Artificial Intelligence, Genuine Concern (1): Creativity
Many say that creativity is one of the fundamental human traits that differentiate us from machines: we do not only treat input information to produce an output through a mere calculation process, we create. This view is giving some magical properties to creativity (i.e., auto-determinacy). However, in our world, magic often relies on the spectator’s ignorance of the underlying trick.
An evolutionary perspective is quite helpful when it comes to better understanding human cognitive abilities. When a biological system encounters a problem to solve (e.g., a direct threat), it usually processes the information available using some sensory input (e.g., visual signal) and produces an output (e.g., motor response) that will contribute to its strive to survive and develop — this latter part is largely determined by internal information (e.g., DNA). I would argue that this is what we could call proto-creativity: both internal and external sources of information are used to produce an appropriate output. Since each new threatening situation will be different from previous ones, it is of interest of the living system to produce an original output for each new situation using both internal information (e.g., recorded past experience, either through epigenetics mechanisms or cellular/cognitive memory) and external information specific to the new, unique, situation it is facing. Here I will argue that (1) proto-creativity is not qualitatively different from what we commonly call creativity and (2) the difference between living systems and AI in terms of creativity is the processing path and not the output. In other words, while living systems rely on some adaptive flexible information processing and AI mostly on big data processing, both outputs could be equally considered as “creative”.
What do we usually call creative then? An entity’s production from which the (internal and external) causes are hidden enough from the observer so that it looks new and surprising, unanticipatable. There are other human constructs that exist thanks to our ignorance as finite beings. A striking example to me is free will, which is relevant to evoke here as it shares a requisite with creativity: both notions need a concept alike to the one of soul to make sense. Following a folk definition, they would need to emerge from virtually nothing, to break the causal chain and be their own starting point. Both tell something very central regarding our perception of ourselves.
Don’t get me wrong: human creativity is wonderful (at least from my own, ignorant, perspective). But knowing that it might not be fundamentally different from any other information processing does not change anything to it. Plus, its underlying processes remain — for now — unique in their efficiency. My point is rather that acknowledging this progressive decentralisation of human cognitive abilities is necessary as progress is made in the field of AI. As we are slowly accepting that we are not radically different from other animals, we might soon have to extend our family nucleus to machines. However, I do not think that machines will ever reach human cognitive abilities as defined nowadays, but rather that we will reevaluate our beliefs regarding human nature.
In this series, I will argue that we should be concerned about future development in AI not because of ethical considerations regarding machines (i.e., artificial consciousness, suffering, slavery, etc.) but because of how much human values might be impacted by future technological advances. This reflection that AI will give us might quickly change our current moral system and impact human societies like never before. Although this might not be a problem per se, we might want to somewhat anticipate these potential radical changes in how we, human beings, perceive ourselves, as this is the very basis of our societal system.
