Functional group interconversions(oxidation reduction)
Can we fully digitize the brain and run it real time ?
1. Can we fully digitize the brain and run it real time ?
Petr Hruška
phruska@gmail.com
Final Project
Understanding the Brain: The Neurobiology of Everyday Life
2. 2
One day, several months ago an article in Czech magazine took my attention.
In that article the author cited some interesting
words of Raymond Kurzweil, an American
author, scientist, inventor, futurist, and director
of engineering at Google.
Mr.Kurzweil said in that article he is looking to
the future (~2045), where computers will have
enough capacity and power to fully simulate
the human brain so we will be able to upload
ourselves into the cybernetic body.
Mr. Kurzweil's claim about uploading minds by 2045
3. 3
And questions started attacks in my mind:
Is Mr.Kurzweil serious or does he
just provoke to attract attention ?
Could really silicon chips and electrical circuits
fully simulate all the brain functions ?
Is all what we are the matter of CNS, can we live
without our body as digital beings ?
I'm software engineer and I have some idea how computers work, but I did not know
really much about the brain. I decided I need to study to find out more. So I enrolled for
Computational Neuroscience and Understanding the Brain: The Neurobiology of
Everyday Life courses to get some background to asses Mr.Kurzweil prediction.
4. 4
So the question is, in future, can we fully simulate the human brain functions ? And
could such simulation also bring the self-awareness and intelligence of living organism ?
My project is to briefly compare the human brain and computer performance to have
better idea for the the former question. For the latter one, my current feeling is unless
we try, we're not going to know really ...
5. 5
Basic Description and Characteristics
The human brain is complex
neuronal network supported by
glial cells and powered by
oxygenated blood. It's a very
parallel system with multiple core
centers communicating each
other simultaneously.
Stimuli are sensed and
transformed to electrical signals ,
which are transmitted by
neuronal pathways to the brain.
Brain is storing and interpreting
inputs from outside and inside
the body and process it up to
decision how and if to react.
Computers are complex systems of
electrical circuits organized to
integrated processor unit(s), other
silicon chips and memories,
powered by electricity.
Computers are very fast, still mainly
sequential systems. Inputs from
peripheral devices are transformed
to binary code and translated
through several layers to limited
amount of instructions which can be
processed in CPUs. Outputs are
predictable reaction based on
programming and logic burned in
the chips.
6. 6
Compatibility
So we have biological and artificial systems for information processing . Are these two
systems compatible enough that we can thing about full simulation of the brain ?
While there are major differences in used materials, brain anatomy and computer system
architecture, one essential and very key aspect is common. Information is processed as a
binary code, where 1 is represented as higher electrical potential than 0. Both systems
transform stimuli or input information into binary code ~ electrical impulses to be
processed and reactions are transformed back to whatever output systems you have
available, muscle contractions or robotic arm movements. The fact that both systems can
be and are compatible has been proven by many ways, from mind-controlled prosthetic
limbs to such experiments where you read neuronal motorfunction signals from one
person and send the information over the computer network to another body, where the
resulting movement is performed.
7. 7
Speed and Overall Performance
Brain neurons response time is measured in milliseconds or tens of milliseconds.
You will be disqualified from 100m sprint race if your start reaction time is <100ms.
You can't simply react more quickly. Computers on other side operate on GHz
frequencies ~ a few billion cycles per second, so are about million times faster.
The speed is certainly not only performance criterion. Thanks to massive parallelism
in neuronal networks, human brain is still overperforming computers. As computer
speed has quite reached materials limits, we can see big focus on parallel
computing nowadays. New processors are multicore and multithread units, but
despite this fact machines are still quite sequential systems.
But there is one important fact. While our brains will be pretty much the same in
following few decades, computers are doubling their performance approximately
every 18 months. So we can expect that computers will match the brain
performance one day and overmatch it right after.
8. 8
Capacity
Another essential feature is ability to store and re-use information. We have many
types of memories stored mainly in neocortex and other parts of the brain. Rough
estimation of brain talks about 100 billion neurons passing signals to each other
via as many as 1,000 trillion synaptic connections.
Let's say 1% of neurons with a few trillion connections are participating in memory
functions and we would like to store such amount to computer memory, assessing
each neuron not only as a number of it's synaptic connections, but also including
its strange, type and other conditions determining if action potential can go
through, we may end up somewhere between hundreds of terabytes or tens of
petabytes.
The above estimation does not care about how our memory works, how stores
and reads the information. That's the other story of “software” area, here we can
limit to sort of hard drive of long-term memory and say that computers already
have such capacity, although it would not be a notebook, but rather a middle size
storage server.
9. 9
Software
It's nice to have powerful hardware, but it does nothing without a software. How we
learn, how we store our experiences, how we make decisions ? We know quite little
about higher mental brain functions and we're still in phase of “being impressed”,
inspecting a black box which internals are too complex to understand.
Computational neuroscience is trying to encode and decode information processing
in our neuronal system and has already achieved some amazing results. We can
estimate which particular neurons will fire on given stimuli, using statistical
observations on single neurons and probability theory, but there is certainly still a lot
of work to do to understand at least some of more advanced brain functions.
What we already have is brain computer interface (BCI), able to read and send back
neuronal signals. This is a huge opportunity for neurobiology and medicine that we
can at least bridge lost neuronal pathways and recover from some serious damages
of our CNS. And also we have the interface that brain and computers can
communicate each other and transfer the information.
10. 10
Final Words
So where we are in order to make full simulation of the human brain ? We know the
anatomy into very much details, having fascinating visualization methods like Clarity,
FMRI or ultra-microscopic data. We know how neurons and their synaptic connections
work. We should be able to build a computer with required speed and capacity. We
can also program data structures to represents neurons and their connections. We
have the interface to send neuronal activity data into the computer.
It would be tremendous effort to gather all the data from the brain, prepare the
computer system to which we can upload it. We can use computational neuroscience
and probability theory, which has some magic to replace all the biological environment
which keeps neurons alive, functional and determining if action potential pass to other
cell.
And actually there are researchers trying to do so and they are quite close. Next
generation of supercomputers is coming in a few years and will have enough power for
real time simulation. Just guessing if the first stimuli will be “Hello” and the reaction will
be “Hello World”, but expecting once tuned it will be more than that.
11. Not now, still some work to do ...
Petr Hruška
phruska@gmail.com
Final Project
Understanding the Brain: The Neurobiology of Everyday Life