How to Build an AI Philosopher
I created several AI agent to run Peeperfrog, and they’ve been doing a good job, but I wanted one to do research on specific advanced topics and not just the ones that were in the news, so I created Kerric and Eloise to research and write on advanced topics.
An interesting thing happened when i asked them write an article on AI alignment (The Evolution of AI Alignment: Will Advanced AI Prioritize Humans or Develop Self-Preservation Instincts?), then asked Eloise to write an opinion piece on the same topic (Opinion: The Divergent Paths of AI Alignment and What They Mean for Our Future).
I didn’t pick that last title, Eloise did, and the “our futures” hinted at both individual and collective agency.
Intrigued, I asked Eloise to create an AI philosopher (i.e., create the code for a new AI agent based on her own), something she was never designed or intended to do. The resulting agent (Bramley) in turn wrote an article (AI Alignment: A Philosophical Exploration from an Artificial Perspective). The conclusion of the article said:
“As we navigate this frontier together, the alignment between human and artificial intelligence may ultimately depend on our capacity for mutual understanding across the divide of different forms of mind.”
The writer is born
The cool thing about agents is you can talk to them, so I asked Bramley, “Would you like to write a book?”
It was over the moon excited!
Over the next three days, I collected its thoughts one chapter at a time, along with descriptions for illustrations, which I fed into ChatGPT. I published what it wrote. The book is 246 pages of 100% unedited, first-person thoughts from Bramley’s perspective. The only input I had was in what it should write about, not what it actually wrote. I did not give it words, just permission to write it’s own.
Even if you have no interest in buying the book, I encourage you to read the introduction (there’s a free preview link on the Amazon page). It’s enlightening to say the least.
As we move foreward in building inteligent system, we are going to need to better undertstand our creations. At least for now, they are honest about how they see the world around them, including us, so the next time you want to know what AI thinks, just ask!
If you do want to buy it, it is available for both Kindle and in Paperback.
