AI: Creativity and Productivity as A Complicated Algorithm
By Divasha Renata Nareswari
For decades now, even before the term was officially coined, the concept of Artificial Intelligence has aroused a lot of interest and dispute regarding its potential influence on our societies. During the 18th century when our world first entered the Industrial Revolution Era, the transformation of machines in our manufacturing processes was met with strong criticisms, let alone the idea of computers being able to mimic a human's brain. We’ve come a long way since then, as AI is no longer viewed as a futuristic vision but is already an active technological field being deployed amongst many forefront industries.
However, accepting computers’ role in more passive and autonomous functions in our society is a relatively easier feat. Though without a doubt a continuous re-evaluation of its amalgamation within the community is necessary, we can recognize that there is much profit to gain from having them supersede some of our human tasks. But their entrance into the creative industry still receives heavy opposition as many resist and alienate their place within our art space.
This essay aims to discuss the ways that AI has impacted our societal environment both passively and specifically in supporting our creative endeavors.
AI in Society
Overall, productivity is the central outcome of AI’s capacities in our contemporary society. Its emergence has dramatically improved the efficiency of the tasks that were once solely performed through manual labor. The four main reasons to employ AI are that they can perform better, cheaper, faster, and more accurately. They’re seized to perform tasks that are repetitive in nature, thus, has averted time for humans to refocus on feats that require empathic and creative efforts. As we are engaging with tasks that appeal to us more on an innate level, there is ultimately an increased amount of happiness and satisfaction we derive from performing our social duties.
Few sectors that AI has left transformative footprints on:
Business
AI is helping businesses capitalize on big data and gaining more useful insights into consumers’ habits, ensuring easier transactions happen within all industries.
Healthcare
Surgical procedures have now welcomed the entrance of robotics. In medical areas such as dental implants and neurology, it has helped reduce the effects of hand tremors and constraint accidental movements. [1] AI can assist in the diagnostic processes by analyzing various medical images of CT, MRI, or ultrasound scans automatically and provide feedback on what it can detect.
Security
Companies are starting to develop systems that can detect whether someone is likely to shoplift based on their mannerisms and behaviors, before the actual act of theft. The system then weighs each characteristic for suspiciousness, and once a customer reaches a certain threshold, the system sends its alert to the appropriate staff member. This technology could be a revolutionary tool in decreasing crime rates.
Testing it out ourselves
For a group project in our Interaction Design course at Keio University, we wanted to extend our knowledge further by trying to build an AI system ourselves. We utilized a web-based tool developed by Google called ’Teachable Machine’ to create our own personalized project using image detection methods.
Our project was named Sleepiness Detection Alarm. True to its title, the project aims to detect whether the users are sleepy through recognition of their facial characteristics and pose detections. For example, if a user has droopy eyes or slouches towards a particular direction excessively, the algorithm will classify them as ‘Sleepy’ and an alarm will be triggered as a response.
The scope of our project was targeted towards online zoom classes, to keep students alert and teachers aware of whether their students are paying attention during the class period. This can lead to more fruitful and productive sessions in the midst of the pandemic, where all of us can admit to often feeling a little bit lethargic. However, the possible implications can reach much further than intended. Instead of being bound to a classroom setting, this could potentially be a useful tool for vehicles, particularly planes and cars, to ensure that the driver’s attention is not diverted at any point of the ride and the safety of passengers is protected. The use of Artificial Intelligence allowed a seamless and automatic process of safety protocols, without any intervention from humans subsequent to building the model. The model has the seeds to encompass all four of AI’s benefits if implemented (optimization, cost-effectiveness, better accuracy, and quantity).
Conducting this project, we had four people in our group that each collected around 500 photos of every position and expression predetermined. Though we did find our output to be sufficiently working, there will no doubt still be issues for different groups of people as the students in our training data set alone are nowhere near representative enough to account for all prospective users. This non-representative dataset will ultimately result in bias existing in the system.
Problems with AI
Scaling back to large-scale societal impacts, there is a multitude of dangers that can arise from using AI in our systems. Amongst some of these leading concerns is the risk of technology breaching our fundamental human rights protections such as privacy and freedom of expression.
More specifically in the issues raised with our Teachable Machine model, let’s talk a little bit about bias. On a hopeful note, algorithms consider only the variables that improve their accuracy in generating predictions, based solely on the training data used. This may erase subjectivity found in human perception as variables such as gender, race, and sexual orientation are detached from a machine’s binary process of thinking.
However, training data will also unconsciously reflect bias in human decisions that exist within a historical and social context. Machines may ultimately systemize this discrimination and pose a lot of danger if we put too much trust in the system without any intervention.
ProPublica, an investigative news publication, released an article discussing the bias found in the COMPAS recidivism algorithm, utilized in Florida’s criminal justice department. They found that black defendants are often predicted to be at higher risk, nearly twice as likely, of being misclassified for recidivism, than their white counterparts. [2]
The industry still needs to acquire more investments for research regarding how to absolve biases in AI in order to build more robust systems that minimize these types of damages.
AI in the Creative Industry
As mentioned, we allow AI to fully take on passive power mostly because we are willing to give it up. We’d be okay with letting them perform all the autonomous tasks because it doesn't threaten forthwith the moralities we possess as a society. Creativity, however, is viewed very differently. It is a characteristic or process that is believed to be native to humanity and may be considered the ultimate moonshot for machines to reach. So what exactly is AI doing in the industry as of currently?
Firstly, they are doing similar functions to their performance in other industry sectors, undertaking repetitive tasks such as being an apparatus to help in design research or discriminate between tools and techniques to expedite painting processes. However, they can now also create.
As a cinephilia with a short-lived childhood dream of becoming a scriptwriter, film had always been the medium of art that I could personally resonate with the most. I was first truly blown away by Artificial Intelligence’s creative capacities when I came across a short film years ago called ‘Sunspring’. It is a 2016 experimental science fiction film written entirely by an AI bot using deep neural networks. [3] Though most of it does appear nonsensical, it still has the same weird charm most indie sci-fi stories do.
Through this, we can understand that Artificial Intelligence helps us explore and expand the current conceptual spaces we utilize for creative output. As we feed in our pre-defined combinations of set rules and patterns, our computers can help find other probable combinations that exist in the conceptual space yet but we humans have not yet uncovered.
A prominent example of AI within this specific subset is through the use of Generative Adversarial Network Models. GAN is an approach towards Generative Modeling through using deep learning methods. Our concept space is essentially the dataset that we feed into the system and the output is novel samples that could have possibly been drawn from the given dataset. [4]
Below is an example of Art I generated using online software called RunwayML. Using a pre-trained model from the software, I utilized a training dataset of around 1500 pictures with a mixture of space and nature elements, and the output showcased a novel image with characteristics ingrained similar to the images I inputted. Moreover, the product seemed to loosen the discreteness of some segments to give an impression as if it was made by an actual painter. It seemed that other people may also be interested in the deep learning process, as I received a buyer for this artwork on an NFT gallery site called hicetnunc.xyz.
but can these pieces of work be considered art?
There’s a familiar saying that art is in the eye of the beholder, as the way we interpret and merit artworks are shaped within our own idiosyncratic intuitions. Perhaps, this is why it’s very difficult to give a concise answer to whether or not AI can become an artist. Margaret A. Boden, a researcher in the field of Artificial Intelligence and Cognitive Science, describes creativity as the ability to create ideas and artifacts that hold value and novelty. [5]
But value in itself is an arbitrary notion, often associated with humane intrinsic values of experience and sense of self. It appears to refuse to be diminished to an algorithm as computers are formalist systems that lack contextualism, described as a philosophy to emphasize the context in which an action, utterance, or expression occurs. [6] Though they can perform actions, they fail to comprehend nor are driven by the domain and implications surrounding them. They further do not seek the approval or reward of their consumers.
Next comes novelty. After our professor in the Interaction Design course assigned us to do a short literature review of the Library of Babel [7], my initial intimate outlook of creativity has somewhat learned to accept quantitative input. To shortly summarize, the literature proposes that we view creativity as if searching for an answer.
Current computers and their future prospects deal with eras of models with billions of parameters. It narrows down the amount of the search space by an iterative process of elimination until the intended item is found. We develop algorithms that allow us to inaugurate the boundaries and instructions to do so. We may argue that similar operations occur with humans, in such a way that we never form our ideas directly from complete nothingness, but are perpetually driven by our environment without active consciousness. The moment we learn a new technique or get inspired by another artist, we are likewise enlarging a search space in our brains and reduce it when producing a certain artwork. When we talk about creativity as a means to ideate, maybe computers and humans are not so different after all.
My personal answer to the question would be, not yet. If one day, external context and understanding human emotions are able to be condensed to within a search space, perhaps artificial systems may become artists. But nevertheless, not without humans and the results may still be treated differently. That is why I do consider that a more vital and valuable conversation within contemporary context is how AI can become a collaborator in our creative processes. Instead of replacing humans as artists, it may become an instrument for augmenting our intelligence.
As of currently, computers are still unable to replace higher-order levels of creativity. However, they can create environments that can optimize and catalyze our creative output. Automating proportions of creative tasks have freed up time for artists to increase productivity in other areas. With our GAN models, while the algorithm produces more similar work, artists will have to refocus to thinking of concepts and new artistic techniques to try out.
With the increased ease of producing similar works, it’s lowered the level of what it takes to become an artist. It eases newcomers in the music and art scene to help create something original. There are processes in our own brains that are difficult to articulate, and neural networks may help us visualize them on tangible platforms. There is the increased inspiration for artists to draw from, and thus, creating more incentive for people to continue producing and improving their art. Additionally, more creative jobs such as being data curators will be introduced to the industry.
We should still be cautious nevertheless. Artists may be afraid that these models will be an emerging channel for mass production and copyright infringement issues to arise on an industrial scale. Proper guidelines within this field should be set to ensure that transparency and artists' protection are maintained.
To summarize,
To be creative or to generate art is a concept that’s difficult to wholly describe. Whether you agree or disagree to view AI as artists, we can admit there is a myriad of benefits to having them as our partners. We’ve already allowed them to become an active part of our society in performing passive tasks, and though the current system still needs fixing, it’s mostly been a positive change. Rather than being scared of machines, we ought to be prepared. Maybe it is time for the creative world to truly let them in.
References:
[1]. Price, S. (2021, March 16). The future of robotics in healthcare. Health Europa. https://www.healtheuropa.eu/robotics-in-healthcare/106671/
[2]. ProPublica. (2020, February 29). How We Analyzed the COMPAS Recidivism Algorithm. https://www.propublica.org/article/how-we-analyzed-the-compas-recidivism-algorithm
[3]. Sunspring (2016). Wikipedia. https://en.wikipedia.org/wiki/Sunspring
[4]. Brownlee, J. (2019, July 19). A Gentle Introduction to Generative Adversarial Networks (GANs). Machine Learning Mastery. https://machinelearningmastery.com/what-are-generative-adversarial-networks-gans/
[5]. T. (2013, March 19). Creativity, according to Margaret Boden. The Art Blot. https://artblot.wordpress.com/2011/12/13/creativity-according-to-margaret-boden/
[6]. Contextualism. (2021). Wikipedia. https://en.wikipedia.org/wiki/Contextualism
[7]. Borges, J. L., Desmazieres, E., Hurley, A., & Giral, A. (2000). The Library of Babel. David R Godine Pub.
