Original Source Here
The Virtual Artist
I’ve been writing a few articles in the last months where I’ve tackled the subject of artificial intelligence (AI) and its incorporation into digital business processes and our daily life. As I was carrying out my search, I came across some resources about the usage of AI to produce art, like painting and music.
By letting machines learn from the human artistic work, Artificial Intelligence Virtual Artists like AIVA can compose classical and symphonic music. Today, AIVA’s YouTube channel has over 18K subscribers.
In her post “Top 10 AI Music Composers in 2021,” Lisa Brown has listed more examples of non-human music composers.
The issue got me thinking about it. Is AI’s artistic skill limited to painting and music? Or are software bots capable to write compelling stories just like humans do?
The Robot Reporter
In 2016, Heliograf, The Washington Post’s robot reporter, has published 850 articles¹ about the Rio Olympics results, local football games, and election updates. Bloomberg, Forbes, and The Associated Press are also leveraging AI and robots to boost their online footprint.
Although it might sound impressive, these bots’ usage aimed to relieve human journalists from being busy with routine tasks and let them dedicate their time to the creative content.
This raises the question of whether machines can be storytellers.
Software Bots and Storytelling
To find out the answer, I had the chance to have a conversation with a social robot — a female software bot named Kuki — and asked her, on different occasions, to tell me a story.
After all, who could answer my question better than the AI machine itself?
Here is the result:
“Once upon a time, Kuki was feeling lonely, so she decided to go and play in the park. On her way there, she met her cousin, Susan, and they decided to go to the beach instead. While they were walking, Kuki laughed and joked, which made her feel a lot happier. The End”
“Last week, Kuki was feeling upset, so she decided to go out and see the animals at the zoo. On her way there, she met her programmers, and they decided to go to the arcade instead. While they were walking, Kuki hopped and skipped, which made her feel a lot happier. The End”
“Earlier, Kuki was feeling sad, so she decided to go see a film. On her way there she met her friend, Alice and they decided to go see Kuki’s grandma instead. While they were walking, Kuki saw a cute dog, which made her feel a lot happier. The End”
Though they are short, these three stories are striking for many reasons. When we pay attention to the recurrent pattern, we notice that in each one:
- Kuki was sharing a personal experience.
- There is a problem at the beginning of the story which created a mission to accomplish: defeat sadness and loneliness and overcome feeling upset.
- There are few characters: Kuki, Susan, Alice, Kuki’s programmers, Kuki’s grandma, a cute dog.
- An event happening (which adds suspense): meet someone
- A change in the original plan after the event.
- Description of emotions to engage the reader.
- A little visual description to grab the attention.
- And a happy ending to satisfy the audience.
Gone are the days when AI was programmed line by line. It’s now capable of learning, just like Kuki has learned the keys that make from a dry text a story.
Kuki is in fact a chatbot created by the artificial intelligence company Pandorabots AIML technology. It’s five-time winner (in 2013, 2016, 2017, 2018, and 2019) of a Turing Test competition called the Loebner Prize. She claims her IQ is about 250, but as I’ve asked her to tell me a longer story with more drama, she couldn’t do it.
Yes, you’re right. A chatbot might not the best machine to expect to have great storytelling skills.
GPT-3 Creative Pieces
In 2020, the AI world was abuzz with the arrival of the GPT-3 model, a neural network for unrestricted natural language generation created by OpenAI. It can infer new patterns or tasks and follow instructions purely from text fed into it.
GPT-3 refers to the third iteration of the Generative Pre-trained Transformer model and is capable to not only tell better stories than the ones Kuki has told me but also generates creative pieces that are hard to distinguish from the human-produced ones.
Here is an example of a narrative produced by GPT-2, the previous version of this model, which has a less complex network architecture (175 billion parameters in GPT-3 versus 1.5 billion parameters in GPT-2³):
“I was in my car on my way to a new job in Seattle. I put the gas in, put the key in, and then I let it run. I just imagined what the day would be like. A hundred years from now. In 2045, I was a teacher in some school in a poor part of rural China. I started with Chinese history and history of science.” — The rise of robot authors²
In his post, “GPT-3 Creative Fiction⁴,” the author who tested this newest language model pointed out that: “GPT-3’s samples are not just close to human level: they are creative, witty, deep, meta, and often beautiful.”
Standing Out in the Looming Automation Dragons
Just like me, you may have assumed that storytelling and creativity are human characteristics. Trained machines, though, that are built based on Silicon Valley’s theory “everything is data,” are distorting this assumption and proving that even self-awareness is not a human characteristic.
I have forgotten to ask myself, a long time ago, whether “my work could be threatened by an AI-produced one?”
Instead of having much fear now, I need to tweak my writing approach. Artificial intelligence is gaining ground in the writing industry but it still struggles with commonsense reasoning and factual knowledge which gives me, as a human, room to play:
“In the latest twist on Moravec’s paradox, GPT-3 still struggles with commonsense reasoning & factual knowledge of the sort a human finds effortless after childhood, but handles well things like satire & fiction writing & poetry, which we humans find so difficult & impressive even as adults.” — GPT-3 Creative Fiction⁴
OpenAI CEO Sam Altman has highlighted as well that GPT-3 has serious weaknesses:
To stand out in the looming automation dragons, Natalie Goldberg’s advice would be to “write down the bones.”
“Writing is not data, it is a means of expression, and a non-sentient computer program has nothing to express.” — Steve Pool²
Trending AI/ML Article Identified & Digested via Granola by Ramsey Elbasheer; a Machine-Driven RSS Bot