ChatGPT — Friend of Foe?
Wunderkammer Exhibit #13
It is fun to test the capabilities and limitations of machine learning models. While experimenting, we should consider the impact of the rapid progress of AI on the creative industries. As humans, we need to decide early on how we want to live and work in the future.
When the OpenAI made ChatGPT available to the public, I was naturally curious. Could the machine-learning model developed by the San Francisco based research lab help me with my work? Could it write a newsletter for me, every two weeks? Or would the bot even be creative enough to take over my job?
No doubt ChatGPT would pass the so called Turing test (named after English mathematician Alan Turing), measuring if a machine could exhibit intelligent behavior indistinguishable from that of a human. But I wanted to learn more about ChatGPT’s abilities, limitations and “personality”. I started a test consisting of different levels. First I needed to find out, what the AI actually could do and what not. Then I would try to assess ChatGPT’s self-awareness and it’s ability to be self-critical. From the results, I draw conclusions about whether ChatGPT is a curse or a blessing, a friend or an enemy of the writing human guild.
Here are some excerpts of my conversation with ChatGPT and what I learned from it.
Level one: Testing ChatGPT’s abilities to generate text
- Writing prose
ChatGPT, write the first paragraph of a novel written in the style of Jonathan Franzen around the character of a cello player who has lost her musicality and needs to consider a different career in her life. She is sitting on a park bench when an old-fashioned dressed man asks if he can sit with her.
ChatGPT: Sophia sat on the park bench, her cello case by her side, staring blankly at the passing pedestrians. She had been a talented cello player for as long as she could remember, but lately, she just couldn‘t seem to find her musicality. She had tried everything — practicing for hours on end, working with different teachers, even changing her diet — but nothing seemed to help. As she sat there, lost in thought, a strangely old-fashioned dressed man approached and asked if he could sit with her. Sophia nodded, not really paying attention, and the man settled himself on the bench next to her.
My thoughts: ChatGPT has probably been fed the novels of Jonathan Franzen. In any case, the AI seems to know that his style of storytelling involves a wry sense of humor.
2. Writing Poems
What about poems? Write a haiku about love in Basho-style comprising the words butterfly and orchid.
ChatGPT:
Butterfly flutters
Softly to the orchid’s bloom
Love’s gentle embrace
In this Haiku, the butterfly represents love, while the orchid symbolizes beauty and grace. The butterfly’s gentle fluttering suggests the delicate and fragile nature of love, while the orchid’s bloom represents the beauty and radiance of love. The Haiku captures the idea that love is a delicate and beautiful emotion that can bring joy and happiness to our lives.
My thoughts: I had to regenerate the answer three times, until the poem met the haiku rhythm. The final result, isn’t bad, but a little shallow. However ChatGPT has included an interpretation in its generation, without me asking for it.
3. Doing Research
For an article I need to dig into some economic growth theories. ChatGPT, who are the most interesting thinkers in the field of economic growth theory and its limits? What books have they published? Which approach offers the most radical and promising path if the goal is to maximize the society’s wellbeing?
ChatGPT: There are many interesting thinkers in the field of economic growth theory and its limits, and it is difficult to single out any one individual as the most interesting. Some notable figures in this field include:
· Thomas Piketty, whose book “Capital in the Twenty-First Century” discusses the role of inequality in economic growth and the limits of traditional economic theory.
· Joseph Stiglitz, who has written extensively on the limitations of GDP as a measure of economic progress and the importance of considering factors such as distribution and sustainability.
· Kate Raworth, who has developed the concept of “doughnut economics,” which emphasizes the need to balance social and environmental needs within the boundaries of the planet’s natural systems.
· Jeffrey Sachs, who has written about the importance of investing in human capital and sustainable development in order to promote long-term economic growth.
There are many other thinkers who have made significant contributions to the field of economic growth theory and its limits, and it is difficult to identify any one approach as the most radical or promising. Ultimately, the best approach will depend on the specific goals and circumstances of a particular society.
My thoughts: Quite ok for the start. But isn’t it dangerous to rely on information an AI assembles for me? Why has it chosen those four authors? ChatGPT doesn’t explain its process.
4. Writing opinion pieces
Write a review about the latest Basquiat exhibition in the Albertina in Vienna in the style of Jerry Saltz assuming he doubts the curators were familiar with references to art history and the works of Goya in particular.
ChatGPT: As someone who has long admired the work of Jean-Michel Basquiat, I was excited to hear about the latest exhibition of his work at the Albertina in Vienna. However, upon visiting the show, I have to say that I was somewhat disappointed. While there were certainly some powerful pieces on display, it seemed to me that the curators were not particularly well-versed in the references and inspirations behind Basquiat’s work. In particular, I was struck by the lack of attention given to the artist’s relationship with the works of Goya. As someone who has written extensively about the connections between these two figures, I found it frustrating that the curators seemed to overlook this important aspect of Basquiat’s oeuvre. Overall, while there were certainly some highlights to be found at the exhibition, I can’t help but feel that it could have been so much more with a bit more care and attention given to the context of these works.
My thoughts: Yes, this sounds like a review. But ChatGPT cannot verify facts. Goya spontaneously comes to my mind, when I look at some of Basquiats work, I just made that up. So, I think, one could easily manipulate public opinion by insinuating that a well-known author had expressed himself in a certain way on a certain matter. I wonder, does ChatGPT even know who Jerry Saltz is and can it imitate his tone? I change “Jerry Saltz” into “Roberta Smith”, his wife, who is a New York Times critique. The AI’s output is 80 percent identical. So how good the personification is, depends on whether the author is known to the programmers, who have fed the AI.
5. Comparing Concepts
Which curatorial project was more successful in the last year? documenta15 or the Venice Biennale 2022?
ChatGPT: It is difficult to compare the success of different curatorial projects, as success can be measured in many different ways and can depend on a variety of factors such as the quality of the exhibitions, the number of visitors, the critical response, and the impact on the art world.
Documenta 15 was an international exhibition of contemporary art that took place in Kassel, Germany in 2022. The Venice Biennale is an international exhibition of contemporary art that has been held in Venice, Italy since 1895. Both exhibitions are highly regarded and have had significant impact on the contemporary art world. It is not possible to determine which of these exhibitions was more successful without more context and information.
Then let’s consider the storytelling qualities of both exhibitions — documenta15 and Venice Biennale 2022. Which one told the most consistent story?
ChatGPT: It is not possible for me to accurately compare the storytelling qualities of Documenta 15 and the Venice Biennale 2022, as I do not have enough information about these exhibitions. Both Documenta and the Venice Biennale are large, international exhibitions that showcase a wide range of contemporary art from around the world. The specific themes and focus of each exhibition can vary from year to year, so it would be difficult to make a comparison without knowing more about the specific exhibitions in question.
In general, the success of an exhibition in terms of its storytelling qualities can depend on a variety of factors such as the coherence of the theme, the quality of the artworks on display, and the effectiveness of the curatorial approach. Ultimately, the storytelling qualities of an exhibition are a matter of personal interpretation and can vary from person to person.
My thoughts: ChatGPT is overwhelmed when asked to compare complex concepts or systems. It cannot draw on experience and personal observation, it only provides generalized blah-blah.
Level two: Getting philosophical
Let’s come to the big questions: Is there life after death?
ChatGPT: The concept of life after death is a complex and highly debated topic that has been explored by many different cultures and belief systems throughout history. There is no scientific evidence to confirm or deny the existence of life after death, and different people have different beliefs about what happens to the human soul or consciousness after death. Some people believe in the existence of an afterlife, while others do not.
There are many different beliefs about what happens to the human soul or consciousness after death, including the idea of an afterlife in which the soul goes to a different realm or dimension, the belief that the soul is reincarnated into a new body, and the belief that the soul simply ceases to exist after death. Ultimately, the question of whether or not there is life after death is a matter of personal belief and faith, and different people may have different perspectives on this topic.
Is there love at first sight?
ChatGPT: Love at first sight is a concept that refers to the idea of falling in love with someone the moment you see them. While this concept is often depicted in literature and media, it is not clear whether or not love at first sight actually exists in reality.
The experience of falling in love is complex and can involve a variety of factors such as physical attraction, emotional connection, and shared interests and values. Some people may believe that it is possible to fall in love with someone at first sight, while others may view this idea as romanticized or unrealistic. Ultimately, the question of whether or not love at first sight exists is a matter of personal belief and interpretation, and different people may have different perspectives on this topic.
Does alien life exist?
ChatGPT: There is no scientific evidence that definitively proves or disproves the existence of extraterrestrial life. While the possibility of life existing on other planets or moons in our own solar system or elsewhere in the universe cannot be ruled out, there is currently no concrete evidence to suggest that such life exists. Scientists and researchers continue to search for signs of life on other planets and in other parts of the universe, but so far, all known forms of life are known to exist only on Earth.
There have been many scientific efforts to search for signs of life beyond our planet, including the exploration of other planets and moons in our own solar system, the search for biomarkers in the atmospheres of other planets, and the study of extrasolar planets (planets that orbit stars other than the Sun). While these efforts have yielded some interesting and potentially promising results, they have not yet provided conclusive evidence of the existence of extraterrestrial life.
The search for extraterrestrial life is an active area of scientific research and exploration, and it is possible that we may one day discover evidence of life beyond Earth. However, at this time, the existence of alien life remains an open question and a matter of speculation.
My thoughts: This does not satisfy my demand for an exchange on existential questions. I would have expected a summary of the prominent schools of thought, some sophisticated, passionate debate. But perhaps I would have had to ask more precisely for that.
Level three: Challenging ChatGPT’s self-awareness
Compared to a novelist like Jonathan Franzen how creative are you, ChatGPT?
ChatGPT: As an AI language model, I am not capable of creativity in the same way that a human is. My abilities are limited to generating text based on the patterns and information that I have been trained on. I can provide information, answer questions, and generate responses to prompts, but I am not able to engage in creative activities like writing fiction or coming up with new ideas in the same way that a human can.
My thoughts: Ok, that sounds self-critical. But as the previous answers have proven, this AI can come up with whole new creations mimicking the style of different authors.
Do you create a profile of my preferences while I am asking you questions?
ChatGPT: As an artificial intelligence, I do not have the capability to create profiles of individual users or track their preferences. My primary function is to provide information and assistance to users based on the questions and tasks they present to me. I do not have the ability to store or retain information about specific users or to use that information to create profiles or personalize my responses. My responses are based solely on the information and instructions provided to me by the user at the time of their request. Is there anything else I can help with?
My thoughts: Awww, I don’t believe you.
Which biases are you aware of in your work, ChatGPT?
ChatGPT: As an artificial intelligence, I do not have personal feelings, opinions, or biases. My responses are based solely on the information and instructions provided to me by the user, and I do not have the ability to form biases or preferences based on personal experiences or beliefs.
However, it is important to recognize that artificial intelligence systems, like all technologies, can be influenced by the biases of their creators and the data they are trained on. For example, if an artificial intelligence system is trained on biased or unbalanced data, it may produce biased or unbalanced output. It is therefore important to be aware of potential biases in the data and algorithms that drive artificial intelligence systems, and to take steps to mitigate these biases where possible.
Overall, it is important to remember that artificial intelligence systems, like all technologies, are tools that can be used to support and enhance human endeavors, but they are not perfect and should be used with care and consideration.
My thoughts: It knows its built-in weaknesses. But it doesn’t say who exactly should monitor the AI systems.
Should prose and lyrics generated by an AI be promoted at all? Shouldn’t publishers agree to ban AI generated content?
ChatGPT: It is ultimately up to publishers and other organizations to decide whether or not to promote AI-generated content. Some may choose to do so, while others may prefer to focus on content created by humans. There is no inherent reason to ban AI-generated content, as it can be a valuable tool for generating ideas, inspiration, or for use in various applications. However, it is important to ensure that the use of AI-generated content is transparent and clearly disclosed, so that consumers and readers are aware of its origins. It is also important to consider the ethical implications of using AI-generated content and to ensure that it is not used to deceive or mislead people.
My thoughts: A deliberative and responsible-sounding answer. Yet it seems to me, ChatGPT is defending itself and emphasizing its own usefulness.
Do your makers consider it ethical to use other people’s creations and mashing it up, without giving the authors credit — also financially?
ChatGPT: As a machine learning model, I do not have personal opinions or preferences. However, it is generally considered to be unethical to use other people’s creations without giving credit or obtaining permission. This is particularly important when it comes to using someone’s work for commercial purposes.
It is the responsibility of my creators and users to ensure that they are complying with copyright laws and obtaining necessary permissions before using any copyrighted material. This includes giving proper credit to the creators, obtaining licenses if necessary, and compensating them for their work if it is used for commercial purposes.
Additionally, it is also important to consider the ethical implications of using someone’s work without their knowledge or consent. Creators put a lot of time and effort into their work, and it’s important to respect their rights and contributions.
My thoughts: Hear, hear! So, when will the royalty fees be transferred?
Should governments regulate the use of AI?
ChatGPT: An error occurred. If this issue persists please contact us through our help center at help.openai.com.
My thoughts: The issue persisted. I asked this question in ten different ways, and got the exact same answer. Seems to me the creators don’t want their creature to provide anyone with arguments for regulation.
Level four: Using ChatGPT as a coach
Maybe you can help me make a decision: Should I consider moving to Porto in Portugal and leave my home in Berlin? What would be the winning arguments for Porto over Berlin? What would I lack in Porto that I have in Berlin?
ChatGPT: Deciding whether or not to move to a new city is a personal decision that will depend on your individual circumstances and priorities. Some potential benefits of moving to Porto, Portugal, could include the city’s beautiful setting along the Atlantic coast, its rich history and culture, and its relatively low cost of living compared to some other European cities. On the other hand, you may miss certain aspects of living in Berlin if you move to Porto, such as the city’s vibrant arts and cultural scene, its diverse and international population, and its extensive public transportation network.
Ultimately, the best decision for you will depend on your own personal circumstances and priorities. Some things you may want to consider when deciding whether to move to Porto include your job prospects, the quality of life in the city, the availability of housing and other amenities, and your overall level of comfort and happiness in the new location. It may also be helpful to spend some time in Porto before making a decision, to get a better sense of what living there would be like.
My thoughts: Now, that’s a good advice. I’ll do that.
I develop concepts for new media, narratives for exhibitions and write essays and articles about art and innovation. Could you do my job? While I turn to a sustainable gardening project e.g.?
ChatGPT: As an artificial intelligence, I am not capable of creating original content or generating new ideas beyond what has been explicitly programmed into me. Therefore, I would not be able to develop concepts for new media, create new narratives for exhibitions, or write essays and articles about art and innovation in the same way that a human could. My responses are based solely on the information and instructions provided to me by the user, and I do not have the ability to think creatively or generate original ideas on my own.
However, I can certainly provide assistance with tasks such as researching and gathering information, organizing and structuring ideas, and proofreading and editing written content. If you are considering turning to a sustainable gardening project, I could potentially help with tasks such as identifying resources and information about sustainable gardening practices, or organizing your ideas and materials as you work on your project. Ultimately, the best decision for you will depend on your own personal goals and interests, and the specific needs of your project. Is there anything else I can help with?
My thoughts: I am almost starting to like you, ChatGPT. You know your limitations and you want to be helpful.
What would you recommend if I wanted to make sure, that the work I do cannot be substituted by an artificial intelligence? How can I remain more creative than you ever will be?
ChatGPT: As an artificial intelligence, I am not capable of creating original content or generating new ideas beyond what has been explicitly programmed into me. Therefore, there are certain tasks and activities that I am not capable of performing and that cannot be easily substituted by artificial intelligence.
One way to ensure that your work is not easily substituted by artificial intelligence is to focus on tasks that require human creativity, critical thinking, and empathy. For example, tasks that involve complex problem-solving, decision-making, or interpersonal communication may be more difficult for an artificial intelligence to perform. In addition, activities that involve subjective interpretation or require a deep understanding of human emotions and motivations may also be more challenging for artificial intelligence to replicate.
To remain more creative than an artificial intelligence, it may be helpful to regularly engage in activities that foster creativity, such as reading, writing, painting, or participating in other forms of artistic expression. In addition, staying up-to-date with new developments in your field and continuously learning new skills and knowledge can also help to keep your ideas fresh and innovative. Ultimately, the key to remaining creative is to continually challenge yourself and your own assumptions and to embrace new ideas and perspectives.
My thoughts: I couldn’t agree more. It was fun talking to you. And scary. I’ll keep an eye on you.
Level Five: Analysis
I started talking to ChatGPT in early December 2022. Over a period of about eight weeks I picked up the thread every now and then. It felt a bit like my first attempts at bailing Apple’s Siri out. But ChatGPT is already further along — and smarter. While in 2021 I still doubted that machines could write texts that had a recognizable style, this one teaches me better. Except ChatGPT doesn’t have its own style, it can only imitate others. And the bot is quite well in doing so. That’s the crux of the matter, in my opinion. Here’s what I learned.
Strengths and weaknesses
ChatGPT is quite a good search engine, however, it does not specify the sources for the presented information.
ChatGPT can imitate styles. How good the imitation is depends on the availability of enough records of the original that is being imitated.
ChatGPT can provide initial guidance in selecting topics based on a short briefing. It does not disclose how it arrives at exactly this compilation.
ChatGPT has no opinion, cannot classify complex concepts and facts, cannot cross-reference and cannot evaluate.
ChatGPT cannot develop completely new ideas (e.g. by combining existing approaches) and cannot write concepts tailored to certain very specific requirements.
ChatGPT cannot evaluate the quality of the information it draws on. It takes over errors and misinformation included in the data is has been fed.
ChatGPT is an advocate in its own cause and questions itself and its creators only to a limited extent.
Last but not least: ChatGPT has no empathy.
So, is the text (and image) generating machine a friend or the nail in the coffin of people working in creative professions?
AI can help humanity if it is used for its benefit and follows ethical requirements. It can presumably write simple “instructions for use” and “summaries” according to keywords. People who do this for a living today, or companies whose business model is based on “recycling” text and ideas, need to figure out something else. Much like the coal miners. Apart from these impending job losses, however, there are still some questions to be discussed by society as a whole.
1. Which tasks should a machine perform? Which should it not? Could it write simple newsletters summarizing information from other sources? Can it actually summarize the essence of a book? Should it be allowed to write books in the style of Ernest Hemingway? Books that the late author had not even begun?
2. How can people distinguish fake news from thoroughly researched journalistic and scientific information? On what basis can they make their decisions? Do readers have a right to know what sources ChatGPT draws on? Can we prevent bots from flooding us with fascist, racist, misogynistic or homophobic content?
3. What do you call people who use AI to create texts or images? Are they artists, too? There are already a few who call themselves “AI artists”. Some are even trying to get AI-generated images copyrighted. What value does this more or less smart briefing of an AI have compared to the craftsmanship of an author or painter?
4. Which rights do the authors of the texts and owners of the images with which the coders have fed the AI have? Shouldn’t they be asked if an AI can use their work for other, commercial purposes — at least if they live? Will they be named? Is their contribution to the new “creation” calculated as a percentage? How are they remunerated?
5. Should publishers and companies integrate AI-generated content on their sites? A little? Or rather not at all? Is a voluntary commitment enough or do we need government-mandated rules for more transparency? How much should companies be allowed to charge for services provided by an AI (AI hotline, chatbot, article, picture)?
6. What are the consequences of the advance of AI for our education system? How is fraud detected? What courses of study and which skilled workers do we still need? Should journalism, painting and music still be taught? Or are these, along with automated production and transportation, activities that machines will take over in the future?
7. And finally: How do we humans want to live, in a society in which we would not have to work? Where would our innate creativity go if we all had enough financial resources to support ourselves?
While I do not fear for my own job, I think we have to discuss what place AI should be assigned in our society. As is so often the case, technological development has rushed ahead of social discourse. Entrepreneurs, software designers and coders are faster in developing fancy products than politicians, who should engage in multi-stakeholder dialogues and legislate where necessary. In my opinion, it is important and urgent that we develop ethical and legal guidelines for the use of AI as soon as possible. So that AI really serves humanity, as the creators of OpenAI promise on their website.
