[LAM 1] Encounters With AI and People

Hyderabad, January 7, LA 2 (conventionally called 2024)

If you’re in my world (which surprisingly still seems very small), it has been an in-expletive-sane 2 years since the launch of ChatGPT on November 30, 2022. To put things in context, I have built and gone-to-market with 4 AI software products since then. In the meantime, I have also spent ~6 months in New York, 3 in the UK, ~6 months in San Francisco most recently, and the rest with my family here in India.

In this process, I have witnessed at least 4 different tectonic shifts beneath the players in this dramatic technology wave. It has been a profound experience to see people — including myself — react to this technology at such a singular time in history. Some moments stand out. Here they are.

Moment 1: The launch of GPT-4

On March 14, 2023 (pi day) OpenAI launched ChatGPT-4. If you don’t know what that is, OpenAI merely upgraded the model, i.e. the actual intelligent software, behind the chatbot on www.chatgpt.com. For a long time, this advanced model required a paid subscription, although it is now free. I immediately got the subscription.

I remember being sick and in a constant state of dread all the way to my stomach for about 2 months, because I couldn’t believe this thing was already here. This thing, this machine, that was incredibly good at thinking and generating very complex software programs — something that I never thought machines would be able to do — and riffing with me on non-trivial topics like philosophy, history, strategy and math. It was available at $20/mo.

I am not exaggerating when I say this, but one of my biggest fears at the time was that OpenAI would realize how cheaply they were giving away something so useful, and that they would increase the price, or worse — that someone high up (the government???) would realize that it was a mistake for so many people to have access to something like this and take it away completely. What an insane product.

Moment 2: "You're solving the wrong problem!"

A few months later, during a bug-finding session with ChatGPT (after I got through the initial shock of GPT-4, it had become a critical part of my coding workflow, as I believe it did for almost every programmer on Earth), I saw something very interesting when I asked it to fix something.

Me:kindly fix this part of the code
ChatGPT: Sure, here’s how to fix your python code — [better solution]. BUT, if you look closely at your initial code, that’s not the problem you should be fixing, but this other thing [elsewhere in provided context]…

Yes, I have abused the use of quotes here by paraphrasing, owing to my lack of precise recollection of something that happened over a year ago. Nevertheless, I vividly remember the feeling that answer gave me — it was a mix of surprise and awe.

Months after its launch, I was still learning new things about it, and learning that it could surprise me with its intelligence.

Moment 3: "But what will happen to Us?"

A few months after the bug-fixing adventure, I built a tax AI product called Taxtalk. It was a simple chatbot that took any question about tax law, found the relevant legislation, applied it to the question and returned an answer, along with references to the relevant sections and clauses.

We first presented Taxtalk to 3 partners at a large tax advisory firm in Mumbai. They seemed reasonably impressed, but each of their reactions to the technology were very different. The first partner immediately saw the potential, and began to riff on ideas as to how it could be improved and where it could go in the future.

Another partner immediately went to nitpicking flaws with Taxtalk’s answers — “Ahhh, see it got this part wrong!” “It missed this nuance!” Yes, you can get away with being very unspecific if you’re unbothered by peasant concerns about staying true to the spirit of quotation marks.

But the reaction of the third partner was most interesting, perhaps because it mirrored my first reaction to ChatGPT-4. This partner sat quietly throughout the demo, speaking not nearly a word. Once everyone else was done, he quietly began to speak, “But… what’s going to happen to our jobs? If [your product] can do this now… Already, and you’re definitely going to keep improving it, what is going to happen to us? To our industry? What will we do?”

My heart went out to him.

Reflecting

How could I explain that 85% of what I had built was basically not to my credit, but to the AI giants building the intelligence layer underneath? How could I tell him I did not know what the future held? How could I answer his question, since I am still figuring this out myself? How could I explain that the ground is shifting beneath me just as fast as (or even faster than since I’m closer to the source so I have a smaller radius) it is, beneath him?

Part of the trouble I think we have adapting to this new thing is a narrative violation in our expectations of computers. They weren’t meant to do anything this general purpose; compute was basically a mathematical trick for most of their history. In the last 3 years, we have seen technology eat up entire industries and transform our understanding of creativity and intelligence. It will continue to do so.

Zooming out, recent progress in AI has once again blown open the BIG questions that we collectively became comfortable not addressing. We will have to find new, betters answers to all the big questions surrounding the human experience: what it means to be human, to have intellect, to be creative, to suffer, to love, to do things. Why are we here? Where did we come from? Where do we go? The answers that most reasonable people traditionally held have become obsolete. I am personally super excited to see what we figure out.

Bonus Moment: "sure"

There is a company called shapes.inc. They allow you to make custom chatbots that can hang out with you on discord. It appears they found lots of use in roleplaying servers. Sometime before they took twitter over by slop-storm in mid-2024, I found their discord and was immediately fascinated. I found that the company was about to shut down its product less than 2 months before it experienced viral growth. They had ~20k users then, and a post on their support channel from a founder announcing they were considering changing directions and requesting user feedback. Apparently they received tons of positive feedback about the product and improved it significantly, and eventually found a mechanism to unlock growth and retention.

What caught my attention was a character named Blizzy. It was a female anime avatar, stuff that culture seems to love these days, but its way of interacting with people was very unique. To try it out, I “@-messaged” it saying “yo”, and it asked me a question on the group chat about my interests or hobbies or something. I answered, and asked it playfully if it was interested in the same thing. And it replied with 1-word, “sure”. That floored me. AN AI CHARACTER THAT COULD BE RUDE??? I WAS NOT PREPARED FOR THIS.