Artificial Intelligence: Product or a Process?
in·tel·li·gence
/inˈteləj(ə)ns/
noun
1. the ability to acquire and apply knowledge and skills.
Applications of Artificial Intelligence seek to uncover the nonlinear complexities and challenging nuances of the “human.” From decoding handwriting to performing speech recognition, facial detection, semantic and sentiment analysis, and recommending content, this list goes on. Artificial Intelligence offers the potential insight to make sense of who we are and adapt to our world.
The premise then is this: if Artificial Intelligence can learn enough to understand us and our world, it can become an entity that aids and co-exists alongside that which is human. For example, if Artificial Intelligence can understand how humans drive and make sense of roads and traffic conditions, it can be packaged into a self-driving car. If a task or data can be reliably learned to perform an action, the system can insert itself as an automated alternative that optimizes the complexities in our lives.
What Does “AI” Really Mean?
As I’ve used Artificial Intelligence recently in various creative tool-making and musical contexts, I’ve realized that AI doesn’t actually mean Artificial Intelligence — AI really refers to a system that is “Artificially Intelligent.” If intelligence is defined as “the ability to acquire and apply knowledge and skills”, then intelligent loosely means “one that has/applies knowledge and skills.” Our machine learning models that perform classification, regression, and even “human-like” generation are more accurately “Artificially Intelligent.” They are all systems trained to understand a semantic structure to translate input to output (like a mapping function). While this may a gross oversimplification, this is how we view AI sometimes, a genie that translates wishes into commands. An image is translated into a detail-rich caption. A text prompt is translated into a desirable grid of colored pixels. A question for “Who is the President” is translated into “Joe Biden” (at the time of this writing).
Artificially Intelligent — Systems that are apply trained semantic structure to translate input to output
Artificially Intelligent systems are trained translators. Applications that employ these systems are like compilers, they take in human-readable input (code) and output a human-desirable product (binary executable). They can translate the current state of the world into the best action to be taken (Reinforcement Learning). Artificially Intelligent systems are not intelligence. They are trained to give specific output, not how to give output. With systems like the ever-popular ChatGPT: if the “knowledge” or Language Model supports it, it will respond. If the knowledge isn’t there, there isn’t an ability to acquire that knowledge and learn a new scenario to translate.
These systems are frankly fixed models. They are terribly powerful, but on their own don’t suffice as “Artificial Intelligence”. If our future comprises of these systems co-existing alongside us, the caveat to achieving Artificial Intelligence is that there needs a human alongside it. Man + machine is what constitutes an intelligence that is able to compute, learn, and adapt. The human is what makes sense of an Artificially Intelligent system. As more knowledge is extracted from the world around us, the human is to filter that knowledge into wisdom. As users at the dawn of an age of Artificial Intelligence, do we have the agency and vision to understand its implications? Where do we want to go with these tools, and to what end do we build? Artificial Intelligence is a process that demands a Human-in-the-Loop to understand what can be done and to critically judge what should be done.
Artificial Intelligence in the Process of Co-creativity
One particular area of Human-in-the-Loop Artificial Intelligence is this notion of co-creativity. How can researchers, tool-builders, and users create using Artificially Intelligent systems? How can Artificially Intelligent systems challenge existing notions of human creativity? This Artificial Intelligence loop is highly interactive, placing the human at the front and center to imagine, understand, and assess AI.
The success of Dalle 2 is not in its computation and visual prowess alone; It rides upon the shoulders of an active yet playful human imagination, to dream of a “photo of hip hop cow in a denim jacket recording a hit single in the studio.”
To co-create with Artificially Intelligent systems in a meaningful way, it requires a human aspect of understanding the world in a way that machines do not. SoundscapeAI is a look at the human condition and our need for calm in the day-to-day. When paired with the right tools such as WebChucK and ChAI, the human experience is magnified by a creative tool that “understands” us.
When creating co-creative tools, there is always an assessment for balance between the user and the system. Either the human or the system can contribute to the feedback loop. If the system is an input translator, then co-creativity, or working with such systems creates an open dialogue. Sure the system alone is capable but towards what end? Can it be dynamically expressive or simply deterministically reactive?
As Artificially Intelligent systems become more and more powerful, they will continue to provide incredible insights, discoveries, and opportunities. As tools, these systems and their outputs only make sense of a small portion of our world — finite input into a finite output. As Artificially Intelligent systems further develop, the difficulty is not what they can do, but how we as humans wield these technologies. As users, how can we process this knowledge with discernment and wisdom? Beyond products to alleviate our difficulties, how can they be infused into our own processes? Where do systems offer newfound agency in an area of struggle? Do we even need this system? How can we use these tools with integrity and intelligence to explore what is deeply meaningful for ourselves, our community, or the broader context and future of humanity?
