Before I delve into my research and what it proves, a good understanding of the basic issue and concepts first. So, what is generative artificial intelligence and how does it harm creators in the arts especially?
Generative AI is a type of artificial intelligence which creates original content, such as texts, images, and music. It uses common patterns it can learn from vast amounts of existing data.
Models are trained on massive datasets in order to understand a greater concept of the prompt it was told to answer by a user. It finds relevant information such as texts online, images, and posts, and condenses their meanings or uses to create these new pieces that match the request.
It simplifies many tasks for much of the general public, but hurts those it feeds from. The individuals who made the many projects it summarized and used as a basis for providing the requests asked.
These people impacted by it includes those who produce any form of media that is consumed in creative fields; individuals who paint, sell or upload their works, write, create sound design, make films, etc., whether it is done on a small or wide scale.
Rome was not built in a day and neither was the millions of projects that are created and uploaded on to the internet.
While some fields use AI for their aid, the areas in the arts are given more disadvantages than benefits.
The problem with generative AI in the arts is that it creates competition for these creators. It uses their work, for the majority without compensation, consent, or credit to rapidly produce content, that the original makers cannot keep up with doing the same at the pace set and needed for stability.
SURVEY
For my research I wanted to demonstrate how not only does the topic of AI hurting the arts remain a widespread problem, but can be found locally as well. Local for me was in the state of NC, where the problem has a home in as well, but isn't discussed as much as in other communities.
To gage public opinions, I decided to conduct a survey with various questions centered around the topic. It also included questions on demographics as well, so I could better understand my audience.
Circumstances: Submissions were anonymously recorded and questions were framed in a neutral manner as to eliminate the possibility of response bias from participants.
It was published on the internet and encouraged the form to be shared with others for a wider audience, which would make a more realistic random sample.
Results: In total there were 47 responses. The results are much more generalizable to women, who consisted of 72.3% of respondents. As for the age group most applicable to the data was tied between the 18-29 and under 18 crowd that each had 34%.
Most of the random sample lived within the Triangle area or that which closely borders it, although a considerable portion lived out of NC.
A majority of 76.6% percent, when asked what their main role with creative works answered to be a fan/consumer only.
Topic Based Questions: They were mainly framed on the likert scale, although multiple yes or no questions were asked as well as a short response based one.
"Were you aware that AI models are typically trained on large amounts of existing creative works and data?"
In response to this question, most were aware of that information.
"Do you personally use Generative AI tools in your creative or professional work processes?"
Occasionally was the highest selected option, and when the data is grouped, most people have used AI in these routines.
"Have you taken preventions for your work from being used to train AI models, such as using opt-out options, watermarks, and/or using specific platforms?"
Few people have protected their works in this way, and some weren't aware or had considered it an option.
"AI developers should be required to obtain explicit permission from creators before using their work for training data."
When this statement was posed, the most selected option was strongly agreed and the majority in general agreed.
"Creators whose work is used for AI training should receive some form of compensation."
The overall consensus for this statement was that majority agreed.
"I believe that Generative AI is a positive tool that will ultimately benefit the creative industry."
The results were more varied on this question than previous. When the strongly agreed and agreed was combined it was the highest percentage, with both versions of disagreed following, and neutral not far behind.
"What is your biggest ethical concern with the sourcing of data for AI models?"
To this short answer question, the common themes in responses were:
-
Copyright/legal issues, plagiarism, theft
-
Lack of credit
-
Jobs taken away from artists, writers, and other professionals
-
Art loses its meaning or loss of the human experience, spirit, and expression
-
Relying too much on AI
-
Lack of creativity
-
No financial compensation
-
Undecided on the matter
-
No informed consent
-
Deepfakes and not being able to differentiate AI from reality
-
Bias in data & effects on marginalized communities
-
Environmental & resource concerns
-
Privacy
"I am concerned that the current legal framework is ineffective to protect creators' rights against AI training."
In general, the majority of respondents agreed.
"What solution do you think is most necessary to address the issue of generative AI training impacts?"
Most people said stronger copyright laws.
"Do you have concerns of AI made works potentially devaluing or over-saturating the human work in creative fields?"
A strong majority of participants answered yes.
"Has the rise of generative AI made you less likely to share your work publicly?"
Most of the people selected no.
SURVEY CONCLUSIONS
In the survey's aim to gain a sense of mainly local people's perceptions on generative AI's interruption in the arts, it can be shown how the majority of individuals do think that there is an issue present and that more should be done. The type of action that the public thinks should be taken however, is greatly varied and uncertain. The firmness by the majority that something should be done, but the widely distributed answers in other questions that propose more detailed stances suggests that many people have not been exposed to this topic before or in much depth. The presence of selected answers that were neutral, or more complex than the common 'agree-or-disagree,' etc. also aligns with the idea that the issue is not having enough recognition due to informational outlets not exposing audiences to it.
By knowing that a central factor to the problem is not because people do not care, but do not are not solidly educated about it proves the need for an awareness campaign and more discussion in regard to it. Validity would have been improved if questions' data that were more applicable to creators were separated from those who are consumers only. This could have been solved by creating and specifying these questions as optional to those who it is not relevant to. This would have contained less room for confounding variables. The research also could be advanced by finding the percentages as to what each demographic was likely to answer as their beliefs on the issue. In summary, the data allowed me to find how to best convey my points to a wide audience, and the necessity for the word to be spread with greater understanding before significant change can be accomplished.
INTERVIEW INSIGHTS: INFLUENCES ON THE WRITING COMMUNITY & BOOK PUBLISHING INDUSTRY
I had the opportunity to speak with the Senior VP of Sales Strategic Planning at Penguin Random House, Julie Black, where we discussed the significant effects of AI on the publishing industry and the individuals involved. It allowed me to discover new perspectives on the topic with situations that I would not have considered or thought of, especially with AI's role involved. Hear all about it below for more details & information!
Key Ideas Discussed:
-
Copyright concerns- lawsuits over unauthorized use of content, no compensation, and need for protection of author's properties
-
Despite AI capabilities the creative process comes from human involvement
-
Resources & environmental concerns
-
AI in marketing strategies and overall efficiency
-
Advocation for larger transparency
-
Reader experiences
-
Authenticity in writing
-
Enterprise AI Safeguards
-
Limitations of protections & piracy
Through this interview I realized that the use of AI in publishing has both helpful and potentially harmful implications. For instance, AI can save editors time by scanning for typos, word duplications, word counts etc. and free up a human editor to perform other tasks. AI is also helpful in summarizing books for a general overview or to use in marketing strategies. The negative consequences of AI in publishing center around copyright issues and the loss of human creativity as the AI engines are increasingly able to mimic specific authors and writing styles an can generate “new stories” of their own. Even with the basic protections of copyright put in place, piracy and illegal uploads being fed to the large models of AI that need training cannot be prevented.
FOCUS GROUP
For the last portion of my originally performed research I constructed a focus group consisting of four high school students. Once I contained their informed consent, I gave them an overview of the studies procedures and topic to ensure that they would still want to participate.
I explained how I was hoping to use the conversations' findings to get a better grasp as to what in creative writing, stands out or strongly resonates with people and the ideas of human essence. While this was true, I hoped my real purpose that I would debrief them on afterwards, was to see how undetected and unassuming artificial intelligence could be in literary works of poetry. I chose this form of writing as it is commonly thought of as a purely human space of feeling.
The idea for the research was based off of an example from a source I had found for my paper by David Galef in the Chicago Tribune of an AI copy of poetry going undetected by a professor. For my version I showed various poems, but left out that some of them were AI produced in similar writing styles and themes, and started the first half of discussion flowing on which poems people liked, disliked, noticed, resonated with them, why, etc.
The poems used have been inserted below, and if you would like to test out your own skills in detecting AI poems check out the quiz featured on the home page!
AI Produced:
1.
When the days feel heavy
and the nights grow long,
rest your heart, dear one,
you are not alone.
I am in the quiet moments,
the gentle pull of peace,
the stillness that surrounds you
when all your worries cease.
I am the light upon your window,
the soft wind on your face,
a presence woven tenderly
through every empty space.
So when your tears fall softly,
let them fall without fear,
for love like ours is endless,
and I am always near.
2.
I’ve seen dark nights linger,
long as they please,
but even the stubborn moon
must bow to morning.
And somewhere
beyond the hush of now,
a small bright thing is rising
maybe a dream,
maybe a chance,
maybe just the sun
stretching its golden arms.
But I feel it.
Warm as a hand on my shoulder,
soft as a promise kept.
So I lift my face
to whatever light is coming,
and whisper to the sky:
I’m ready.
Because hope,
sweet, steady hope
always finds a way
to bloom again.
3.
The hesitation of the heart
Is wider than the ocean's breadth
It measures not the certainty of light
But chronicles the shadow after death
To stand upon the threshold of the knowable
And find the key is made of air
Is lesson for the eternity
That faith is only fashioned to repair
The soul, she is a tender gauge
That registers the slightest chill of fear
And every promise whispered from the pulpit
Must pass the question dwelling here
For Heaven is a hypothesis
Until we touch its farthest unseen wall
And truth is but the echo of a sound
That we are not assured
Will fall
We prize the anchor of the word
Yet find the cable frayed and wearing thin
And search the horizon for the beacon
That only flickers from within
Poet Produced:
1. By Emily Dickinson
There is another sky,
Ever serene and fair,
And there is another sunshine,
Though it be darkness there;
Never mind faded forests, Austin,
Never mind silent fields -
Here is a little forest,
Whose leaf is ever green,
Here is a brighter garden,
Where not a frost has been;
In its unfading flowers
I hear the bright bee hum:
Prithee, my brother,
Into my garden come!
2. By Mary Elizabeth Frye
Do not stand at my grave
and weep
I am not there. I do not sleep.
I am a thousand winds
that blow.
I am the diamond glints
on snow.
I am the sunlight on
ripened grain.
I am the gentle autumn rain.
When you awaken in the
morning's hush
I am the swift uplifting rush
Of quiet birds in circled flight.
I am the soft stars that shine
at night. Do not stand at my grave
and cry;
I am not there. I did not die.
3. By Langston Hughes
She,
In the dark,
Found light
Brighter than many ever see.
She,
Within herself,
Found loveliness,
Through the soul’s own mastery.
And now the world receives
From her dower:
The message of the strength
Of inner power.
Conclusions
There were no suspicions of there being AI, until I disclosed it for the 2nd half. All of the generated poems in fact were resonated with by someone in the group. I believe this is because of their copied themes that present vast interpretations on nature, emotions, spirit, etc. along with their overly detailed descriptions since it tries to mimic patterns seen. AI analyzes and perhaps the poems made by it were still liked because of it being condensed in meaning, and possibly more direct for interpretation.
Once I pointed out that there were AI poems, I asked if they could differentiate the real from fake, and if they could elaborate on what qualities each possessed.
The real poems were actually all found when read carefully.
Distinguishers of human work were authenticity, author's voice, overall makes more sense in comparison to AI, and does not feel like a combination of other author's thoughts.
AI was found to be very descriptive, but often not contributing to a singular meaning or theme.
The focus group allowed participants and me to see firsthand how AI can truly be present in plain sight, and that learning how to detect AI is good to be informed with as the technology gets more advanced.
It also demonstrated how competitive fake works can be for writers, as the AI poems were surprisingly discussed more than that of real poets, whose writing styles they copied.