Creativity and innovation arise from motives
How a mind can invent brand-new concepts
This is the twenty-fourth post in a series on AGI. You can read the previous post here. You can also see a list of all posts here.
Here’s a riddle: which came first, the concept of a telephone, or the first physical telephone?
If you follow an empiricist approach to concept formation — which currently dominates cognitive science — then concepts like telephone would only be created in a mind once it has experienced a number of examples of telephones. The mind then unifies these under one umbrella. This is an “objective” approach— as in, the objective world drives how you learn a concept. But when faced with the above riddle, this explanation encounters a paradox. The first person to learn the concept of a telephone would have had to have experienced “telephones”; yet how can someone build a physical telephone without already having a concept of what they are trying to build?
Whatever your answers to these questions, they have significant implications for AI, since they get at the heart of human creativity. Telephones are one of many creative human inventions. No doubt an AI that can invent the next smartphone, or cheap renewable energy would be a welcome addition to our society. Such acts of innovation always involve developing a new paradigm, subtended by one or more new “concepts”. Even outside the field of technology, all human innovations involve conceptual novelty, for example in the arts (impressionism, minimalism), philosophy (idealism, deontology), sciences (the scientific process, double-blind experiments), mathematics (non-Euclidean geometry, algebra), business (lean methodologies, the assembly line), literature (romanticism, absurdism), etc. All are concepts which didn’t exist historically; and then suddenly they did.
The expectation that a mind — or AI — should stick to modelling existing concepts by experiencing examples of them is thus an expectation of the absence of any real novelty. This is a recurring problem with modelling objective truth: it glues you to the past by condemning you to study only what has been. Objective ontologies must always defer to what already exists, not what could exist differently, or be better¹. It entrenches the belief that new things come from doing more of the same, from generating samples within the space of an existing distribution, as is the case with generative AI models. For new concepts to be created, the agent must be allowed to derive them by resolving its own needs and tensions, from the ground up.
Necessity is the mother of concept invention
You’ve heard that necessity is the mother of invention, but it has never truly been recognized how deep this goes. It is not just the act of invention that is driven by a need, but the very concepts inherent in the result arise from it like flowers from a seed. What you invent, and how you conceive of it, both bear the stamp of the original tension they resolve. For example, it is well-known that the invention of the photographic camera in the late 1800s pushed visual artists away from realism towards non-representational (abstract) art². But the influence was more than circumstantial. Abstract art is called abstract specifically because it rejects representationalism —there is no other way to define the abstract art. The need that was satisfied remains the heart of its ongoing identity, even now.
As a human you are a kaleidoscope of diverse needs, a battleground of tensions you revolve through from childhood to maturity and beyond. All the while you are discovering new modes and spaces of thinking that resolve these tensions. Such resolutions may be offered up to you by other people, or you may invent them yourself. The difference between these two paths is merely the difference between someone giving you food, and you foraging for it yourself. Either way you must still be the one to eat and digest it. Every concept you engage in, even ones that are already widespread in your culture (like telephone), you must discover or invent for yourself as though it was brand-new.
No one can give you a concept, you always create your own version for yourself from scratch. And you will only pick one up when it is useful or convenient for you. The fact that other people also happen to have delved into a similar set of situations, and already designated a word for their own use cases does not substantially influence how you yourself arrive at the same ends. Other people are still only part of your external experiences, just like the rest of the physical world. They have no privileged status that allows them to insert concepts directly into your brain. You are always the gatekeeper — you can absorb or ignore others’ suggested interpretations as you please. This is why the definition of a given concept evolves with every new generation that employs it: each generation has their own, slightly different needs.
There are some people who argue that human concepts are innate, and others who argue that they are learned. As you can see from the above, both are right. They are innate since the human motives that underpin them are quite consistent throughout times and cultures, so regardless of the specifics of your milieu, they will arise with little effort. And they are learned because the specifics of the environment must be incorporated into the discovery process. The concept will always be at the intersection of your needs and what reality can provide. The needs are like magnets that pull experiences to them — the resulting clump of metal is a “concept”.
If you’ve been following along this post, and the series in general, then the riddle of the telephone above will be an easy one to answer. Telephones, in practice, are a way of using existing technology to resolve issues of distance and communication. The first person to create a telephone merely worked his way through practical solutions to this need, and did not require an explicit concept of telephone to unify it. The concept of the telephone lay dormant behind an ongoing tension he experienced of being unable to communicate with distant people. It then surfaced once the technology worked. Note, the specifics of the solution matter; it is not just any solution to the problem of communication that is called a “telephone”.
Even once he invented it, telephone was not a concept yet. Explicit conceptualization happened later, when the inventor tried to verbally express what he had made, say, to sell or promote it. It was his desire to communicate his solution to others, along with coming up with a clear definition of its features, that gave birth to the identifying word. However, this latter desire is not why others learned to say “telephone”. They didn’t care whether the technology sold, they only cared about their own needs —i.e. to communicate. The same driving force that originally pushed the inventor to invent, now pushes them to use the word; and it has been the case ever since. Telephones have maintained their conceptual identity through over a century of evolution because of the underlying needs they satisfy have not changed.
In hindsight the original inventor may have realized he was always creating a “telephone”, but this was not what he was doing at the time. The act of unifying thoughts and experiences under a concept-word is a retrospective action. Similarly, from the outside we might imagine that the inventor knew he was always working on inventing a telephone. Our own retrospective desire to clearly define what he was doing causes us to project the concept onto his intentions; but again this misses the point. The tension he was resolving could not have, by definition, contained the resolution as part of its cause. He was not thinking “how can I invent a telephone?”, but rather “how can I communicate using electro-mechanics?”
Nor did this act of invention have to happen in physical reality. He could have imagined resolutions to his driving tensions, and given a name to an invention that had not yet been realized. We do this all the time with unrealized inventions like cold fusion or AGI. If, as discussed in this post, imagination is a type of self-generated experience, your response to real and imagined experiences is fundamentally the same. You can create concepts out of experiences that you have merely imagined (i.e. thoughts) just as easily as ones you experience in reality. This is why the question that began this post seems so tricky — depending on what path the inventor took, the first physical telephone may have either preceded or come after the concept in the person’s mind. Either way, the key to resolving the chicken-and-egg problem is to give up the notion that minds try to model objective reality. Rather they create their objective reality as a moment-to-moment resolution of their tensions.
In the next post we’ll discuss the relationship between identification and interpretation, followed by a discussion of how the act of identifying entities is a critical aspect of logical deduction, induction, and abduction.
Next post: Identification and interpretation
¹ For many people, this is sufficient, and even preferred. Novelty can be scary.
² This simplifies the historic move from representationalism to abstraction. There were many other factors, not the least of which was that abstract art was a criticism of the use of the camera itself, as seen in the works of Warhol.
