Will machines take over??

Preface

Is it really possible to achieve "Singularity" in Artificial Intelligence? This subject that has garnered a lot of interest in the recent years due to the exponential growth in the field of artificial intelligence and machine learning. The best minds in the world are pondering and researching on this subject so this post is just my personal views and read at your own discretion!

What is singularity? The definition as coined by the experts is a term used to a point in time when technical advancements will be so rapid that it will lead to significant changes in human civilization. From discovery of Fire to invention of the Wheel, human civilization has achieved multiple singularities that revolutionized the world but if we look at the timescales, the changes that used to happen over thousands of years have started happening in the matter of decades and years. Machine learning that was limited to academic papers in the 1950's has become easily accessible to everyone, partly due to advancements in computing power but also due to the way the world is connected, the power of the internet, which was the probably one of the singularities that revolutionized the world we live in today. Collective intellect of humankind is working at a fast pace by sharing information and working together, thereby exponentially speeding up the creation of singularities.

AI Singularity has been feared as something that will eventually lead to the growth of an advanced artificial intelligence that will be able to evolve on its own. It's a concept that has been showcased in countless Hollywood movies and embedded in our collective psyche. Will someday machines evolve on their own? Do we really need to worry about an AI super intelligence taking over?

Background

Intelligence is a word and has multiple connotations, animals like chimpanzees, dogs, dolphins, elephants, squirrel, Whales etc. can be termed "intelligent" based on their specific abilities. What makes us different? What makes us human? and is it possible that some day we may be able to build an AI that thinks like a human? Before answering these questions, we have to go back eons and understand evolution. Darwinian evolution tells us "Survival of the fittest" and natural selection triggered evolution and though the initial trigger from the primordial soup to single celled organisms may have been an accident, complex animals evolved due to natural selection because only the ones who adapted, evolved and survived. One critical factor to adapt is problem solving. Every living thing solves problems, but the only goal is to keep themselves alive and to "survive". The beauty of evolution is every living thing also pass their genetic makeup, DNA to the next generation i.e., reproduce! This ensures continuity in evolution hence the survival. So, if anyone asks what the purpose of life is, then for 99.99% living creatures is to survive and pass their genetic makeup to the next generation! But if Humans left it at that, then we would have been still on the trees! The only difference between a chimpanzee and human is two chromosomes, we share 94% of the genetic makeup but that difference gives us Da Vinci, Newton and Einstein! This is because of one key behavior that only we have and that is "being inquisitive". Human intelligence comes from the fact that our genetic makeup (the software) makes us ask questions and challenging it, "critical thinking". Its way beyond the need to survive.

Googles deep mind can defeat the best alpha go human or alpha fold identify structure of over 200 million proteins or a deeplearning model can create art or music without human intervention. We have always looked at nature for inspiration for our inventions and we looked at out brain and tried to replicate it using deeplearning. The way our neurons in the brains interacts via synapse by taking information from our various sensory orgs (a.k.a environment), deep learning does the same by summing up the inputs and applying an activation function to provide outputs. However, the argument that I have is that how is this any different to the way humans train a dog to fetch ball? you train the dog and once you train, the dog will not need instructions, it will fetch the ball no matter where it is and will figure out a way to fetch the ball be it in the water or over a hill or any color of the ball. The current deep learning models and AI is becoming very good at varied type of problem solving and because of the compute power we have at our disposal, these models can do multiple trials and get to the solution quicker but that does not make it super intelligent, it makes it super-efficient.

The Scenario

Imagine a situation that we find a planet suitable for life, but it will be gargantuan task to move the entire population so most likely, we will send a robot to the planet to setup base. We can send something like the 3D printers and get the AI to create self-replicating robots by extracting the minerals from that planet. Hypothetically speaking if this works and the first AI robot creates a first gen of robots and starts working on setting up the infrastructure. Will the first robot feel a sense of "love" towards the children that it created and a need to protect them? Probably yes, because mathematically survival of these replicated robots is critical to the successful completion of the mission "objective". Humans will not need to code "love", mathematics will handle that. The next generation that was replicated probably also evolved (may be created a new arm because of the type of soil on that planet- without a pre-defined set of instructions from humans back on earth), this is completely plausible. Now let us take this one step further, imagine an asteroid hit the earth and entire humanity is vanished and the humans never arrive. What happens to these robots who are now stuck in this far away planet? They were sent with a task to setup infrastructure for humans but then the humans never arrive, will they feel their sense of purpose is complete and shut down or will they continue to look up to the stars? Will they ask, "what is our purpose ", become inquisitive and in pursuit of answers, continue to evolve and keep replicating one gen after another and eventually the original purpose is somewhere lost, and they are only trying to survive !

What will happen in this hypothetical scenario is unknown, but the answer may start with the question itself.

So where do we start... firstly does the neurons, synapses or our brain (the computing power) give us consciousness? Nope, that's just the mechanics, the engineering... what truly makes us human is our DNA and that's where the "inquisitive" human mind may find what it is looking for!! It's not the compute, it's the code!

The exploration continues!