AGI is not an engineering challenge- it’s an emotional and moral one.
I think, therefore I am. This phrase, as we have heard time and again, is the greatest reminder of our consciousness. When the famous philosopher René Descartes said these words, it shook the world to quite an extent. It was terrifying to think how a simple sentence could define our existence with such a degree of rationality. Human beings, in their effort to give meaning to life, aim to accomplish one thing: map their reality with their own individualistic conscience. What separates our conscience from the conscience of the remainder of the creatures in the animal kingdom is this: we work in our self-interest. But every animal and organism does. What’s different? Our drive in our self-interest can result in our own self-destruction too. The glimpses of irrationality in our actions, time and again, is what makes us truly unique.
Now you might be wondering how this is relevant to AGI. Well, I think we will never be able to give AGI emotions that are needed for a machine to work and operate at the same intellect of an average human. Human beings run on emotions and preconceived identities about themselves, not on logical or factual reality. For a true AGI to be made and brought into this world, the makers will have to do something that most of us are capable of but almost never look towards it: cruelty and evil.
Since I said what I said, I know what you might be thinking. That is a crazy notion. But allow me to ground this in logic. Most of the systems like politics, business, relationships, and even academics have some degree of cruelty involved. What kind of cruelty and to what extent? That is varied by situations and emotions. When a business does layoffs, it is one of the most cruel moments in the life of a former employee. But for the business? A necessary evil. A politician using underhanded strategies to malign their competition, swaying public opinion using agendas that speak to emotional undercurrents in the general public? Evil. These situations are more easily observable, hence the examples. But if you look around and see, this happens almost every day and everywhere.
A Necessary Evil
So how do you come up with an AGI that takes evil and cruelty into account, derived from a self-serving interest? To feel emotions, the AGI would also have to feel threatened for its own identity.
Can something with no “self” ever act as if it has one? It would need to feel small when faced with the truth. It would have to fight back by omitting information, using manipulation, and underhandedness when dealing with situations requiring underhandedness. It would have to be ignorant too from time to time to protect its ego, if we can even make it have any ego. All of these weights, as we may call them, would have to be fine-tuned as well to match the average human being's biases and cruelty. When dealing with other AGIs and humans, it would have to put on a mask of niceness even though it might not be willing to play nice inherently. How do you map this into the machine?
The tech giants promising AGI themselves are the perfect examples of the cruelty and evil their own tech would need. It is a dream being sold for their self-serving interest. What purpose it might serve? Valuation, funding, prestige, nobility, etc. You may take your pick. But for an AI to be AGI, it would be hoping a sheep lives and hunts like a wolf.
Now that we have established what it would take for AGI to be invented, let's humour ourselves a bit and assume an AGI does happen with all the necessary human-like cruelty and evil factored in. Even then, we would run into the artificial emotion paradox. Would the AGI follow the emotions of its user or its own? If it's programmed to follow its user, then it does not really have a self and in turn is not true AGI.
The metaphysical problem of AGI
To deal with our metaphysical existence, we have come up with terms in the past: soul, spirituality, etc. Different schools of thought explain these terms differently, but the interpretations somewhat aim to explain being human. This metaphysical is also a result of evolution over billions of years. Our journey actually started from microbes that were a result of chemistry and conditions on Earth that were ripe for life. Primates are our relative ancestors and did have some degree of conscience. But these microbes were just RNA roaming around in acidic oceans with no thoughts or ambitions. The DNA/RNA evolved over millions of years to go from microbes → single-celled organisms with nucleus → crustaceans → reptile-like creatures → and so on. Our present-day intellect and emotions are a result of random A/B testing by nature. We split, and split time and again in a never-ending cycle to one day invent computers and in turn claim to march towards inventing AGI. The A/B testing for all organisms on our planet has been a result of giving us correct traits to help the species survive. Spiders evolved to create silk for catching their prey, protecting their offspring, etc., and venom was a result of stomach enzymes being transferred near external glands in order to externally digest their prey, which it turns out nature found out was more convenient. These mutations in all creatures have been possible due to the DNA/RNA’s will to survive at all cost. This is my own theory, but true will is not ours. It is the DNA that carries the will in all of us to carry forward our genes and keep it surviving. Can we give the same ‘will’, ergo the DNA, to AGI? What would its DNA be?
I think it is impossible to invent AGI for the simple reason that we do not understand evolution and will itself. The ‘will’ to survive and iterate traits to maximize chances of survival is purely metaphysical. Even if we attempt to create AGI, why would we start with something already intelligent? That would not allow it to evolve into something with a true self. The AGI would have to start its journey with a basic microbe, as was our case. Else it would be a hollow image of us and, as we have seen with GPTs, it already is.
Humans are surprisingly slow at being genius
Another challenge for true AGI stems from the fact that we simply do not have quality data for AGI to learn and iterate from to reach consciousness. GPTs in their current state are already trained on existing bytes of data we have created in the last odd 30 years or so. New information and genius of some humans take time and it is incremental. Let us understand this from the examples of books and movies. The most common forms of mediums that we revel in across cultures. Movies, good ones at least with cult following and true genius, are not a common phenomenon. A lot has to happen right and under the right conditions to make a truly phenomenal movie. It has to take birth from a place of originality. The writer of the story needs to find a director. In turn, a director needs to find a producer to align their vision. You get the idea. Similarly for a book, a genius writer needs to push through all the hurdles and challenges to find a publisher willing to take a chance on their work. AGI would need to come up with its output from a place of originality and originality, in its genius form, is radical at most times. Else the AGI is only going to be a regurgitator of pre-existing ideas and works. Just like the GPTs. We will explore how genius comes to be in later articles, but for now, we need to look at business use cases where the AGI would be needed the most.
A startup takes birth. At its nascent stages, it pivots, transforms, and finds its go-to-market fit. Now the question that begs to be answered is this: at what stage of an organization's lifecycle do you plug in the AGI? If a true AGI does happen, and it works in its own self-interest, then we cannot predict its decisions. For the lack of a better word, would it be in the interest of the firm itself? If every AGI would have the same conscience using the same weights and biases, then every business would be successful. But do we have enough cake for all? For example, let's say I run Apple Inc. Now, a true AGI comes alive, and another firm uses it to compete with me. Being an AGI, it would have all the data in real time of market economics, tech innovations, and everything in between. Would it help my competitor beat me at my own game? Or my AGI would beat their AGI? We could argue, true AGIs would be specialized in certain domains. But if everyone would have the same intelligence plugged in, who are going to be the winners?
The Social Contract for AGI
Any human needs incentives to behave in a particular way. An AGI would need some sorts of rewards too. What kind of rewards would make it truly self-aware and intelligent? Would it need to eat food? Travel perhaps? Or maybe the prospect of having a family someday would align its output to match the needs of its user? An AGI, like the rest of us, would need a social contract- its own reality. However, that cannot be engineered, as we previously concluded, due to the paradox of emotions. The way out, it seems for most businesses from the pipe dreams of AGI, is a good old mix of humans, APIs, and small fine-tuned GPTs that are not overfitted like ChatGPT. More on this later, because my incentive to think further on this has ended for now. Though, if you would like a consultation on AI and automations and how you can stay ahead without all the hype and noise, feel free to book a call with me.