Research in this profile theme addresses health, cognitive functioning, disease and wellbeing throughout the life cycle. By studying the complex interplay between genes and environment, and the working of the brain over the life span, we aim to provide the tools for the early diagnosis and prevention of diseases and psycho-social problems, and for lifelong cognitive, emotional and physical enhancement, resulting in medical literacy and healthy ageing across the population.
Reverse engineering the human brain
The most profound transformation will be in "strong" AI, that is, artificial intelligence at the human level.
To recreate the capabilities of the human brain, we need to meet both the hardware and software
requirements. Achieving the hardware requirement was controversial five years ago, but is now largely
a mainstream view among informed observers. Supercomputers are already at 100 trillion (1014)
calculations per second (cps), and will hit 1016 cps around the end of this decade, which is the level I
estimate is required to functionally simulate the human brain. Several supercomputers with 1015 cps are
already on the drawing board, with two Japanese efforts targeting 1016 cps around the end of the
decade. By 2020, 1016 cps will be available for around $1000. So now the controversy is focused on the
To understand the principles of human intelligence we need to reverse-engineer the human brain.
Here, progress is far greater than most people realise. The spatial and temporal resolution of brain
scanning is progressing at an exponential rate, roughly doubling each year. Scanning tools, such as a
new system from the University of Pennsylvania, can now see individual interneuronal connections,
and watch them fire in real time. Already, we have mathematical models of a couple of dozen regions
of the brain, including the cerebellum, which comprises more than half the neurons in the brain. IBM is
creating a highly detailed simulation of about 10,000 cortical neurons, including tens of millions of
connections. The first version will simulate electrical activity, and a future version will also simulate
chemical activity. By the mid 2020s, it is conservative to conclude that we will have effective models of
the whole brain.
There are a number of key ways in which the organisation of the brain differs from a conventional
computer. The brain's circuits, for example, transmit information as chemical gradients travelling at
only a few hundred metres per second, which is millions of times slower than electronic circuits. The
brain is massively parallel: there are about 100 trillion interneuronal connections all computing
simultaneously. The brain combines analogue and digital phenomena. The brain rewires itself, and it
uses emergent properties, with intelligent behaviour emerging from the brain's chaotic and complex
activity. But as we gain sufficient data to model neurons and regions of neurons in detail, we find that
we can express the coding of information in the brain and how this information is transformed in
mathematical terms. We are then able to simulate these transformations on conventional parallel
computing platforms, even though the underlying hardware architecture is quite different.
One benefit of a full understanding of the human brain will be a deep understanding of ourselves, but
the key implication is that it will expand the tool kit of techniques we can apply to create artificial
intelligence. We will then be able to create non-biological systems that match human intelligence.
These superintelligent computers will be able to do things we are not able to do, such as share