Saturday, 8 March 2014

122. Hawkins’ Model for Intelligence and Consciousness

Jeff Hawkins, in his 2004 book On Intelligence, proposed the so-called memory and prediction theory of how human intelligence arises. The basic idea of Hawkins’ theory of intelligence, in his own words, is as follows:

The brain uses vast amounts of memory to create a model of the world. Everything we know and have learnt is stored in this model. The brain uses this memory-based model to make continuous predictions of future events. It is the ability to make predictions about the future that is the crux of intelligence.

Kurzweil's recent (2012) theory (or model) of the how our brain functions has many ideas in common with those in Hawkins' book. Moreover, there is also some discussion over credit sharing. I quote from Kurzweil (2012): 'The pattern recognition theory of mind that I present here is based on recognition of patterns by pattern recognition modules in the neocortex. These patterns (and the modules) are organized in hierarchies. I discuss below the intellectual roots of this idea, including my own work with hierarchical pattern recognition in the 1980s and 1990s and Jeff Hawkins (born in 1957) and Dileep George's (born in 1977) model of the neocortex in the early 2000s'.

Hawkins pointed out in his book that the neocortical memory differs from that of a conventional computer in four ways:

1. The cortex stores sequences of patterns. For example, our memory of the alphabet is a sequence of patterns. It is not something stored or recalled in an instant, or all together. That is why we have difficulty saying it backwards. Similarly our memory of songs is an example of temporal sequences in memory.

2. The cortex recalls patterns auto-associatively. The patterns are associated with themselves. One can recall complete patterns when given only partial or distorted inputs. During each waking moment, each functional region is essentially waiting for familiar patterns or pattern-fragments to come in. Inputs to the brain link to themselves auto-associatively, filling in the present, and auto-associatively linking to what normally flows next. We call this chain of memories, thought.

3. The cortex stores patterns in an invariant form. Our brain does not remember exactly what it sees, hears, or feels; the brain remembers the important relationships in the world, independent of details.

4. The cortex stores patterns in a hierarchy.

Storing sequences, auto-associative recall, and invariant representation are the necessary ingredients for predicting the future, based on memories of the past. How this happens is the subject matter of Hawkins’ book. According to him, making such predictions is the essence of intelligence.

Hawkins takes the view that perhaps consciousness is simply what it feels like to have a neocortex. He suggests that the self-awareness aspect of consciousness is synonymous with the formation of declarative memories. These are memories we can recall and talk about.

Hawkins, while formulating his theory of intelligence, took very seriously the so-called Mountcastle’s hypothesis. Since the same types of layers, cell types and connections exist in the entire cortex, Mountcastle (1978) had put forward the following hypothesis:

There is a common function, a common algorithm, that is performed by all the cortical regions.

What makes the various functional areas different is the way they are connected. He went further to suggest that the reason why the different functional regions look different when imaged is because of these different connections only. Hawkins suggests that, although hearing, touch, vision etc. are processed by the same algorithm in the neocortex, they are handled differently in the R-brain: ‘Hearing relies on a set of audition-specific subcortical structures that process auditory patterns before they reach the cortex. Somatosensory patterns also travel through a set of subcortical areas that are unique to somatic senses. Perhaps qualia, like emotions, are not mediated purely by the neocortex. If they are somehow bound up with subcortical parts of the brain that have unique wiring, perhaps tied to emotion centres, this might explain why we perceive them differently, even if it doesn’t explain why there is any sort of qualia sensation in the first place’.

The structure of the inputs (i.e. the spatio-temporal information pattern) is qualitatively different for, say, the auditory nerve and the optic nerve. The optic nerve has a million fibres, and the auditory nerve has only thirty thousand. The optic nerve caries information that is more spatial than temporal, and the auditory nerve carries information that is more temporal than spatial. This may have a bearing on why is red red and green green. No matter how consciousness is defined, memory and prediction play crucial roles in creating it.

Here is how Hawkins answers why our thoughts appear to be independent of our bodies: ‘To the cortex our bodies are just part of the external world. Remember, the brain is in a quiet and dark box. It knows about the world only via the patterns on the sensory nerve fibers. From the brain’s perspective as a pattern device, it doesn’t know about your body any differently than it knows about the rest of the world. There isn’t a special distinction between where the body ends and the rest of the world begins. But the cortex has no ability to model the brain itself because there are no senses in the brain. Thus we can see why our thoughts appear independent of our bodies, . .’.

I end this post by stating two points made by Kurzweil (2012) regarding the brain model described above.

The first is that Dileep George also contributed to this model: 'In 2003 and 2004, PalmPilot inventor Jeff Hawkins and Dileep George developed a hierarchical cortical model called hierarchical temporal memory'.

The second point is that Hawkins' model differs in some important aspects from the model presented by Kurzweil in his 2012 book: ' . . . As the name implies, Hawkins is emphasizing the temporal (time-based) nature of the constituent lists. In other words, the direction of the lists is always forward in time. His explanation for how the features in a two-dimensional pattern such as the printed letter "A" have a direction in time is predicated on eye movement. He explains that we visualize images using saccades, which are very rapid movements of the eye of which we are unaware. The information reaching the neocortex is therefore not a two-dimensional set of features but rather a time-ordered list. While it is true that our eyes do make very rapid movements, the sequence in which they view the features of a pattern such as the letter "A" does not always occur in a consistent temporal order. (For example, eye saccades will not always register the top vertex in "A" before its bottom concavity.) Moreover, we can recognize a visual pattern that is presented for only a few tens of milliseconds, which is too short a period of time for eye saccades to scan it. It is true that pattern recognizers in the neocortex store a pattern as a list and that the list is indeed ordered, but it may also represent a spatial or higher-level conceptual ordering . . .'.

Dileep George's 2008 doctoral dissertation presents a more up-to-date description of the hierarchical temporal memory method [Dileep George, 'How the Brain Might Work: A Hierarchical and Temporal Model for Learning and Recognition' (PhD dissertation, Stanford University, June 2008)].

No comments:

Post a Comment