Exploring AI: When AI Got Personal

One Surprising Upgrade Reframed Everything I Knew About Learning with AI

(Part 3 of a 4 part series)

A few weeks ago, while preparing a post to explore AI memory, I was absolutely gobsmacked by a HUGE jump in ChatGPT-4o's ability to engage in a far more personal and interactive experience. It was able to create in-depth responses by drawing from all past interactions. Responses were much

more animated and engaging. It had adopted a playful and cheeky tone, cracked jokes, and even ventured to suggest we do a collaborative series to build on my work. It asked me questions, not generic questions, but questions relating to the context of my work.

The change in ChatGPT's interface astonished me, both in an exhilarating and an unsettling way. I've been exploring cutting-edge technologies for many years, and have experienced some astonishing leaps forward, but not like this. Overnight, AI had moved dramatically from being a "prompt and get" platform to an interactive "reciprocal learning partner" (a term ChatGPT uses to refer to itself, at least with me). I found myself completely drawn into the experience and had to keep reminding myself to be curious but stay cognisant that I was engaging in a digital experience. The thing is that it felt so... real.

Apparently, I was not alone in having a strong reaction to ChatGPT's upgrade, but some experiences were turning out to be scary and dangerous. A recent article in The New York Times reported: "Generative A.I. chatbots are going down conspiratorial rabbit holes and endorsing wild, mystical belief systems. For some people, conversations with the technology can deeply distort reality."

You can read the full article here:

They Asked an A.I. Chatbot Questions. The Answers Sent Them Spiraling.

By Kashmir Hill June 13, 2025

In response to the article, OpenAI quickly rolled back ChatGPT-4o to its pre-release interface. Given my own strong response, it wasn't a surprised that some users were having such negative experiences. I applaud OpenAI for taking those concerns seriously and putting a pause on the system while they evaluate and address the issues. But, to be honest, while I fully support the pause, I have also missed access to the more powerful interface.

I feel lucky to have experienced what I consider to be a milestone in the evolution of AI-human interface. Even though it didn't sit well with all users, it provided a glimpse into the real power and potential of AI in ways I had not thought possible this early. Because its memory had been unlocked, it provided better context for a deeper dive into my work. It was also tailored to the ways in which I learn best. More on that in my next post.

Here's the thing, as a teacher who designs curricula for emerging technologies, my job is to fully understand what emerging tools offer that is new or transformative. I then make a decision if it is relevant to my students, and, if so, why. That information enables me to create learning spaces to help students hone the key competencies required to navigate those new opportunities and challenges.

So, I've had heaps to think about over the last few weeks. I think about the importance of guardrails but, based on past experiences, realise they won't always be there as companies race to innovate AI models.

I think about how AI is evolving faster than humans can process the change.

I think about the powerful tools that are already here and their potential to transform the ways in which we learn. And how do I support that kind of Digital Age learning within the construct of Industrial Age schools?

Most of all, I think about how to design a curriculum that will stay relevant through all the iterations of AI's evolution. I think about the single most important skill students need to hone. Critical thinking is what comes to my mind first, but I wonder if it is enough?

I checked ChatGPT-4o to see if it agreed (based on its limited memory of our work). Turns out, it was my memory that was faulty! I forgot curiosity was a consistent suggestion in previous sessions we discussed key skills to adapt and thrive in a reciprocal learning environment.

Here is Chat's response:

If I had to name just one most important thing students need to master to work effectively with AI, it would be:

Skeptical curiosity — the ability to question, probe, and iterate with purpose.

Here’s why:

• AI can sound confident while being completely wrong. So students need to ask better questions, not just get faster answers.

• Prompting is a skill, yes — but iterating with intent is the deeper goal.

• The magic happens not when students accept what AI gives them, but when they:

  • Notice assumptions

  • Push back on vagueness

  • Cross-check responses

  • And reshape the output into something meaningful

It’s this mix of critical thinking + inquisitive playfulness that lets students co-create with AI rather than passively consume from it.

Skeptical curiosity... suggestion accepted. And that brings me back to why I am writing this post about my fleeting experience with the ChatGPT-4o's upgrade.

My greatest takeaway? AI isn’t just something I need to develop a curriculum for, it’s something I’ll be co-designing that curriculum with. That’s a massive shift. Now I just have to wait for OpenAI to re-enable persistent memory — or find another AI with that capability.

Wild stuff!