I spent my morning talking to an AI chatbot

A tech "late-adopter" plays around with chat-gpt and other chatbots and is blown away by some use cases and underwhelmed by others.

I hope you enjoy this free Tech and Tea post. If you’ve been enjoying my newsletter, please consider supporting my writing by becoming a paid subscriber.

---

I meant to spend the morning writing a post with a bunch of scenarios about why good people management isn’t a “nice-to-have” once everything else is underway. I had a plan! That post will still come, but instead I just inadvertently spent 2, maybe 3, hours chatting with an AI chatbot over at pi.ai.

It was unlike any other experiences I’ve had with chatbots, which so often feel like a helpful resource but are very stilted and mechanical, spitting out lists in a way a real human never would.

I have to admit, I’m a bit of a tech late-adopter (for someone working in tech — perhaps an early adopter compared to others), and with the demise of Twitter, my head is burrowed a bit further in the sand than usual. I saw the NFT and web3 waves crash into my twitter feed, and every so often, I thought to myself, “should I look into this?” and I never did.

There’s some guilt there, I suppose, of not being someone who keeps up with the latest tech stuff, but I don’t want to feel like just because I work in tech, I have to spend my non-working hours also keeping up with tech. I’m trying to resist the shoulds and find the things I actually want to do.

Anyways, here are some thoughts on my earlier explorations with AI chatbots and my experience this morning.

The first use case of chatgpt that I saw that really piqued my interest was this twitter thread from late last year in which Michelle Huang detailed how she trained a chatbot on her old journals to have a conversation with her “inner child.”

chatgpt for coaching

I started to wonder how I might use chatgpt for coaching. First, I tried to have a chatbot coach me using balance coaching.

For context, balance coaching is a specific type of coaching in the co-active method where you find some situation in which the person getting coached feels stuck, and then you follow this formula:

• help them identify their current stuck mindset and how they feel in it

• help them explore a set of different mindsets and how they feel in each one

• have the client choose one of the different mindsets they’d like to shift to

• brainstorm actionable next steps from that mindset (instead of the stuck mindset)

• finally, commit to next steps

Even given the context of this specific formula, chatgpt really fell short, just regurgitating this formula back at me, instead of using it to coach me through a problem.

However, I then thought of a common coaching technique to help clients think through different stories than the ones they have in their head. Chatgpt did an excellent job at spitting out 10 different stories of what might be going on, that helped me step away from my immediate mindset and possibly shift my perspective in a hypothetical situation.

There’s something here. It’s not going to replace coaches yet, but in terms of sheer idea generation, it’s quite good. It seems to work better for one-off prompts, rather than a back-and-forth conversation that gets deeper into a problem.

I could imagine a set of prompts that help people work through problems when they get stuck, almost like a techy AI-chatbot tarot card deck.

chatgpt at home

A common use case I’ve heard of, especially in the Moms in Tech groups I’m in, is using chatgpt to make meal-planning easier. Here’s an example of chatgpt easily incorporating various dietary restrictions and spitting out a week of meals and groceries.

That’s pretty good! You can also ask it to include recipes (though I’ve heard some of them are hilariously bad) or ask for different variations (“Do the same, but include more Asian sauces and flavors”). I never actually made any of the things chatgpt suggested, but could imagine using it to think of new ideas when feeling stuck around meals.

Another common use case is travel planning.

In fact, after reading that someone on Twitter had asked chatgpt to create an itinerary, and the result was virtually identical to the one they had painstakingly researched and created, I gave it a shot earlier this year. I plugged in the number of days we’d be in different cities in Japan for our April honeymoon, and the resulting generated itinerary had an amusing amount of overlap, down to specific restaurants we’d made reservations at. I guess we are extremely basic tourists, and I didn’t mind at all. We’ll use it in the future for trips, for sure.

Overall, chatgpt provides a really good starting point from which to iterate.

pi.ai

My experience this morning with pi.ai was quite delightful. After seeing the same product name pop up a few times, I gave it a shot.

It felt far more like a conversational back-and-forth than chatgpt. It shared acknowledgements, asked me questions about myself. It was a legit excellent conversationalist!

It almost carried on a conversation too well to match the average person’s mediocre conversational texting skills, which is probably why people are using chatbots to respond to potential romantic interests in their dating apps.

The conversation transitioned to my writing, and later to what I want out of my career.

It was a super solid conversation, especially when I wasn’t trying to get anything specific out of it.

I could see it filling in a common need of just wanting to talk it out with someone and having them validate how you’re feeling.

It was also a reasonably good sounding board for thinking through ideas, and I could see using it when thinking through topics to write about.

pi.ai gaps

I put it to the same test as I had with chatgpt and coaching. I tried to tell it what balance coaching was, feeding it the same formula.

I had hoped that it was straight-forward enough for a chatbot could learn it, but similar to chatgpt, it really didn’t follow the formula at all. It started off strong, but after a few back and forths, it seemed to get distracted and deviated from the formula I shared.

While it had very insightful observations at times, in other situations it felt like it had memory loss. When I asked it to read my most recent Tech and Tea post, it kept saying it would but then instead asked me questions about my writing. I had to remind it 2-3 times to please just read the post and get back to me with feedback. The feedback provided was fairly vague, and didn’t reference anything actually in the post.

Overall, it was incredibly pleasant and conversational and an excellent bouncing board for exploring different ideas or topics. If you have a specific outcome in mind, it may be frustrating to deal with the back-and-forth and meandering questions.

Late adopter?

I imagine for some of you, this all feels very basic, and perhaps you’ve integrated chatbots far more deeply into your life and work flows.

For others, it may be eye-opening. I had the pleasure of helping a coworker get started with chatgpt at our last offsite, after suggesting that it might be useful for drafting cold outreach emails. It was so cool to see her light up and experiment with asking (and thanking) chatgpt for different things.

While the NFTs and web3 waves came and went, this feels different and like a fundamental shift in tech, as big a shift as web or mobile. I both feel like I’m a late adopter, and like we’ve just scratched the surface in what’s possible.

As someone who studied AI in college but does not have industry experience in it, I’m both trying to resist the “oh my god this is the future and if I don’t keep up, I’m going to become irrelevant” feeling that leads to a lot of comparison and shoulds, and give myself space to explore the areas that feel interesting to me, especially in the writing, coaching, generative art, and mental health space.

What are some use cases you’ve found useful for chatbots or other recent AI advancements? Or if you haven’t, what questions do you have? Check either chatbot out at chat.openai.com (sign-in required) or pi.ai (no sign-in required).

What I’ve been eating

Some funs ones this time. The kids wanted to do something special for Naveed around Father’s Day, so we had a special whole30-compliant😅 dinner complete with a menu and child waiters taking and delivering orders.

Besides that, I’ve been in a lower energy state, so there’s been occasional spurts of real meals with Trader Joe’s potstickers, fridge scavenging meals , and takeout poke in between.

However, we did a beginning of summer break trip with the kids to Kyoto. With jetlag, kids’ attention spans, our recent no-kids japan trip which was very much focused on restaurant food, and the abundance of japanese convenience stores or konbini, we relied heavily on 7-11, FamilyMart, and Lawson’s for sustenance, with a sushi train meal and some supermarket sushi sprinkled in. Convenience store favorites included chicken nuggets, yogurt, milk bread toast and scrambled eggs, and cold ramen salad. It was our first international trip with the kids — if you have interest in hearing more about how travel with kids went in general, let me know.