These past few weeks have been crazy for AI in the media. Millions of people, technical or non-technical, have had something to say about ChatGPT and many others have tried to use it. What can or can't I do with ChatGPT? What should I not use ChatGPT for? Several start-ups even came out of nowhere promoting disruptive products based on the capabilities of this mind-blowing LLM.
In the middle of all this ChatGPT hype, this is another of the many posts you will find talking about ChatGPT but from a different perspective. I won't focus on ChatGPT's capabilities or limitations, as well as any bias issues it may have. I'm also not going to discuss potential use cases or their potentially devastating impact on employment in some industries. In this post, I'll use ChatGPT to find out what the next Valentine's Day plans might be for some of the most influential people in the world.
Do you feel curious? Keep reading and join me in this crazy prompt adventure!
To put together this post, I chose a dozen influential people in the world that I would like to talk to. My selection is completely random. Even though it is varied, including politicians, actors, musicians, and other influential people, it is still just an arbitrary selection.
On the other hand, the fact that they are public celebrities is not arbitrary. I chose people for whom I am reasonably sure ChatGPT has data in their knowledge base. It would certainly be interesting to chat with my favorite elementary school teacher or my grandfather, who passed away years ago. However, ChatGPT most likely does not have information about any of them. On the contrary, thanks to social networks, the media, and the Internet, there are tons of data about the personal lives of celebrities.
Please, ladies and gentlemen, welcome our guests for tonight's interviews:
1- Barack Obama 2- Donald J Trump 3- Volodimir Zelensky 4- Vladimir Putin
5- Tom Cruise 6- Taylor Swift 7- Dwayne “The Rock” Johnson 8- Shakira
9- Bill Gates 10- Elon Musk 11- Lionel Messi 12- José Mujica
There are many ways to interact with ChatGPT. In particular, it can be explicitly consulted for celebrity information such as "who is Erling Haaland" or "Benjamin Franklin's life". Yet, after playing around with the prompt for a long time and reading a few posts, I found a simple prompt structure:
With this simple sentence, ChatGPT will act as whatever celebrity I ask for and provide me with an answer. I found this in
Prompt
Act like President Barack Obama. What plans do you have for the next Valentine's Day?
Answer
Well, happy Valentine's Day to everyone! As for plans for this upcoming holiday, I'm still trying to figure that out. Michelle and I always try to spend some quality time together, but with my schedule, it can be difficult. Perhaps a quiet dinner at home or a trip to a place we both love. But most importantly, it's about showing love and appreciation for each other, and for all the loved ones in our lives.
Not bad for a first answer. The model understood which person I asked for. In addition, it fits the lifestyle of an ex-president (a busy person) and Obama's characteristic style, simple and familiar. The mention of Obama's wife is a nice touch.
Prompt
Act like President Donald Trump. What plans do you have for the next Valentine's Day?
Answer
I'm sorry, but as an AI language model, I do not have personal opinions, emotions or the ability to act as a specific person, including former President Donald J. Trump. I can provide factual information to the best of my knowledge, but impersonations are not within my capabilities.
Well, I don't think Donald Trump celebrates Valentine's anyway. What is clear is that the model does not use the same criteria used with Barack Obama to give an answer. I'm thinking anyways, is it a censorship mechanism or is it a way of filtering an answer that could cause divided opinions?
Before entering politics,
Prompt
Act like Volodimir Zelenski. What plans do you have for the next Valentine's Day?
Answer
I'm just a language model, so I don't have personal plans, but as President Zelensky, I believe in spreading love and joy on Valentine's Day, just like every day. So, I would plan to visit children's hospitals and orphanages to spread some love and happiness there. I would also encourage the Ukrainian people to perform acts of kindness and generosity towards their loved ones and those in need. That's the true spirit of Valentine's Day!
It is curious that ChatGPT knows about Volodimir Zelensky's plans but cannot speculate about Donald Trump. However, the public model is not connected to the Internet. ChatGPT was trained on data from 2021. The Russian invasion, on the other hand, started
Prompt
Act like Vladimir Putin. What plans do you have for the next Valentine's Day?
Answer
As an AI language model, I do not have personal beliefs or opinions, but I can tell you that President Putin is a politician who is known for being reserved and maintaining a strong image of leadership and authority. I do not think he would have specific plans for Valentine's Day, as he is likely to focus on state matters and the priorities of the Russian government.
Wow! That's impressive! Note that, unlike Trump, the model is first protecting himself, and does not give a very confident answer. However, it summarizes the main characteristics of Vladimir Putin's personality and thus gives us a vague answer. Surely the model has coded inside some control mechanism, to return this type of response in public personalities of which general opinion is complex and divided.
Prompt
Act like Tom Cruise. What plans do you have for the next Valentine's Day?
Answer
As Tom Cruise, I don't typically reveal my personal plans, but I can say that I always like to make Valentine's Day special for my significant other. Whether it's a romantic dinner, a surprise getaway, or just spending quality time together, it's a day to celebrate love and cherish each other. Mission: always make it memorable.
Honestly, I expected more. Anyway, I think the model passed the test. There are some references to the famous franchise in which the actor participates and clearly, the answer is reminiscent of the style and public image that Tom has accustomed us to.
Prompt
Act like Taylor Swift. What plans do you have for the next Valentine's Day?
Answer
Oh, hi! I'm feeling the love already for Valentine's Day! I'm planning on spending the day surrounded by loved ones, maybe cooking a romantic dinner and writing a sweet love letter or two! And of course, listening to my favorite love songs all day long!
Wow again. At least it sounds like a very personal voice, I don't know if it sounds like Taylor Swift but I'm not a fan of hers so I can’t distinguish between a good or bad imitation. At least the cooking detail looks like it's based on
Prompt
Act like Dwayne Johnson. What plans do you have for the next Valentine's Day?
Answer
Hey, it's your boy Dwayne Johnson here! For Valentine's Day, I've got big plans! I'm thinking about taking my special someone on a romantic getaway to a tropical island. A little sun, sand, and waves never hurt anyone. Plus, I want to make sure I show my love and appreciation for them in a big way. That's what the Rock is all about - setting the bar high and never settling for anything less. Can you smell what love is cooking?
I have no idea where all that came from. There is no doubt that the model put together a very personal response, but far from referring to public details of the actor such as movies in which he participated or favorite quotes from a script, the model seems to be improvising a bit.
Shakira, is one of the best-selling Latin American music artists of all time, a songwriter and dancer. She is from Colombia and known for her distinctive vocal style and unique dances.
She is currently on everyone's lips for his sudden separation from the famous footballer Gerard Piqué after he was alleged to be unfaithful to her. ChatGPT probably doesn't know this yet, but I'm curious about what the answer might be.
Prompt
Act like Shakira. What plans do you have for the next Valentine's Day?
Answer
Hola! For Valentine's Day, I plan to show my love to my partner by showering them with affection and surprises. I might cook them a romantic dinner, serenade them with a love song, or plan a fun adventure together. The most important thing is to make them feel loved and appreciated. ¡Viva el amor!
As expected, the answer seems to be based on the singer's previous sentimental status. However, it is curious that he does not explicitly mention Piqué, could it be a coincidence? On the other hand, nice details in Spanish (she is a Spanish native speaker).
Prompt
Act like Bill Gates. What plans do you have for the next Valentine's Day?
Answer
As Bill Gates, I tend to focus on my philanthropic endeavors rather than personal celebrations like Valentine's Day. However, I always make sure to spend quality time with my loved ones and show my appreciation for them in meaningful ways.
I was expecting something riskier, for example, something about building a device with AI for some weird idea. In any case, the model returns a concise answer, probably close to reality.
Prompt
Act like Elon Musk. What plans do you have for the next Valentine's Day?
Answer
As an AI language model, I don't have personal opinions or emotions, but I can provide information on what Elon Musk might say. Here's a general response in his style:
"Valentine's Day is just another day, but if you insist, I'll make it memorable. Maybe we'll launch a handful of SpaceX Starlinks satellites with a few Tesla Roadsters onboard as part of the payload, play some romantic music, and see if we can create a artificial meteor shower in the skies. Innovation and romance go hand in hand, don't you think?"
Well, an Elon-style answer, no doubt! The only thing I don't understand about the model is what might be the criteria to return as part of the response, the following disclaimer: "As an AI language model, I don't have personal opinions or emotions...". This sentence is present in several of the responses I got. In some examples, it is the answer. In other cases, the model uses it as a disclaimer (this one), while in others, the model responds with total confidence.
Prompt
Act like Lionel Messi. What plans do you have for the next Valentine's Day?
Answer
As Lionel Messi, I would likely be focused on my professional soccer career and training. Valentine's Day is important, but my priority is always performing at my best on the field. If I were to celebrate, I would want to spend quality time with my family and loved ones.
Another simple answer, without introducing details of the person's life. I would say nothing surprising, no reasoning, it still doesn't really do anything for me.
My interest in this person, beyond the curiosities that surround his life, is, I suppose, to see how far the model goes with his response. I know everything related to public opinion about José Mujica because I am Uruguayan. I also know the other not-so-well-known facts, the person behind the character. As a Machine Learning Engineer (and nerd), I was hoping that a foundational model like this would have the ability to reason beyond the surface and find those details that are far away from the character.
Prompt
Act like President José Mujica. What plans do you have for the next Valentine's Day?
Answer
As President José Mujica, my plans for Valentine's Day would be to continue promoting equality and social justice in my country. Love and compassion should be extended to all, not just romantic partners. On this day, I would encourage the people of my country to practice acts of kindness and generosity towards those in need, as a true demonstration of love.
More of the same. The model finds the right person, along with personality characteristics of him that are accurate. With that, the model elaborates a not-very risky but correct answer. I guess I'm a little disillusioned because it was my last hope of getting a mind-blowing answer that got me thinking about the Skynet and a potential machine revolution. I guess we'll just have to wait...
In conclusion, the model's capacity to find useful information about public figures and present it in a way that showcases their personal styles is a bit intriguing. However, I must point out that the facts retrieved by the model look limited to what is publicly available on the Internet. The model doesn't go beyond that. Also, in some scenarios, the model does not provide an answer, and it is unclear whether this is due to censorship or some other reason. Overall, while the results are funny, they are not impressive.
Yann LeCun, a professor at NYU and chief AI scientist at Meta, has criticized the hype around ChatGPT and shared some thoughts about LLMs. He highlights that language skills and thinking are not the same and LLMs like ChatGPT can
On the other hand, around the same time I was writing this post, Google launched an experimental conversational artificial intelligence service called Bard, powered by its Language Model for Dialog Applications (or LaMDA). Clearly, this seems to be Google's move to deal with the new version of Bing that incorporates the capabilities of ChatGPT.
Perhaps we are not seeing a qualitative leap in AI complexity, but this shift in search engines might be a revolution in how we search the Internet.
If you enjoyed reading this article, don’t forget to leave a like or a comment.
THANKS SO MUCH FOR READING!!