lattices for Manifold Mathematical Morphology’ and a paper by Carvalho et al (2014) ‘Lattices from Combinatorial Game Theory’ in the journal Theoretical Computer Science, vol 527 pp37-49.

[4] We refer to this as isomorphic asymmetries in a lattice game dimension. The patterns could be presented in geometry a (lattice) game on manifolds, and we refer to torus and neoteny at http://www.patrickmcnutt.com/news/meso-data-omega-circles-origami-manifolds/

[5] The morphological opening of the word ‘mirror’ could be regarded as the opening move in the game.

[6] The reasoning here is not dissimilar to Heidegger’s hermeneutic circle – the idea that neither the whole text nor any individual part can be understood without reference to one another: Heidegger’s work published in 1927 is Being and Time

[7] To be proven by the author and beginning to revisit John Nash’s mathematics on compact real manifolds.

[8] Neuralink is the company financed by Elon Musk to develop brain-machine interfaces.

[9] The axiom holds unless and until AL. Gorithm as an artificial moral agent connects ‘act-like’ human thought and emotions.

[10] AGI is artificial general intelligence with Microsoft and Google competing to develop human-like OpenAI and DeepMind Google Brain algorithms respectively with common-sense reasoning and conceptual thinking. The dream is to replicate the human brain.

[11] Zimmer’s conjecture was solved in 2017 –  symmetries can exist in a higher dimension that cannot exist in lower dimensions.

[12] The values are subject to the constraint of a non-empty well-defined core of ‘the big equation’, the metric equivalent of a gesture of human understanding, knowledge, reasoning and wisdom