Simulation and Neural Networks

I was looking up “Science Fiction” in the iOS Podcasts App when I came across this interesting podcast called The Sci Phi Show. The very first episode that caught my attention was an interview with Dr. David Kyle Johnson on The Simulation Hypothesis (which I’ve referred to earlier in this blog as The Simulation Argument).

https://player.fm/series/the-sci-phi-show-exploring-science-fiction-and-philosophy/the-simulation-hypothesis-with-david-kyle-johnson-sps501

In the course of this podcast, Dr. Johnson summarized the Simulation Argument by saying that if you believe our civilization will one day be capable of making a computer simulation of the universe, then by inference, there must one day exist many computer simulations of universes (because such simulations would be so darn useful), and the likelihood then that we are currently living in such a computer simulation is many (maybe billions) to one (the true physical universe), or in other words, very likely. It’s an interesting mind flip, of the sort that I imagine philosophers find really appealing. But on the surface at least, I can’t argue with its logic.

However, when questioned about whether we will ever achieve such a simulation, Dr. Johnson scoffed and admitted that we are nowhere close to say, simulating the function of a human brain, let alone building an entire universe for such a brain, or brains, to perceive. A simulation like the game, The Sims, is a far cry from The Matrix in which the subjects are conscious. But wait a second, I thought. Might it not be true that if a civilization, such as ours, becomes capable of making ANY simulation, that a universe replicating simulation MUST one day arise from such a society? And if that were true, that proves the Simulation Argument holds for our existence.

By definition: “Simulation is the imitation of the operation of a real-world process or system over time. The act of simulating something first requires that a model be developed; this model represents the key characteristics or behaviors/functions of the selected physical or abstract system or process. The model represents the system itself, whereas the simulation represents the operation of the system over time.”

And of course, we have those already. What’s really accelerated our ability to run civilizations is the computer. We can model behaviors with programing language, imbue objects with state and set them into “time”, model randomness in such a system, and see what happens.

Further in the podcast, Dr. Johnson imagines that if we were all in a simulation, then we would all be on some hard drive somewhere, existing. But the computer engineer in me begged to quibble: actually, we’d exist not on a hard drive, but in RAM – active storage that is available to be acted on by events and perceive them at the same time. In other words, conscious.

The thing is, our conventional computers of today will probably never be able to model the human brain. Recent reading I’ve been doing on neural networks suggests that humans are not, in fact, computational beings. Digital computers are exact. Human brains are by necessity inexact. Our brains observe a vast amount of sensory data, analyse this data by associating billions of neurons and synapses with each other in many complex layers of memory, and act (free will) based on this awareness of data. The analysis portion of this chain can only be done in real time with massive parallelism and necessarily inexact correlation and ranking of many layers of previous experience. It’s what we do in an instant every time we understand a sentence, or identify a cat in a picture.

Such a brain could never be simulated by a rudimentary hard drive only capable of storing and fetching data sequentially. Furthermore, the vary notion of a data moving from a hard drive, into RAM, and then waiting to be processed mathematically in a CPU, cannot work to make a neural network. The data and the use of the data would have to exist simultaneously in the same structure. This would probably have to go beyond the general purpose Neural computing instances currently being built and used by Google, although those are worth further investigation to understand how they work. You’d probably have to go to quantum computing for the kind of fuzzy, simultaneously indeterminate math that is needed. Or resort to a physical construction of our biological, cellular brains.

Which is conceivable.

 

 

Advertisements

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s