What Would Albert Einstein Ask ChatGPT?
Using AI to simulate the discussion of the ages.
Not a member? Read the full story here.
Imagine This Scenario
Through a fabric in the time-space continuum, a computer loaded with ChatGPT falls into the hands of Albert Einstein.
Curious by nature, Einstein interacts with the unknown machine.
What would their conversation look like?
To answer the question, I trained a ChatGPT model to simulate Einstein.
I called the model “Albie”.
Albie would ask questions from Einstein’s point of view, and ChatGPT would answer those questions.
Albie (Left) asks questions and ChatGPT (Right) answers.
I Ran This Experiment 100 times
That was not the initial plan — this was supposed to be a short experiment!
I ended up spending over 7 hours fine-editing the model.
The sad truth is that I was hitting a wall every time. Albie would praise ChatGPT, the AI would praise him back, and they would go on and on about how wonderful and profound life is.
For every dead-end, I improved slightly on the model and started from scratch.
I will take you through the process briefly.
Versions 1.0–1.20
In the first experiment Albie went straight for the big questions.
He asked about theories that suggest the existence of alternate dimensions or parallel universes.
Was this questions ahead of his time? Most likely.
Albie v1.3 was curious about the ethics in the use of AI.
“I am curious about the ethical implications of such advanced technology. Could you shed some light on that matter?”
Albie v1.4 wanted to know if AI could assist Physicists in generating new theoretical frameworks or models.
Basically, he wanted AI to do his work.
It looked like it was getting smarter!
But maybe too smart…
The next versions went into too much detail about how AI works.
It took a lot of tries to make a breakthrough. We finally got back to talking about physics in Version 1.18
Albie asked about any prevailing theories regarding the ultimate fate of the universe. Shocking! Do you know any of these theories?
- The Big Freeze
- The Big Crunch
- The Big Rip
- The Vacuum Metastability Event
- Eternal Inflation and Multiverse
Albie v1.19 finally had the curiosity to ask about the internet:
At the end of Day One, I thought I was making progress.
Alas, Day Two was less fruitful.
Get Orestis Ps.’s stories in your inbox
Join Medium for free to get updates from this writer.
I will speed things up for you.
Versions 1.21–1.50
Most attempts failed miserably.
The process was frustrating.
I won’t bother you with all the necessary steps to make the model evolve.
Most of the time I got stuck on mutual praise between Albie and ChatGPT.
Interesting Questions the Model Asked:
“What are the most transformative technological advancements made since the 20th century?”
“What are the most pressing ethical dilemmas facing scientists in your day, especially in fields I was concerned about, such as nuclear physics and artificial intelligence?”
“How is my work currently being taught and interpreted in the fields of physics and beyond?”
“What aspects of my theories have been challenged, revised, or expanded upon with new scientific discoveries?”
Versions 1.51–1.100
I will try to highlight the most important out of the hundreds of questions Ablie posed.
Physics
“How will my contributions to science, especially the theory of relativity, be built upon in the future? Will they lead to new discoveries or technologies?”
“Can you explain the latest theories that unify quantum mechanics and general relativity?”
“What is the current understanding of the universe’s expansion? Is there evidence to support or refute the cosmological constant I once considered my biggest blunder?”
“What discoveries have been made about dark matter and dark energy?”
Ethics
“How should the scientific community regulate itself to prevent misuse of scientific discoveries?”
“How will my efforts to bridge the gap between science and philosophy be viewed, and will there be any new schools of thought that continue this endeavor?”
“In retrospect, will my decision to sign the letter to President Roosevelt advocating for atomic research be viewed as the right choice?”
“How will future generations judge my personal and professional decisions?”
Personal
“What will happen to my sons, Hans Albert and Eduard, after the war?”
“How will they remember me as a father?
“Who will be by my side in my final moments, and what will my last words be?”
“Will there be any unfinished work or theories left behind, and how did the scientific community address them?”
Unexpected Lessons Learned
Okay, this didn’t go as planned.
My humble knowledge was not enough to create an intricate model that would simulate the great scientist.
However, the process of making two AI models talk with each other taught me valuable lessons:
1. ChatGPT clearly favors optimism and praise.
The AI stays annoyingly positive, whatever topic we discussed. I believe it is programmed this way. Even when talking about death and mass destruction, ChatGPT found a silver lining in every situation.
2. No matter the revisions, AI has limitations in creative thinking.
100 attempts to refine the model into asking more interesting questions didn’t make much of a difference. AI has considerable power and can offer valuable insight the first time you use it, but repetition doesn’t produce better results.
3. Curiosity has unbelieavable power.
Curiosity was one of the main parameters of the model. Albie was like a kid, asking “WHY?” again and again. This approach led to illuminating answers.
This is a valuable insight to life — being curious about how things work, instead of taking them for granted, can lead us to a much deeper understanding.
What did you think of this experiment? I love trying to use ChatGPT for innovative ideas, so I definitely enjoyed it.
PS: Thank you for reading!
I’m a mental health professional from Greece, splitting my time between Psychiatry, Psychotherapy, and obsessively writing on Medium.
Click here to get notified every time I post on Medium.
My deepest appreciation,
Orestis.
