DALL·E 2023-01-22 21.53.17 - an impressionist painting of the complexity of the internet

AI for Personal Growth and Self-Reflection

I’ve been poking around with ChatGPT to analyze my rather extensive collection of blog posts and journal entries; a quite manual and tedious process but ChatGPT with Web Browsing helps speed up the process quite a bit. And the insights have been well worth the effort, and quite profound (will talk more about it below).

Seeing this work so well has gotten me thinking about how AI aids could be incorporated more meaningfully into my self-reflection/action loop, and by proxy, my life. Often, I’ve found myself trying to prompt ChatGPT to try to ask me questions; I’ve been partially successful, but it needs me to do considerable prep in my own prompt to essentially, give the baton to ChatGPT to “prompt me”.

So when I saw this Tweet, I was curious to know where the thread would go:

Following the tweet chain, I discovered Mindsera and just started using it to see how it works. It’s too early to give any real learnings, but I will credit them for a slick, stable, usable, distraction-free interface for journaling. Overall, I love what they’re trying to do and hope it can provide real value for people.

For me, I have some reservations about using an app for such a personal purpose that specifically is trying to make money from what I do for fun, as well as probably doing some form of QA on my content and output. That said, if it encourages me to write more or be more reflective, then even a single entry could be worth the monthly cost to me, though the privacy aspect is why I’ve been so interested in hosting my own LLM’s locally (a whole other entry to come!).

I’m using the free version to test out what it can do; so far, it seems to take in portions of my post as a prompt and generates a question based on it. More features come with a paid version. I’ll be comparing it with my use of ChatGPT for the same purpose. Some of the “questions” seem pretty generic and canned, and I’m curious to know the distribution of what it has internally. I can see it working well for many people, and of course, the free version is ‘free enough’ for most people.

https://www.mindsera.com/#pricing

Anyway, let’s see what it can do!

(btw when writing journal entries, I tend to not do a lot of polish; typos, grammatical errors, and half-baked thoughts are probably all over the post–for a clean interpretation, just read what ChatGPT thought).

Mindsera Entry 1

Hello new journal!

I’m going to try testing Mindsera for my personal use. In my own personal use of ChatGPT w/GPT-4, I’ve been finding the rise of Large Language Models (LLMs) and their ability to analyze and synthesize insights quite powerful, and rather than treating GPT-4 like an ‘enhanced search’ (where I ask a question to get a set of answers out there), I’ve been trying more in-depth analysis of content and actually having a conversation about it (with ChatGPT). Instead, the most powerful results have come when, at least to my perception, ChatGPT actually appears to understand the details of what I’m saying.

Mindsera Question 1: How do you determine whether ChatGPT’s apparent understanding of what you’re saying is a true comprehension of your meaning or just a clever approximation based on pattern recognition?

Well, to be honest, I can’t know for sure–but I have two big thoughts here:

While, it’s true that I can’t really know for 100% sure if ChatGPT is actually sharing an ‘understanding’, I actually would argue that that’s also true for any human we interact with. We use language to communicate with each other and can exchange words and other communication, but at some point, we can only hope and have faith that the other person actually gets it.

More important though, is the actual experience interacting with it. I’ve basically fed it single posts at a time, or sometimes several in sequence, to try to provide it sufficient data to work with.

Unfortunately, ChatGPT, even with GPT-4 is quite limited in maximum prompt length. I’ve heard some anecdotes that ChatGPT can read “hundreds of pages” to synthesize a summary, but in my own use of it, it seems to fail at integrating more than a few thousand words. GPT-4 With Web Browsing does seem to be able to incorporate a chain of webpage-sized inputs together and provide output that more holistically interprets the writing (and in turn, me), but even then, it seemed quite limited in what it actually uses to generate the final output (~4 pages).

How I’ve used it… now this is the interesting part. to see if it can generate summaries, offer insights into my own thoughts, offer perspectives on how others see me, and even offer suggestions. It does all this amazingly well, and as my journals are relatively private, I don’t think I would have been able to get this kind of benefit anywhere. Well, at least, until recently. Apps like MindSera clearly are recognizing this opportunity with LLMs to build a thoughtful journaling partner. I’ll say… so far, it’s been quite good! Look at all that I’ve written already!

And for fun, I used my standard prompt for single-entry interpretation to see what ChatGPT thought about my post:

ChatGPT (GPT4)’s take on my post:

Prompt: Can you summarize what I’m saying, how I present how I’m feeling, and any insights into my thoughts about my own life? How do you think I or my post might be perceived generally by someone reading it?

(Above Mindsera post text)


Response:

Journal Entry Analysis:

Summary: The author is experimenting with Mindsera, an AI-based journaling application, after previously using ChatGPT with the GPT-4 model. The author acknowledges the power of Large Language Models (LLMs) to analyze and synthesize information, noting the most impactful results when ChatGPT appears to truly understand the context. They express a degree of skepticism about the LLM’s understanding, suggesting it could merely be a clever approximation based on pattern recognition. However, they make a parallel between this and human interactions, stating that understanding is largely based on trust. The author also mentions the limitations of GPT-4 in integrating and synthesizing large amounts of data. Their main usage of the tool has been to generate summaries, offer insights into their thoughts, provide perspectives on how they might be perceived, and suggest possible actions. The author appreciates these capabilities and the privacy of their journals.

Presentation of feelings: The author appears to be intrigued and positive about their interactions with AI models. They show curiosity and excitement about exploring new applications like Mindsera. There’s also a sense of skepticism and practicality regarding the capabilities of AI, but these feelings do not diminish the overall enthusiasm.

Insights about life: The author values self-reflection and introspection, as they use AI to analyze and interpret their thoughts. They seem open-minded, valuing the insight and perspectives that AI provides, indicating a willingness to engage with new technologies for personal growth. The value they place on their journals being private might suggest that they find writing to be a personal, solitary activity.

Perception of the post: The reader may perceive the author as technologically literate, self-reflective, and open to new experiences. They would see someone who values their privacy but is also curious about the possibilities of AI in personal growth and introspection. The author’s thoughtful approach to using AI tools may make them appear explorative and analytical. There’s a sense of excitement and positivity in the post that makes the author seem optimistic and forward-thinking.

Potential reactions: The reader may feel intrigued by the author’s exploration of AI for personal growth and might be inspired to investigate these tools themselves. They might appreciate the author’s balance of optimism and skepticism towards AI’s capabilities. The author’s inquisitive and analytical approach might also lead to a feeling of respect for their thoughtful engagement with technology.

Not bad, ChatGPT! I’ll continue to use it and MindSera as I found the initial phases of writing the hardest, and the quick prompts from Mindsera were actually great at taking little seeds of ideas and encouraging me to expound them into something worth writing.

But I have found that as my posts get longer (and perhaps less coherent), MindSera seems to struggle to give any meaningful prompt that incorporates the entire post. Perhaps a paid feature?

A reflection on the moment (forming thoughts on LLMs and GenAI)

(Note: This post is a test of ChatGPT’s ability to complete a blog post with perhaps only 20% started–I’d say it kept to the spirit of what I wanted to say, though it lost a bit of the original voice)

The artificial intelligence domain has been a hotbed of innovation lately, especially with developments surrounding large language models (LLMs). With regards to (LLMs) it’s hard not to feel like we’re at a pivotal moment simultaneously in human-computer natural language interfaces understanding, epistemological foundations, neuroscience, and computer science, and the entire potential application space with the convergence of all these technologies.

And it is absolutely clear that these are such early days that even months and weeks are showing vastly new possibilities. Google’s Transformer architecture and OpenAI’s pioneering work in LLMs have unleashed something special, but the ingredients of this explosion have been brewing long before this moment and they will be far from the only players in this space. Still, this is as watershed a moment as there ever has been in artificial intelligence and the relationship we have with our technology.

The pace of progress is breathtaking. What was new last month seems almost outdated today. While OpenAI has been a major catalyst in this explosive growth, the current advances are built on a foundation laid down over many years by countless researchers and developers.

But this isn’t just another technological advancement. We are potentially reshaping the way we interact with technology, opening doors to possibilities that extend far beyond mere tool utility. Let’s consider some of the potential real-world implications of this evolution:

  • Education could see a radical transformation with personalized learning becoming the norm rather than the exception. With LLMs, students can have access to tailored learning experiences, helping them grasp complex concepts with ease and at their own pace.
  • Healthcare stands to gain immensely. LLMs could analyze patient data to suggest treatment options, and even converse with patients to provide comfort during trying times. The implications for improving patient outcomes and the quality of care are enormous.
  • In the world of art and literature, AI has already started to make an impact. But this is just the tip of the iceberg. Imagine AI not merely as tools, but as creative partners, offering fresh perspectives and pushing our creative boundaries.

These groundbreaking innovations in artificial intelligence are not merely changing our external world; they are also fundamentally altering how we perceive and understand ourselves. They raise intriguing questions about human cognition, creativity, and even consciousness. What does it mean to think, to create, to be aware? Are these uniquely human traits, or can they be replicated, even surpassed, by our own creations?

In the face of such profound questions and the promise of even more astonishing breakthroughs, the future is a vast canvas filled with potential. As we continue to traverse this exciting path, it is not only the technological advancements that thrill me but also the opportunity to continually redefine our understanding of ourselves and our place in the cosmos. In every sense, we’re not just shaping the future of AI — we are, indeed, reimagining our very existence.

XXXVIII – Day 13879

Haha, I’m kind of scared that ChatGPT is going to take away my will to write, given how well it seems to do. But nay that’s how we cede ground in the human-vs-AI wars no? Must try harder!

So here am, at the turn of 38, back here to blog about it! That it’s been almost two decades since my turning 19 post is kind of a shock, but seeing how much I had changed between then and 20 I do feel that since then, growth in my (writing) voice has come a little slower. I do wonder occasionally if I’m ‘supposed’ to be writing in far more elegant prose by now, but I think the chief reason is that I’ve not been writing for fun nearly as much since short-form social media came to dominate our online discourse.

(Are folks closer to 40 supposed to act like they’re suddenly old people? Dunno, but I’m gonna keep doing me, so 😛 ya)

Anyway, wanted to write a real post, something I could reflect on later, and ask ChatGPT to incorporate into my ongoing meta-analysis of my blog.

Onwards to 38!