Mary and the Simulation hypothesis
In the last few decades one of the most interesting scientific and philosophical concepts being discussed by the likes of Stephen Hawking, Jaron Lanier and Elon Musk is the simulation hypothesis. This is the idea that one’s experience of life only exists as a very complex set of data in a larger, more complex simulation that resembles the universe we perceive.
Nick Bostrom proposed in an article in 2003 entitled ‘are you living in a simulation?’ that our reality is an ‘ancestor simulation’ created by a post-human civilisation who, with vast computing power might choose to simulate the lives of their ancestors.
A later paper called “Constraints on the universe as a numerical simulation” concludes that Bostrom’s suggestion is indeed possible as the universe is proven to have a finite resolution or ‘underlying lattice.’ This led digital physicists to hypothesise that the universe, as we perceive it, is not made up of matter and energy but rather of ‘quantised bits of information’ which are perceived as energy and matter.
This argument relies on a few things, such as the willingness and capability of future (or indeed past) civilizations to conduct such an experiment. The more interesting question however, surrounds the concept of consciousness and whether an immersive and convincing human experience could ever be created artificially.
An increasingly popular belief held by dogmatic scientists is that the human experience is nothing more than a complex product of biological machine learning. Essentially, this means that the all human desires, thoughts, hopes and dreams do not come from an immeasurable and intangible soul, but rather from complex, yet predictable (and perhaps more importantly, computable) biochemical reactions in the brain. In the words of Yuval Noah Harari, “There is no true self, there is just a complicated network of biochemical connections, without a core, there is no authentic voice that lives inside you.” Put simply, it is increasingly evident that humans do not have souls which in turn suggests that it might be easier than we think to simulate our experiences.
From a technological perspective this has significant implications. It is starting to become apparent that the exponentially rapid advancement of computer processing power (as defined by Moore’s law) coupled with artificial machine learning programmes means that artificial intelligence that can compete with a human is right around the corner. This artificial intelligence is governed by a set of incredibly complex algorithms which will provide responses to stimuli based on an enormous amount of collected data. Most importantly though, these algorithms can adapt and learn independently of human supervision. This happens through a process called ‘algorithmic machine learning’ and is wonderfully demonstrated by the programme “Mar/io” (which can be found on Youtube) whereby a computer taught itself to play the popular Nintendo game ‘Mario World.’
The method of all machine learning algorithms is one of trial and error repeated over many simulations incorporating slight variation. This process very closely matches the way that humans evolved biologically. That is to say, random mutations being added into the ‘system’ such that if the mutation proves beneficial, the individuals with the given mutation will be more successful and the previous ‘generation’ will be rendered obsolete. Does this suggest that given enough time, enough data and enough processing power that an accurate human experience as we know it could be replicated wholly artificially and inorganically?
The counter argument is that there is more to the human experience than the algorithmic nature of our neural networking. Each human has unique experiences which give rise to a unique emotional profile and personality. The way that a human experiences an event is special because it is fundamentally subjective. This differs from an algorithmic experience of an event which is objective and measured against a fixed criteria. This point is illustrated beautifully by Frank Jackson’s famous thought experiment of ‘Mary and the black and white room’ which goes like this:
“Mary is a brilliant scientist who is, for whatever reason, forced to investigate the world from a black and white room via a black and white television monitor. She specializes in the neurophysiology of vision and Learns everything there is to know about colour. She discovers, for example, the wavelengths of every colour in the visible light spectrum, their emotional associations and even that particular objects are certain colours: tomatoes are red, grass is green etc. She learns that the sea is blue and associated with tranquility, she understands that red is visceral and associated with anger and passion. But Mary has never actually seen these colours.”
What will happen when Mary is released from her black and white room or is given a colour television monitor? Will she learn anything or not?
This experiment tries to establish if there are non-physical properties and attainable knowledge that can be discovered only through conscious experience. It attempts to refute the theory that all knowledge is objective, physical knowledge. In this experiment, when Mary is locked in the room, she represents an algorithmic experience of the world, one that is based on knowledge and hard data rather than subjective experiences. Once Mary leaves the room (assuming that she DOES learn something new), she represents a human mind that is defined by its own conscious experience of events, a mind that can attain knowledge that cannot be physical or computable.
So, while the simulation hypothesis is a fun conversation, especially given the rate of technological advancement, it relies too heavily on speculation and philosophy to give rise to any serious scientific discussions. As Sir Isaac Newton said, “Anything that cannot be settled by experiment is not worthy of debate”.
As a fun final exercise, consider a hypothetical, vastly powerful computer was tasked with generating your life as you’ve lived it so far, as accurately as possible. Everything from your emotions and feelings, to your experiences and interactions with the world around you. Consider that your current digital footprint is used to help produce this fake reconstruction of your life.
How well do you think it would do?
How accurately do you think it could guess your emotions and the voice inside your head based on every message, tweet, email or status you’ve ever sent or received?
Could this algorithm build authentic, convincing memories from your phone’s pictures, as well as those from snapchat, Instagram and Facebook?
Would it know your music taste and sense of humour based on which Spotify playlists you listen to and which Youtube & Netflix videos you watch?
With how much precision could this algorithm pinpoint your lifestyle habits by observing your online bank statement and Amazon purchase history?
Could it even expose your deepest secrets, insecurities and wanton desires by looking at your private browsing history?
Your smartphone already knows your location to within a few squared metres at any given time and graphics engines can already achieve a resolution and frame rate better than real life. This, coupled with up to date satellite imagery suggests that the audio-visual aspects of this simulation might be strikingly accurate. But, because of the complexities of what we perceive to be consciousness and a subjective understanding of the world around us, it is too difficult to say if this simulation could ever achieve a level of consistency and immersion that would lead the simulation to be completely convincing.