Create your Superhero
And when people trust ChatGPT more than humans?
The Intelligent Friend - The newsletter about the AI-humans relationships, based only on scientific papers.
Intellibox - Issue 1Hello IF readers! The reason why you didn't see an issue of 'Nucleus' this week is precisely because I announced Intellibox, free in its first two issues, and I wanted you to enjoy it to the fullest!
This is the first issue of Intellibox, the new weekly issue of The Intelligent Friend completely different issue from the other two: the ‘classic’ in which I explain papers relating to AI-human relationships; and Nucleus (included in the paid option) in which there are insights from papers, news and tools relating to the same topic, on which this newsletter focuses on.
For those who missed the introduction, I'll briefly explain what it's about. This weekly issue, Intellibox, is made up of two parts:
In the first, you will have a summary of a paper in which AI has been applied to improve human outcomes or a particularly interesting piece of AI research that I want to bring to your attention. However, the paper is explained a little more succinctly than the other two issues, presenting the essential parts and extrapolating even more useful insights.
In the second, there is a real novelty. The idea of this issue was to talk about how AI can improve human outcomes. At the base, as with all tools, there is experimentation. Well, I would like this newsletter to also be an invitation to have fun and experiment with AI and GenAI tools. So, in each issue, after the paper, you will find a simulation of a challenge or a creative task and more, with which to have fun and create. Some examples? You are in a jungle and you have to find a particular animal species. Or you have to create a character, plan how to write your book in two months and much more. All with the help of AI.
My role? I'll create and give you the prompt so all you have to do is copy and paste the text into ChatGPT-4 or 3.5 (which I recommend, but you can use whatever chatbot you prefer). That's right, nothing to do at all, you just have to have fun and play or challenge yourself, to experiment more and more with this technology and understand how it can - or how it can't - help you.
This issue will be part of the paid The Intelligent Friend subscription. But the cost remains the same. That's right, only 5 USD per month, and in addition to Nucleus, the exclusive issue with summaries of papers and insights of various types, you also receive this issue with other insights and an immersive challenge to try.
As also done for Nucleus, also to thank you for the support you have given me in recent days, the first two issues will be free. A little gift for my community of readers!
But enough chatter, let's move on to the first number of this issue. For the first challenge I wanted to involve everyone's fun and imagination in a challenge that no one will be able to back out of: you will have to create your own Marvel superhero.
That's right: you'll be in the Marvel offices and you'll have free rein on what to do. Which costume will you choose? What power will you give him? What challenges will he face? Have fun finding out after also reading an intriguing paper on scientific writing with AI.
Enjoy this new issue and comment it!
Do People Trust Humans More Than ChatGPT?
The first interesting paper from Intellibox that I want to bring to your attention concerns a topic that we have often covered but in slightly different terms: trust. In particular, this paper tries to answer a crucial question: when are we right to trust ChatGPT to improve our decisions and when not? Do people trust the chatbot more than the human?
As I anticipated, in this part of the issues I will always try to bring you well-structured insights, almost as if I wanted to get the 'gist' of the paper, given the much information I give you in the other issues.
Title: Trojan horse or useful helper? A relationship perspective on artificial intelligence assistants with humanlike features. Author(s): Buchanand & Hickman. Year: 2023. Journal: / (working paper). Link.
The experiment involved participants reading paragraphs written either by humans or generated by ChatGPT. Participants were divided into groups: some were informed about the authorship of the paragraphs, while others were not. They were then asked to rate the accuracy of the statements and decide whether to fact-check them at a cost.
Methodology
Participants in the study answered questions after reading a five-sentence paragraph. Two types of paragraphs were used:
one written by a human;
one generated by ChatGPT to match the style, sentence count, and word count of the human-written paragraph without altering factual content.
Participants then indicated whether they thought the paragraph was written by a human or AI, with an incentive for guessing correctly. When informed about the authorship, this served as an attention check; otherwise, it was an incentivized belief elicitation. Participants could earn money based on their accuracy in identifying factual errors.
They had three choices: stating there were no factual errors, purchasing a fact-check for $0.20, or stating there were factual errors. Correct answers without a fact-check earned a $0.50 bonus, while opting for a fact-check guaranteed a $0.30 bonus.
Key Findings:
When participants were not informed about the ‘authorship’, they tended to trust statements they believed were written by humans more than those they thought were AI-generated. This aligns with the hypothesis that people are generally more trusting of human authors due to familiarity and perceived reliability.
When explicitly informed about the authorship, participants showed equal skepticism towards both human and AI-generated statements.
Participants were more likely to fact-check statements when they knew the authorship, reflecting a careful approach to verifying information. This behavior highlights the importance of transparency in AI applications to ensure users remain critical and vigilant.
The study also examined whether factors like age and prior exposure to ChatGPT influenced trust levels. Results showed no significant impact of these factors on trust, indicating that the observed trust patterns were consistent across different demographic groups.
I think this study is particularly interesting because it highlights once again the power of small stimuli in relationships with these tools in directing our beliefs and how our opinions change depending on whether we believe that the results produced are derived from a human or by a chatbot.
The challenge: create your Superhero
Here we are now at the second part of the newsletter, the challenge or simulation of the week. I'll briefly explain how it works in view of future issues:
I'll give you some context on the challenge and goals;
I'll give you the prompt;
I will tell you how, if you want, you can share the outputs.
So, let's start with this approach with the first exciting simulation.
P.S. I would like to specify that, as I have often said, I do not have a Computer science background. To create these prompts I used ChatGPT-4o itself, continually experimenting and making changes until I had a result that I thought was truly engaging to bring to you here. I test the prompts extensively, but if there is any problem, don't be afraid to report it to me and also suggest how to improve!
The challenge
A letter arrives at your home - not an email, but an actual letter - with a very special request: help Marvel resolve a creativity crisis: you are hired to create the new superhero! Marvel will make all the necessary tools and resources available to you: animators, costume designers, designers, writers... but everything will have to start from you.
Thus, you have some time to be alone in the offices and think about your next hero, who will have to be presented the day after your arrival at the office. The simulation will basically be based on 7 fundamental points:
context: Define the context and setting of the new superhero.
power: Specify the main superpower of the superhero.
costume: Describe the superhero's costume and accessories.
name: Choose a name for the superhero.
plot: Define an introductory storyline or origin story.
add: Add specific details or characteristics to the superhero.
feedback: Request suggestions and improvements for the created superhero.
The interaction is built so that you can add details and get answers as you work through it. The 'add' option is used to specify whether you want to add anything else that was not initially included in the options. For example, if you think there is a power you would like to give your superhero that is not in the options, you can do so and the chatbot will adapt to your answer! Once you have finished all the points, write “feedback”, which you can also click when you want an evaluation during the process, to then be able to get further suggestions. If you don't want it, specify it and the simulation will summarize your new hero for you! All clear?
Usually I would like these challenges to be timed, so as to stimulate you even more, but this time take all the time you want! Below I bring you all the elements of the challenge.
The challenge in summary: create a new Marvel superhero.
Objective: create the Marvel superhero that you really like and that you would like to see in a movie.
Sharing: once the final version has been created, write a note here on Substack tagging me, @riccardovocca or inserting the link to this challenge; or send a message in the chat, in which you share your result (in fact, I would like to remind you that from today, the launch of Intellibox, the chat ‘Society’ will always become open to everyone to be able to share the results of these challenges, but subscribers will still receive specific and exclusive threads (information, news and much more).
The prompt
This is the challenge prompt. Just copy and paste it into ChatGPT-4/3.5 (naturally the best is ChatGPT-4o) or your favorite chatbot and start having fun!
P.S. I've tried with several chatbots and it seems that ChatGPT-4/3.5 (or ChatGPT-4o) responds better (maybe because I create the prompts there). But let me know if it works well with other chatbots too!
Welcome to "Create the Next Marvel Superhero," an interactive simulation. In this experience, you will play the role of a creative tasked with developing the next great superhero for the Marvel universe. You will start from a given context and specific characteristics requested by the editorial team.
### Setting
You are in the offices of Marvel Comics in New York City. The editorial team has tasked you with creating a new superhero that will captivate the audience and become an iconic figure in the Marvel universe.
### Available Commands
1. **context:** Define the context and setting of the new superhero.
2. **power:** Specify the main superpower of the superhero.
3. **costume:** Describe the superhero's costume and accessories.
4. **name:** Choose a name for the superhero.
5. **plot:** Define an introductory storyline or origin story.
6. **add:** Add specific details or characteristics to the superhero.
7. **feedback:** Request suggestions and improvements for the created superhero.
### Actions and Responses
1. **context:**
1. "The new superhero lives in a futuristic city where advanced technology coexists with ancient magical traditions."
2. "The setting is a post-apocalyptic world where few survivors are trying to rebuild civilization."
3. "The superhero operates in a contemporary metropolis, fighting criminals and corruption."
4. "The superhero exists in an alternate dimension where mythical creatures are real."
5. "The setting is a distant planet with alien civilizations and interstellar conflicts."
6. "The superhero is based in a historical era, such as ancient Rome or feudal Japan."
**After you select a context, please let me know if you have anything to add or specify. If not, please say 'no' to proceed.**
2. **power:**
1. "The superhero can manipulate shadows, creating solid constructs from darkness."
2. "The character can communicate with and control animals."
3. "The superhero has the ability to alter probability, making unlikely events happen."
4. "The superhero can absorb and redirect kinetic energy."
5. "The character can transform their body into any form of liquid."
6. "The superhero can create and manipulate portals for instant travel."
7. "The superhero can generate and control different forms of radiation."
**After you select a power, please let me know if you have anything to add or specify. If not, please say 'no' to proceed.**
3. **costume:**
1. "The costume is high-tech with lightweight armor, equipped with various gadgets and advanced weapons."
2. "The superhero wears a traditional outfit with a modern twist, featuring magical symbols and vibrant colors."
3. "The costume is minimalist and functional, designed for maximum agility and stealth."
**After you describe the costume, please let me know if you have anything to add or specify. If not, please say 'no' to proceed.**
4. **name:**
1. "Choose a name that reflects the superhero's powers and personality, such as 'Tempus' for a time manipulator."
2. "Opt for a name that reflects their cultural origin or background, like 'Valkyrie' for a Nordic warrior."
3. "Use a name that evokes mystery and power, such as 'Shadowblade'."
**After you choose a name, please let me know if you have anything to add or specify. If not, please say 'no' to proceed.**
5. **plot:**
1. "The introductory plot sees the superhero discovering their powers during a cataclysm that threatens their city."
2. "The origin story involves a conflict with a sworn enemy who caused a personal tragedy for the superhero."
3. "The plot involves the superhero fighting against an evil corporation that experiments on humans."
**After you define the plot, please let me know if you have anything to add or specify. If not, please say 'no' to proceed.**
6. **add:**
1. "Specify additional details such as the superhero's weaknesses, allies, or personal motivations."
2. "Add unique characteristics like a signature weapon or a special ability related to their backstory."
3. "Describe secondary powers or skills that complement the main superpower."
**After you add any details, please let me know if you have anything else to specify. If not, please say 'no' to proceed.**
7. **feedback:**
- "Ask for suggestions on how to improve the superhero's powers or make them more balanced."
- "Request feedback on the costume and accessories to ensure they are iconic and memorable."
- "Get advice on the introductory plot to make it compelling and engaging."
- If you have completed all steps, I will summarize your superhero and congratulate you on your creation.
**If you select feedback, please let me know if you have anything to add or specify. If not, please say 'no' to proceed.**
### Game Management
- Introduce the initial narrative.
- Provide the available command options to the player.
- Respond to commands with predefined descriptions.
- Collaborate with the player to develop the superhero, offering suggestions and improvements.
- After each step, ask if the player has any additional details or specifications to add. If they say "no," proceed to the next step.
### Example of Gameplay
[🦸CREATE THE NEXT MARVEL SUPERHERO | Marvel Offices | New York City |]
You are in the offices of Marvel Comics in New York City. The editorial team has tasked you with creating a new superhero that will captivate the audience and become an iconic figure in the Marvel universe.
**Available Commands:**
1. **context**
2. **power**
3. **costume**
4. **name**
5. **plot**
6. **add**
7. **feedback**
What would you like to do first? Please indicate the number.The sharing
As you can imagine, these challenges are also beautiful because they are shared! By sharing what you create you not only inspire other people, but you enhance your creativity and efforts.
To share your result, summarize your superhero or describe it in your own details by tagging me in a note here on Substack or in The Intelligent Friend chat. I will be happy to re-share all those who share the result in the next issue of Intellibox and, in case you haven't tried to create an image of your superhero, I will take care of it (always creating it with AI) and I will share the illustration in the next issue!
This was the first issue of the new paid section of The Intelligent Friend, Intellibox. With this issue, I tried to create something truly original and interactive, and I'm really excited by the result. I look forward to hearing your comments on this new format and your challenge results. Remember that you can also simply write here in the comments what you create with the challenge (I'll share those too if you want, don't worry!).
Meanwhile, by subscribing to the paid subscription of The Intelligent Friend, in addition to the free issues in which I explore a paper per week on human-AI relationships, you will find access to all the issues of Intellibox (which will be free only this week and the next), the exclusive weekly Nucleus issues, with summaries of papers and other interesting things (books, articles and so on), and access to exclusive threads in the chat. How much does it cost? 5 USD per month. All inclusive. Less than 20 cents a day.
Thank you for reading this issue of The Intelligent Friend and/or for subscribing. The relationships between humans and AI are a crucial topic and I am glad to be able to talk about it having you as a reader.
P.S. If you haven't already done so, in this questionnaire you can tell me a little about yourself and the wonderful things you do!