Human Intelligence and Beyond
Over the past few years we have been flooded with a constant stream of articles, papers, books and online content on the topic of artificial intelligence. However, we may need to bring human intelligence back into the picture to understand the tremendous feat of trying to replicate it.
Over the past few years we have been flooded with a constant stream of articles, papers, books and online content on the topic of artificial intelligence. AI’s very definition, current and future capabilities, impact on employment and society, biases and fairness, moral concerns, etc., have been under close scrutiny and will continue to be so — if only to separate hype from fact. However, we may need to bring human intelligence back into the picture to understand the tremendous feat of trying to replicate it. Let’s take a brief glance at some of the main characteristics of human intelligence.
Human intelligence is an embodied intelligence; it resides and operates within a human body and it is open to the world. As a result, our brain — the seat of intelligence — can make use of our body’s motion potential and tools and all the sensory information captured from the environment.
We can observe an object from a certain distance, come closer to it and appreciate the details, walk around it and get a 360-degree perspective, and touch, smell, lick or listen to any sounds produced by it, ascertaining its physical characteristics (cold-hot, smooth-rough, fragrant-stinky, sweet-sour-bitter, loud-silent, etc.). We may also measure (using hands or feet), lift (working out a weight estimate and a corresponding density) and see what’s beneath it, or break it and look inside to learn its constituent parts.
A simple example may serve to further illustrate the idea of embodied intelligence. Imagine a toddler grabbing an object and smelling, licking, listening to it, shaking and banging it against the floor, breaking it apart and trying to reassemble it. She is just trying to get as much information as possible before deciding what it is or what it represents with respect to her.
At the same time, human brains can learn from repeated body activity through muscle memory and proprioception, a kind of sixth sense. Limb sensors (proprioceptors) continuously send detailed information to the brain concerning body position and forces acting upon it. Through this mechanism — turning procedural memory into declarative memory — , human bodies help create and reinforce the brain circuitry required for certain activities or skills (playing an instrument, dancing, drawing, gymnastics, martial arts, etc.).
Human intelligence is a situated intelligence; it is inextricable from the context in which it operates. Our intelligence is immersed in a specific environment in which we perceive (though human senses are limited to certain spectra or ranges), reason (we grasp and comprehend regularities through repeated occurrences and behaviors) and act (time and causality seem to be evident in our macroscopic environment).
Living and operating in such an environment — in which situations and objects are consistently perceived and behave in a certain way and appear to be the probable cause of the observed effects — for thousands of years has helped develop and consolidate what we may call “common sense” — or environment- or context-related reasoning. Making use of such common sense humans are better able to understand and comprehend a situation from its context or make predictions about the future and anticipate their actions and behavior.
Unfortunately, over time this common sense may not make sense at all; it may even end up being just a misconception — science has repeatedly shown the inadequacy of our common-sense view of the world, particularly in small-scale or high-speed environments — or a prejudice — common sense may eventually become obsolete and not applicable, turning into a stereotype or bias, an automatic labeling that is no longer useful, mindful or fair. And there is a reason for that.
Human beings, having evolved in a very particular environment, are unable to understand — left to their own capacities — other types of environment; science and technology, however, help us understand what we were not meant to understand, clearing up prevalent misconceptions. Moreover, like many other living beings, humans label objects and situations (friend or foe, good or bad, positive or negative) to speed up or automate their reaction to them and increase their chances of survival. Additionally, this labeling helps conserve energy since it avoids having to analyze the same challenge over and over again. Needless to say, survival and energy acquisition and utilization, along with reproduction, are basic goals of all living beings and always top priority.
Human intelligence is a multifaceted intelligence; it is not a monolithic cognitive function or ability but instead includes a number of capabilities or facets in complex interplay with one another. Gardner’s theory of multiple intelligences proposes and defines seven “intelligences” — logical-mathematical, linguistic, visual-spatial, musical, bodily-kinesthetic, interpersonal and intrapersonal — with the later addition of three more — naturalistic, spiritual and existential.
We may well say that, to succeed in different realms of life we need different kinds of intelligence. In other words, the instrument to process information, comprehend and learn, reason and make decisions, solve problems and attain goals, in different environments, might be different. For instance, the first two “intelligences”, logical-mathematic and linguistic, would be crucial in school and academic settings while the last two, intra- and inter-personal, would be very much needed to effectively and mindfully interact with ourselves and others. Incidentally, understanding others bestows us with a better perspective of our own emotions and thoughts (our “self”). Further, it allows individuals to collaborate and generate a kind of collective intelligence (yet another one).
Unfortunately, although it seems reasonable and useful to decompose reality into smaller parts to analyze complex problems in detail, we often fail to develop a coherent and mindful synthesis to achieve an integral knowledge. Human intelligence is a case in point.
Human intelligence includes a meta-intelligence; it not only allows individuals to solve problems or attain goals — and the corresponding learning, understanding, decision making, etc. — but also to prioritize or decide which problems to solve or goals to attain. Human intelligence evolved to tackle the constant challenges human beings were faced with, to solve their everyday problems, but also to pursue and attain human goals and objectives, beyond evolutionary pressures (survival, energy acquisition and reproduction).
In a sense, we could argue that humans are intelligent because they have objectives and goals. Humans no longer act driven by instinct alone, using their innate capabilities; they generate goals and objectives and apply and enhance intelligence to build new capabilities in order to pursue and attain those goals and objectives. This is the basis of all human progress. Being purposeful — having goals, objectives, even dreams — makes us intelligent and, consequently, human. Incidentally, living in communities maximizes the number of potential goals and the chances to attain them through human interaction. We could go so far as to say that living together makes us smarter, makes us human. We may even try to collectively eradicate misconceptions, biases or stereotypes.
Human intelligence is judgment driven; it is based on our experiences and worldview (the common sense that we talked about before). It is not necessarily morally grounded or framed, but it is definitely built upon personal judgments used to shape mental processes and subsequent decisions.
As a result, the assumptions, expectations, theories and mental models guiding our analysis may determine, to some extent, the facts we encounter. The opposite is also true; well-known scientists have rejected or tried to modify their own theories because their results or implications didn’t accord with their previous thoughts or expectations. Contemporary intellectuals and philosophers may also feel hampered or constrained in their quest for social progress by the prevailing political correctness, which undoubtedly affects their reasoning and thinking process.
On the other hand, when humans venture into uncharted territories they are able to use past experiences and judgments, even if potentially incorrect, to guess their way around, avoiding getting blocked or stuck. Past experiences become memories and memories are reinterpreted and reshaped, in light of new experiences and knowledge, every time we think about them. Through this continuous process humans draw and distil mental models, theories, hypotheses, assumptions, ready to be used when needed.
Human intelligence has a multi-level organization; it is broadly structured into brain, mind and consciousness, which, in turn, are also multi-level.
At its simplest, the human brain may be considered a powerful biological machine that processes, organizes and interprets information acquired through human senses. The mind could be regarded as the activity generated by the brain (orders and controls, thoughts and ideas, emotions and feelings) trying to make sense of the information gathered and decide the best course of action. Lastly, and probably most important of all, consciousness would be the awareness of that activity and its effects on ourselves (subjective experiences or qualia), of ourselves as the subject of that activity (which entails the creation of the ‘self’) and of the others as subjects of their own thinking processes and mental activity (theory of mind). This awareness, however, is mediated by the brain; we are not aware of reality itself directly but of the representations of reality that our brain constructs.
This brief account of human intelligence may help us gain a humbling appreciation of the limits of current AI. So-called intelligent machines don’t benefit from a mobile, supple and dexterous body to collect information or learn from. They don’t generate comprehensive contextual knowledge, specific to the environment they are operating in, essential to comprehend events and situations. They are specialized in the extreme and lack the necessary versatility to attempt different approaches in different settings. They are given the problems — in a very precise and analytical manner — and the means to solve them — both information and algorithms, or algorithms to generate algorithms — without having to choose or prioritize them or struggle all along to survive or acquire the energy needed. They don’t have at their disposal a huge repository of experiences on which to base judgments and make informed decisions, deeply understanding what they are doing, instead of just choosing according to perceived, but not understood, patterns. Finally, intelligent machines lack consciousness and fail to understand that they are the subject of their own thoughts and workings and that the world is populated by individual living beings and objects and not just data.
While it seems difficult to be able to replicate human intelligence, the acceleration of scientific progress and its repeated breakthroughs may move the issue from a matter of ‘if’ to a matter of ‘when’. There is no arguing that this is a genuine possibility. The more we study the brain the closer we will be to generate real artificial intelligence and the more we develop artificial intelligence the better we will understand our own brain.
However, humans may not need to develop human-like intelligence. In the past, they knew how birds flew but they created planes with rigid wings and propellers; they knew how fish swam but created ships with rigid hulls and sails; they knew how horses ran but created cars with rigid chassis and wheels. Thus, the objective may not be to replicate nature but to create sophisticated tools that help us solve our problems and attain our goals. In the case of intelligence the same rule applies; we may create sophisticated tools to enhance our intelligence but we may not necessarily need to replicate our intelligence. Current artificial intelligence is precisely doing this.
