A body capable of providing the brain with sugar and oxygenated blood supporting the brain during its growth.

One key point is that the brain self-assembles. We do not yet understand this process and there may be some unique property of the self-assembly process on which the brain relies. This means that new brains may be impossible to construct without being self-assembled.

An example of this kind of property is the fact that each human brain is co-adapted to its body. The brain interfaces with the spinal cord and cranial nerves; the nerves interface with the body's musculature and sensory receptors. Each of these interfaces is laid out uniquely in each individual, because it grows along with the individual. A brain transplant could not be done without rewiring the corresponding brain/nerve junctions (and for several other reasons, from immune rejection to psychological trauma). Even if the neurons were successfully connected up, the cerebellum would have to re-learn how to operate the new, differently-sized muscles and the newly-shaped skeleton. You would not be able to run any time soon after the transplant.

One possibility is that the brain may employ systems which cannot be copied. An example of this kind of system, which we do not suspect is vital for cognition, is the quantum anteripositorysuperposition. Quantum effects are vital for biological systems, since they are involved in the most basic chemical reactions and are key to the operation of enzymes. However, we hope that quantum effects are simply involved at the low level - in the plumbing and the wiring - and that, therefore, we can easily swap out the molecular plumbing and wiring for electronics, and still run the high-level operations of cognition - the business logic - in the same way.

Whether the high levels are independent of the low levels is an open question. A controversial theory due to Roger Penrose proposes that quantum effects are directly involved in memory due to microtubules. A more recent and less controversial research programme suggests that RNA molecules are directly involved in the storage and retrieval of memories. This is easier to accept, since we know that the immune system completes complex pattern matching and long-term memory tasks using molecules along - receptors on the surface of immune cells which bind to each other using quantum effects.

This means that, when our immune system repels a measles infection which it has not seen since childhood, it is looking things up in a distributed molecular dictionary. When we remember something - either in the short term or the long term - are we storing information in a distributed molecular dictionary?

We hoped that memories are simply stored in the connections between neurons, in the synaptic weights, in the connectome - in simple categorical information (which neurons are connected to each other) and simple scalar information (how strong each connection weight is). This would allow us to scan these connections and simulate them electronically. However, if the high levels are not independent of the lower levels, retrieving a memory involves, recruits and depends upon the plumbing and the wiring.

This raises some interesting consequences. We have come to terms with the idea that, when we are trying to remember a particular detail, our neurons are working hard to search our memory for it.

But this conception is being slowly put to bed, if only by the demonstration of dendritic spiking, which suggests that the internal structure of a neuron might also need to be faithfully replicated.