GenAI-Assisted Fantasies

Hallucinations can be seen as GenAI-assisted fantasies that help us imagine impossible or very implausible worlds.

Everyone wants to ask: Is artificial intelligence really creative? We look around and enjoy how it helps us imagine. This post focuses on how even in courtrooms, museums, and learning about long-dead artists, people are using AI to explore and imagine.

In our last post, we said the creative mistakes that GenAI (Generative Artificial Intelligence) makes, which are often called “hallucinations,” would be better described as “creative fantasies.” Hallucinations would be ungrounded from facts of reality, while indeed GenAI is grounding basic points in its responses with more knowledge than any person could ever absorb. As people might, it fills in things it is unsure of with plausible responses to mollify the requester. The term “hallucination” takes us away from understanding the value of GenAI connecting the things it is sure of with plausible but fantastical material. These stumbles can be seen as GenAI-assisted fantasies, with huge and exciting potential for use. If a prompt asks GenAI to be skeptical, it will dutifully find and concoct something. When asking ChatGPT to look for literature skeptical of geoengineering, it had many good suggestions. It also included a list of things needing to be written, the silliest title it suggested may have been “The Hubris of the Stratosphere: A Feminist Critique of Solar Radiation Management.”

So before we get going, a bit more about what is signified in common language by these two different terms, hallucination and fantasy. GenAI tells us that “hallucinations” are a false sensory perception that occurs without any external stimulus. Switching to fantasy, that can be defined as imagining a world that works in a way that is different and implausible, but intriguing.

GenAI makes connections and finds patterns from vast data. People use prompts to direct it towards their purposes and goals. So far, people have faulted GenAI for every unwanted statement and fiction; these ‘mistakes’ are labeled hallucinations. 

In our reconceptualization, we call these “GenAI-assisted fantasies”: creative applications of GenAI to help us to imagine impossible or very implausible worlds. And instead of seeing the unrealities in the output as mistakes, these are confabulations, filling in spaces between things that are known with plausible connector stories. 

Such a characterization fits much better the ways that people are using GenAI to assist in their imagining things that don’t exist. It is a market that is beginning to be tapped.

Applications of this kind are already appearing in the news. As a category, it is only just getting started, so we won’t try to develop its scope in this post. Instead, we go directly to examples of how creatives are already working in the industry of creating GenAI-assisted fantasies. Both are supercharged with ethical issues and have the potential to cause history-changing social impacts, but a word on that later.

As a first example: a much loved family member is killed when they are caught up in a violent act and the case is going to court. GenAI, it is claimed, can be a tool to “bring victims of crime back from the grave” to ask them to respond to something happening now. Already, GenAI has presented a victim no longer here. It has responded to questions, speaking convincingly to the courtroom, with a personal perspective on what happened and more importantly, what needs to happen now as a result. This opens new questions of what is evidence for the law. With what guidelines and regulations is it okay to use “AI fantasies” in court? 

As our second example, we describe how GenAI has helped give budding authors the undivided attention and intuition of a best-selling author: a ‘maestro’. Ever since the arrival of the printing press, writers have had technology to distribute the same fantasy to a lot of people. After books, there was the invention of broadcast radio and then broadcast TV, which enabled creating audio-visual experiences of fantasies and getting them to mass audiences. Now GenAI provides a new window to imaginary worlds. In the hands of a talented team of creatives, GenAI can create scripts that blur the boundaries and hide the moments where facts blend with fiction. 

Agatha Christie never wrote a ten-part course on how to write a crime novel, but suppose she did; what words would she use and what secrets could she share? It has been done now with GenAI.  

A mere mortal, like real-life Agatha, only has so many hours in the day and a lot of hobbies and activities they want to fit in. Even if she had the time, we now want this course from her, but she is gone. But GenAI was enlisted to package Agatha’s personality and combine it with a single purpose: explaining the secrets of her success. This is a GenAI-assisted fantasy being offered to the world. 

How much would people pay for this fantasy and how “true to life” does it need to be in order to meet the promise that it makes? According to the commercial arm of the BBC, the answer is £70, and for this one can get 10 superlative lessons with Agatha Christie, who is considered the most successful detective novelist ever. The publicity and preview of this AI-assisted fantasy on the BBC site is confident that the ‘essence’ of Agatha is present in every word. 

The commercial success of GenAI-assisted fantasies depends on characterizing them in a way that clearly explains the service they bring, and reaching the consumers who will buy them. 

There is no way to know where this will go, but the social impact of GenAI-assisted fantasies may likely rival, if not surpass, the impact of TV, radio, and the printing press. It’s important to say that this new industry will need to take care and search with each new application for a moral framework that helps well-meaning creators to produce ‘good fantasies’ and not fantasies that cause harm. 

When “War of the Worlds” was first broadcast on radio, it sent a lot of people into panic. The radio drama is now understood as a powerful way to create and share stories, but back then, when radio itself was a new experience for society, “War of the Worlds” became a major incident when it met the public’s lack of insight into the “rules of the game.” What checks and balances will be important before this application of GenAI-assisted fantasies is approved to go ahead? And where is the decision made?

People use GenAI to reflect, to create stories, and to respond as another might. GenAI  prompts impact the tone and output and should be chosen to not inadvertently make a major fictitious incident as “War of the Worlds” did.

Berry Billingsley is an educator interested in philosophical “big questions.”

Ted Selker is a computer scientist and student of interfaces.

You Just Read

GenAI-Assisted Fantasies