Reward Systems

Every day, whether consciously or unconsciously, we make countless decisions. These decisions are often based on weighing the potential rewards against the costs. In short, we evaluate the balance, and if the outcome seems beneficial, we proceed and remember to make the same choice in the future. This system has become so ingrained in our habits that it permeates every breath we take. In essence, our lives are the result of a continuous evaluation process—deciding between avoiding pain and seeking pleasure.

While may not always frame it in such stark terms, given the complexities of our daily lives and the nuances we factor in, but at the core, it’s about making a choice, often by weighing the benefits against the drawbacks. Of course, there are circumstances where we are compelled to act or where no choice is available. But generally, when we can influence our lives, we follow a simple formula based on the two variables mentioned: pain and pleasure.

Interestingly, when we look at machine learning, we notice a similar pattern. By teaching a system what it should pursue—what the favorable outcomes are—and what to avoid, we essentially grant it the creative freedom to reach that goal in the most efficient way possible. It mirrors how we are wired: although we may not think of ourselves as behaving like machines, we function in much the same way under this principle.

As humans, We claim that consciousness, morals, or other uniquely human traits set us apart, but these have little to do with the mechanics of decision-making. In the end, decisions are based on which outcome promises the most benefit. In a way, this binary behavior mirrors the fundamental workings of computers, which operate on a basis of zeros and ones. No matter how we reach an outcome, the result can often be reduced to this basic notation: we either act, or we don’t.

We employ pattern recognition, rely on experiential memory, draw comparisons, reference past situations, and weigh different options. And if we still can’t figure it out, we fall back on intuition—what we ‘feel’ is right. But this isn’t some magical solution, it’s simply another way of tapping into information that we might not have fully processed yet. We often attribute breakthroughs to religious experiences, artistic inspiration, or moments of epiphany, but it’s usually just a different path to accessing our stored knowledge.

We might question the extent to which we are, in fact, operating under conditions similar to those of computers. After all, large language models like ChatGPT have surprised us after demonstrating the ability to mirror our human methods of constructing information. As of this writing, OpenAI has even introduced enhanced 'reasoning' capabilities, enabling these models to assess data, correct themselves, and justify their responses. To me, this certainly feels like a step toward creating human-like intelligence.

My point is that we may have been anthropomorphizing our digital systems, viewing them as lacking the qualities that make us human, when in fact, we may be more alike than we care to admit. Perhaps we’re too arrogant or, worse, too naïve to acknowledge this. People often argue that computers make decisions in a cold, inhumane way. While there may be instances where they fail to account for more sensitive data, does this mean they are incapable of doing so? Or have we simply "programmed" them poorly, giving them the wrong objectives?

Furthermore, could it be that as humans, with all our emotions, biases, opinions, hopes, and fears, can sometimes be even more unreliable or erratic than a machine? It’s not unthinkable that, at some point in the not-so-distant future, we may witness a scenario where a robot or any digital system makes a decision that is more compassionate or thoughtful than a human’s. When that day comes, we might find ourselves surprised—perhaps even unsettled—by the realization that these systems have surpassed us in the very area we believed was our exclusive domain: being human.

If that were to happen, it could very well trigger a global identity crisis. We might begin to question what role we serve in the world and to what extent we make a difference. We could attempt to find 'flaws' in these silicon systems, but we may ultimately come to the uncomfortable conclusion that robots are, in some ways, more humane than we are. And at that point, we might realize that our goal shouldn’t have been to recreate human consciousness in machines. Perhaps we’ve been chasing the wrong target all along.

Perhaps instead of trying to replicate humans, we should facilitate systems to act in our best interests—systems that create a better environment for everyone and improve our quality of life. This would allow us to shift our focus towards finding new purposes: artistic expression, writing, or simply doing things we love for the sheer joy of it. People could spend more time with their families, pets, and nature. They could pick up hobbies, engage in sports, and celebrate life.

Although these digital systems may not have a sense of self or emotional depth, we might learn that effective decision-making often requires a steady, less emotional hand. Ironically, we may find that we become more valuable once we relinquish control. This realization could be a difficult one to accept, but it’s something we must prepare for, regardless of what scifi films films like I, Robot or The Terminator suggest about the dangers of such a future.