The Integrated Information Theory of Consciousness

Brains, Minds and Machines Seminar Series

The Integrated Information Theory of Consciousness
Speaker: Dr. Christof Koch, Chief Scientific Officer, Allen Institute for Brain Science
Date: Tuesday, September 23, 2014
Location: Singleton Auditorium, 46-3002

Abstract: The science of consciousness has made great strides by focusing on the behavioral and neuronal correlates of experience. However, such correlates are not enough if we are to understand even basic facts, for example, why the cerebral cortex gives rise to consciousness but the cerebellum does not, though it has even more neurons and appears to be just as complicated. Moreover, correlates are of little help in many instances where we would like to know if consciousness is present: patients with a few remaining islands of functioning cortex, pre-term infants, non-mammalian species, and machines that are rapidly outperforming people at driving, recognizing faces and objects, and answering difficult questions. To address these issues, we need a theory of consciousness – one that says what experience is and what type of physical systems can have it. Giulio Tononi’s Integrated Information Theory (IIT) does so by starting from conscious experience itself via five phenomenological axioms of existence, composition, information, integration, and exclusion. From these it derives five postulates about the properties required of physical mechanisms to support consciousness. The theory provides a principled account of both the quantity and the quality of an individual experience, and a calculus to evaluate whether or not a particular system of mechanisms is conscious and of what. Moreover, IIT can explain a range of clinical and laboratory findings, makes a number of testable predictions, and extrapolates to a number of unusual conditions. In sharp contrast with widespread functionalist beliefs, IIT implies that digital computers, even if their behavior were to be functionally equivalent to ours, and even if they were to run faithful simulations of the human brain, would experience next to nothing.

[Stadtpark]

38:47 the part about the Integrated Information Theory starts here

[deeliciousplum]

Exceptionally stimulating talk. I am not certain as to why there is a measurable level of drama around the the idea of there being a hard problem surrounding the phenomenon of consciousness. Regardless of there being hard or soft problems, the ideas and research which help our species to better understand the brain and its emergent properties are a joy to explore. Thank you for sharing this wonderful talk.

[elmaiz]

I am going to comment about consciousness here as if I understand what it is.

[modvs]

I see the neurological sciences now have their own version of string theory.

[DaroG]

Chirstof is wrong if suggest that enough 'good' simulation of the brain will not produce consciousness. The argue with tree pin transistor vs much more pinouts of the neuron  is not proper. Neuron should not be compared with transistor but with some circuit which behave/signal process like neuron. Please note that neuron is assembled by many smaller units which have smaller connection and simpler causal power – sometimes also similar to transistor. Neither neuron nor  transistors are not  consciousness unit – only  working ‘special’ network circuit  with similar causal power effect as brain (also on the inside network level not only at the external human behavior). That for example electronic 'special' network circuit will emulate causal power of the  natural network in the brain – and that should also be conscious.  Moreover simulation of  this circuit by digital approximation on the  some framework  on the powerful  set of computers  with arbitrary precision will produce the same casual power due to the same signal processing of the 'essential'  analysis network level. I agree that eg fire simulation  is not the same as real fire but in the context of the brain is huge difference because the essence is the special signal processing which keep the same internal and external casual effect. Moreover if we add in fire simulation (eg based on the chain of the chemical reactions) the some real effectors controlled by output from this simulation then we have the same final effect: real burn ~= external human behavior). As we known any signal processing can be performed by many kind of media: eg: electronic hardware, software and  in theory by any other physical medium of the same signal computation (functionalism and multiple realisability).

[zoltanpetrik]

Utterly fascinating. I intuitively came to agree that we really do need to build neural networks in the physical world in order for it to gain consciousness, instead of simply being able to simulate it.

Although this might be true, but then why did he demonstrate the evolutionary benefit of consciousness in the real world with the simulation of a path finding robot with a neural network with increasing "virtual Φ"? If consciousness is only measured by the Φ value, then why isn't "virtual Φ" equivalent to the "real world Φ"?

[nanotech_republika]

@1:15:00, Koch says that you can not get consciousness in a computer simulation of the brain, but conciseness only appears in simulation (that brain runs) per Josha Bach. Therefore, Koch might be wrong. A computer simulation might be a perfect environment for creating another simulation within it. And therefore, to be conscious.

[slomnim]

Stuart Hammeroff's quantum microtubule research suggests that those microtubules are the substrate. Much smaller than neurons

[spiralsun]

Love this- the way it is set up is so clear and so much better!! Can see him and the slides. Excellent. He covers a lot of great studies and arguments. My book 20 years ago points out the paradox of using an evolved brain to look at brain tissue. Because we project our ideas of space time onto it etc. yet it is free to operate unconstrained beyond our coherent perceptions adapted to the eternal world. Keeping this in mind helps greatly in knowing what the nature of consciousness is. I posited the glial cells not operating according to time. But it would have to be something like that- the claustrum maybe yes. I would ask if the claustrum connects to the glial cells as well. Anyone it may end up just being something we cannot see- like the absence and ultimate connectivity that is beyond space time.

[arctos]

what i dont understand is how does consciousness not require language? how are we able to experience anything (consciously) without the ability to describe those experiences, through the medium of language?

[charlesqwu]

9:51 C does not require LTM (long-term memory)

[RynaxAlien]

Would it be correct to think that synesthesia, associations cause consciousness? Analogy could 3D illusion in 2D plane with details giving depth

[riptideelectricalservicesl]

It may just be my conscious mind, but it seems the sound quality in this video is awful.

[LogicalBelief]

The brain computes something like a wave function, similar to a Restricted Boltzman Machine. The claustrum could be unifying these and selecting the most probable observation. Wave function collapse leads to consciousness. In this particular case, complex consciousness due to the sheer complexity of the wave function being collapsed.

[_photography_]

I don't understand how we fix the points or mechanisms in the causal state space. There's got to be some equivalence classing going on somewhere that makes functionalism work.

I mean I'm probably misunderstanding, but why are the transistors the points in the computer simulation he discusses towards the end? Why not electrons? Or more abstract data structures? Don't we also have to look at all possible assignments of physical systems to networks that we then compute phi over?

I don't know, lots of good ideas but I feel like this isn't the whole picture. If I'm missing something, someone please correct me.

[vinm]

I thgought Sott Aaronson demolished this ?

[jimillsung]

The yellow squares didn't disappear for me, does it need to be seen on a bigger screen?

[Zzzooooppp]

I wonder if the simulation of a brain at the end, which Christof says behaves and has outputs identical to a conscious brain expresses itself as if it were conscious.

It would be a strange state of affairs for something to have intimate and self-discovered knowledge of the nature of consciousness (or questions thereof) but not be conscious. Any ideas?

[tiborkoos]

Is it convincing that the cerebellum is not involved in consciousness because of the graph properties of its network connectivity ? Isn't it obvious that the reason is that it is not involved in the mental processes that constitute conscious experience ? By his logic the retina is definitely involved in visual consciousness yet its network organization is just as different from cortex than the cerebellum.

[viniislaif]

I don‘t get the irreducible aspect. Anyone care to explain?