By petr zhegin in neuro — Nov 1, 2020

Rethinking consciousness by Michael Graziano. A summary

‘Any sufficiently advanced technology is indistinguishable from magic’.

Arthur C. Clarke

This quote by Arthur C. Clarke in my opinion, is the best epigraph to ’Rethinking Consciousness’ by Michael Graziano.

Takeaways/interpretations

Consciousness can be engineered;
Attention schema and machinery that underlines it create simulations of our state of mind and other peoples’ states of mind/behaviour. In a sense, we live by simulating ourselves and others. I’m not sure I could square it with ideas of ‘living in a current moment’, ‘being here and now’. I feel that meditation (that I probably do wrong) is an attempt to stop these simulations and kind of opposition to how the brain works;
Attention may help to integrate various bits and pieces of information together;
Brain scanning technology is a bottleneck on the way to achieving artificial consciousness;
Artificial consciousness and mind uploading, followed by mind-to-mind communication, will be a revolution bigger than the invention of writing;
At some point, dying and melting into a hive mind may become a privilege.

Summary

Attention schema

Michael proposes to focus on exploring what allows a human, or a machine, to make a claim that it has a subjective experience. While other theories usually focus on the question of how a brain generates a feeling of consciousness.

A core element of the machinery that the brain uses to make a claim about consciousness is the attention schema - a model that allows the brain to collect and process information, monitor itself, and its internal abilities, and also to build predictions about other people. The attention schema is based on attention, the ability of the brain to focus its limited resources on something.

‘An attention schema is a bundle of information that describes attention—not the object being attended, but attention itself. It monitors the state of attention, keeps track of how it can change dynamically from state to state, and predicts how it may change in the next few moments’.

The attention schema is somewhat similar to the body schema, a very high-level schema that describes how to control limbs but that lacks any information about how joints, muscles, etc. work. Our body schema is able to quickly incorporate even non-realistic limbs (rubber hand experiment) into itself but then very slowly retire its elements that don’t fit reality (phantom limbs). A phantom limb is a kind of simulation that persists after a real limb is lost.

‘They [body and attention schemas] are both simulations… The body schema is a model of the physical self and how it works, while the attention schema is a model of another part of the self, the interacting neurones inside the skull and how they work’.

The fact that we don’t have visibility into body/attention schemas makes them seem magical.

Attention: why do we need it and its evolution

Michael provides an overview of how attention evolved, from its simplest form of boosting signals through lateral inhibition (the capacity of an excited neurone to reduce the activity of its neighbours) to centralised attention of vertebrates that allows to analyse of data from all sensors of a being.

‘…as much as half a billion years ago the earliest jawless fish developed some form of overt attention, a tectum to control it, and probably an attention schema to facilitate that control’.

The need for attention is evolutionary obvious. Predicting others’ attention is critical for predicting their behaviour. The book contains an account of how human behaviour might be predicted. That prediction takes into consideration context and affordances in the environment, historical data on the behaviour of an average person, the intentional stance of a particular person (if someone did not eat recently, then the probability of eating is higher) and, importantly, attention of this person to a particular affordance in a given moment.

It’s also made clear that prediction of others’ behaviour is critical, as a simple collection of reflexes and kind of hardcoded rules will not make it. There would be so many rules that the brain would be unable to store and retrieve them efficiently. Rules also tend to oversimplify all the complexity of life.

So the line of thought, as I understood, is the following: the need to predict the behaviour of others leads to the need to predict attention, which leads to a demand for special machinery in the brain that is capable of doing that.

Attention and consciousness

Attention is a data handling method, a process of enchanting signals of one representation and reducing signals of another representation. Consciousness is an inner experience we claim to have, and machinery that tracks attention, redirects it if required, and projects it to others.

‘Attention is something the brain does; consciousness is something the brain says it has…

… In the attention schema theory, the whole point of awareness is to give the brain a running account of attention. Awareness therefore tracks attention closely, something like the body schema tracking the location of the arm’.

‘Consciousness is an ancient, highly simplified, internal model, honed by evolution to serve two main useful functions…. Its first function may have been as a self model, to monitor, make predictions about, and help control one’s own attention. The second function may have been as a catalyst for social cognition, allowing us to model the attentional states of others’.

Consciousness also has something to do with integrating information. It might be one of the most important mechanisms to integrate various bits and pieces. For example, colour is an information connector limited to the visual domain (we could group unrelated objects based on colour). Spatial location is a more general connector that connects multiple senses, e.g. sound and movement from a bird on a tree. The power of location as a connector is due to its relational property - it’s not about an object, its colour, shape, etc. It is about how it relates to you so that it can be applied to anything.

However, some information domains have less obvious connection to location, e.g. feelings, mathematical insights, etc. Attention schema is a universal connector, even wider than location. It has information about attentional relationships with objects.

‘By treating attention as a relational property of the world that is worth modeling, the brain constructs a central connector, the attention schema, to which all other information sets in the range of your attention will necessarily attach’.

Michael explores what parts of the brain have something to do with consciousness. The temporoparietal junction (TPJ) might be an important component of consciousness machinery. The TPJ is involved in processing information about the mind states of other people. When the TPJ is damaged, like in the case of hemispatial neglect, the consciousness of a person about one side of the body is entirely eliminated. A person draws half of a house, eats from half of a plate, etc., without being aware of the other half.

Artificial consciousness, mind uploading

In the book, there’s an amazing overview of how an artificial consciousness can be created. I’d like to highlight the technological challenges of getting there:

The brain has ca. 86 billion neurons and 100 trillion synapses. MRI machines’ resolution is 1/2 of a millimetre. To detect a synapse, one needs micrometre resolution (1/1000 of a millimetre);
Then, one needs to identify a type of synapse. There are probably hundreds or thousands of types of synapses;
There are glial cells, 10 times more than neurones. These cells are not just supporting infrastructure; they have properties relevant to information processing, e.g. secrete chemicals that affect neurones.

It’s hard to predict when artificial consciousness will be possible. Michael shares an exciting benchmark of how long it takes to achieve technology that matches theory. It took almost 100 years to build a machine sensitive enough to confirm gravity waves predicted by Einstein.

Michael also studies mind uploading, scanning the brain, simulating it, embedding it into a virtual body and building virtual worlds, and highlights various challenges. Among them:

Compatibility issue - tech evolves quickly, but mind uploading is about living forever; there will be a need to make tech for the long term.

‘Our capitalist, consumer, high-turnover approach to information technology would have to change substantially before a mind-uploading platform would make any sense for extending anyone’s life’.

The coercive power of the afterlife;
Finding a business model - how to pay for resources needed to upload and store all the data?
Relationship with your own simulation.

In the book, consciousness is considered in an extensive context, and consequences of its engineering are considered as well. Michael uses analogy with writing, which was invented as a tool to keep records of transactions, but allowed to connect generations and to accumulate knowledge. Artificial consciousness and mind uploading is the key to another revolution when those who have knowledge can share it with generations directly.

‘Mind uploading would transform our species in a way that might surpass the evolution of speech and the invention of writing. The change might not always be constructive, however. Just as in the case of writing, television, or the Internet, when society takes on a new technology that increases the flow of information, that change inevitably comes with a rise in misinformation and harmful social memes’.

When mind-to-mind communication is enabled, another fundamental shift in information flows will happen. Instead of a collection of simulated individuals, uploaded minds become a central nexus of intelligence, where individuality will be lost. Dying and melting into this collective will become a privilege.

‘When I try to look as far as I can into our future, the single most important change that I can see—the watershed moment in the history of our species—is the moment when people understand consciousness’.

Photo - Andrew Ostrovsky/iStockphoto