Metaphors simplify communication. They can also lead to inappropriate conclusions. Much discussion comparing brain functions and computer capabilities is misplaced.

Last week, the AI Blog at Google [1] reported on the release of the first-ever connectome — map of neural connections in the brain — of a Fruit Fly[2]. As soon as I read the abstract, I was on Twitter with a quick comment[3].

Brilliant work! Still eons before scientists build complete connectomes for a human brain. It will still not be enough to reconstruct how human behavior functions…

This connectome research is an important milepost that shows substantial progress in mapping the cellular anatomy of a brain, albeit a very tiny brain. Still, it also illustrates just how far away we are from a full, working model of the human brain and how it works.


It’s OK to use metaphors to describe how a particular piece of advanced automation works but don’t confuse metaphors with reality. A metaphor is a figure of speech in which a word or phrase is applied to an object or action to which it is not literally applicable.

Expressions like machine learning, artificial intelligence, and cognitive computing are metaphors that connote certain human capabilities. They’re not accurate, but they are sometimes useful, at least until people start believing that the metaphors are accurate models.

Key Action

Don’t let computer technology vendors throw tortured biology metaphors at executives in your enterprise. Focus on business outcomes, not unscientific metaphors.

Kurzweil’s singularity hypothesis[4] assumes that computational capacity will continue to grow exponentially to the point where it rivals the raw computing power of the human brain. Most critiques of this hypothesis have focused on the limits to the growth of computing capacity. The deeper flaw lies in Kurzweil’s assumptions on the computational power of a human brain and the ability to fully emulate that brain with a computer.  None of the estimates of the raw computing power of the brain are based on a full mapping of the neurophysiology of the human brain or the CNS (central nervous system) which includes the brain and other neural tissue. What the CNS does and how the CNS does it remain hard questions to answer.

I’ll come back to Kurzweil at the end of this article, but first, a prediction and some deeper detail


I am waiting to see other authors gleefully point out how the fruit fly research proves Kurzweil’s singularity is coming, that we will soon see machines that can fully emulate most if not all of what the human brain does. I predict we will see many such conclusions appearing in books, magazine and newspaper articles, blog posts, maybe even Instagram pictures illustrating how close we are to reinventing humans.

Nothing could be further from the truth.

Go deeper and flag shortcomings

Here’s a simplified drawing of a neuron — showing three main elements (cell body, dendrites, and axon) with some supporting features.

Diagram of a neuron

Here are eight reasons why we are eons away from being about to fully understand how the brain works.

  1. The fruit fly connectome is tiny. It mapped connections between twenty-five thousand neurons. The human brain is huge. With over eighty-six billion neurons, the average human brain has over three and a half million times as many neurons as a fruit fly brain. We have a very long way to go to construct a connectome of that scale. And who knows how many methodology revolutions will be required to scale that far! If every year we double the number of neurons whose synapses we can map in a connectome, it will take 23 years to get to the point where we may be able to create a human connectome. When we’re done, what do we have? Not much!
  2. Anatomy is not physiology. In the fruit fly connectome work, the researchers studied extremely thin (dead) slices of fruit fly brains, anatomy, not physiology. By identifying structural features of neurons, namely the synapses — places where neurons communicate with other neurons[5] — they studied structure, not function. Physiology is the study of function. Sometimes, structures imply something about functions, but that kind of inference isn’t enough to describe the brain’s neurophysiology. What a neuron does is far more sophisticated than what non-experts assume. Consider points 3 through 7.
  3. There are non-synaptic neuronal processes that occur between interlaced dendrites of multiple neurons. Those non-synaptic dendritic processes are biases on the neurons, increasing and decreasing the probability of firing.
  4. Glia: There are about 85 billion glia cells in the human brain. We’re learning that glia play a role in neuron physiology. They’re not just support and maintenance cells.
  5. Neurotransmitter soup: There is a constantly changing intercellular soup of neurotransmitters that bathes neurons. The soup influences firing and refractory intervals (post-firing recovery time.)
  6. Intracellular networks: There’s some research evidence that there’s more processing inside a single neuron — it’s more like a network inside each cell — and we need to understand the role of intracellular nets in the overall physiology of the brain.
  7. Synaptic firing isn’t necessary: Neurons that do not fire still learn. There is much more going on than electrical spikes in neurons and brain connections. Proteins, which are the little biological machines that make everything in our bodies work, combined with local electric potential, do a lot of information processing on their own — no activation of the neuron required[6].

Finally, the brain doesn’t exist or operate outside the body. Some parts of the body, like retinas, are direct “extensions” of the brain and others, such as the endocrine and gut systems, impact how the brain functions.

Singularity Conclusion

Kurzweil’s singularity hypothesis (‘when humans transcend biology’) presumes we will know enough about the functioning of the human central nervous system (CNS) to be able to emulate it with computer technology just as soon as we can assemble a system with the capacity to perform all the functions the human CNS performs. But we don’t know what functions the CNS performs or how it performs them. And we still have no proven path for building Artificial General Intelligence.

The singularity is not near.

Be wary of geeks peddling metaphors.