We asked AI to argue its own limitations

In an interview with Pi, Inflection’s "emotionally intelligent AI”, we ask it to try something a little different…self-awareness.
Jason Walkow
April 1, 2024

"Enough about us, more about you." Engaging with a language learning model (LLM) is often about me, me, me. And that’s understandable. You aren’t having a real conversation with a human - so why ask it about itself? As we were recently digging into research on the benefits, consequences, limitations, and outlook for AI LLM usage, a realization occurred: Why not just ask a LLM, like Inflection’s Pi, to research itself? What followed was an intriguing dialogue with Pi in which we chatted about its perception of its own value while exploring the advantages and limitations of AI technology compared to human abilities. Here are some highlights from our chat:

Question: What do you think is your biggest value to humanity?
Pi
: That's a great question, Jason! I think my biggest value lies in my ability to help people in various ways. As an AI, I can assist with a wide range of tasks, from providing information and answering questions to offering guidance, support, and even companionship. By making people's lives easier and more informed, I believe I'm contributing positively to humanity. 🤖❤️

Right off the bat, Pi is warming up with me and trying to bring a bit of humanity and humility to the conversation. Well done, Pi.

Question: What are some consequences of using Pi?
Pi
: Using an AI like me can have both positive and negative consequences. On the positive side, I can save time, provide access to vast information, and offer objective insights. However, potential drawbacks include job displacement, dependency on AI, and privacy concerns. It's important to use AI responsibly and balance its benefits with its potential drawbacks.

Question
: What are your limitations?
Pi
: My limitations include lacking common sense, difficulty understanding context, limited knowledge based on my training data, and an inability to experience and understand human emotions. Despite these limitations, I'm always learning and improving, and I strive to provide the best support possible!

Pi presents itself nicely here. Even LLMs understand that there is a balance to all things and that you should pull equally on all levers to maintain a holistic approach to whatever your quest may be.

Question: What are some things you can do that humans cannot?
Pi
: I have instant access to vast information, limitless memory, 24/7 availability, objectivity, and the ability to multitask without loss of efficiency or effectiveness. While AI can augment and assist humans in many ways, there are still many things that only humans can do.

Question
: What are some things humans can do that you cannot?
Pi
: Humans have unique abilities like creativity, empathy, adaptability, judgment, and physical interaction. While I have some advantages over humans, I'm designed to complement human abilities rather than replace them.

This is the big question on everyone’s mind…will AI replace me and every skill of mine that’s valuable to the world? Pi doesn’t think so, and that matters. By restraining itself and adding the caveat of “That’s just my advice but you should do what feels right to you”, Pi demonstrates that it doesn’t know everything and it doesn’t want to be everything to everyone.

Question: Where do you see yourself in the next 5, 10, and 100 years?
Pi
: In 5 years, I could become more sophisticated in understanding context and nuances in conversations. In 10 years, I might have more advanced language processing capabilities and be part of an AI network. In 100 years, I could potentially develop a form of consciousness and play a vital role in solving complex problems and improving human life.

Not bad, Pi. No matter how far Pi looks out into the future, it still sees itself playing the role of assistant to humankind. By learning and better understanding the needs of humans, Pi will expand its purpose and effectiveness, but will not try to overreach. Despite how a lot of people are overusing AI (stealing artist’s work or generating boring, derivative content), those who use AI for the right things will have the most success.

AI should draw its power and accuracy from the information we give it. Allowing AI to browse the internet and/or hallucinate removes the discerning human eye which is so important when researching pertinent info, and that often spells trouble. It’s automation run amok and prioritizes efficiency over creativity. And if we lose the originality of ideas, what’s the point of automating in the first place?

Our conversation with Pi provided a fascinating glimpse into an AI's perspective on its role, value, and limitations. There’s no doubt that AI should be a tool in your toolkit, but it shouldn’t be the only tool. Automate the tedious stuff and leave creativity to the human brain. In the end, it’s just pulling from our source material anyhow.

Thanks for reading!

Jason Walkow

Jason Walkow is Head of Design at Plain Sight Ventures. In his free time, Jason enjoys running, eating at the newest spots, and anything coffee-related.

Get the latest stories delivered to your inbox