The day ChatGPT landed in my inbox, I had no idea what was coming

The early demo I received on November 30, 2022 was, to me, yet another nothing-burger.

It was a Monday. November 30, 2022. I was working as a senior writer covering AI for VentureBeat, where I had started six months earlier. Elie, OpenAI’s one and only PR person at the time, sent me an email about a new release. It included a link to an early research demo called ChatGPT.

When I took on the role at VentureBeat the previous April, I had never heard of OpenAI. I had been covering enterprise tech as a freelancer, not research, and while I had covered AI, it was not my daily beat. So when I first Zoomed with Elie, I told him I needed some time to catch up.

Six months and probably 100 stories later, I considered myself well-versed in the AI universe. So when Elie sent me that demo, I figured I had it in the bag.

Everyone had been talking about the possibility of GPT-4, and the entire AI research community was communing down in New Orleans at the NeurIPS conference. So that’s what I hooked the latest announcement to: a story titled OpenAI debuts ChatGPT and GPT-3.5 series as GPT-4 rumors fly.

I felt confident that this ChatGPT wasn’t any big thing. After all, that part of the announcement was just an “early demo,” and simply another part of the main announcement: The GPT-3.5 series that is an interactive, conversational model whose dialogue format “makes it possible for ChatGPT to answer followup questions, admit its mistakes, challenge incorrect premises, and reject inappropriate requests.”

Just a couple of weeks earlier, Meta had announced a new open-source LLM called Galactica, that it described in a paper as “a large language model for science” meant to help scientists with “information overload.” The announcement went terribly: Many were appalled by Galactica’s very unscientific output, which, like other LLMs, included information that sounded plausible but was factually wrong and in some cases also highly offensive. Within three days, the Galactica public demo was gone.

There was also Google’s LaMDA, which made headlines in June 2022 when Google engineer Blake Lemoine, told the Washington Post he believed the conversational AI was sentient.

So it seemed to me that ChatGPT might fall into the same realm of “meh,” or “hell, no.” I was pleased to note that as a nice Jewish gal the tool did not seem to be able to spew anti-semitic content, and I noted that ChatGPT was “trained to emphasize that it is a machine learning model.” But then I let it lie.

Within a week, however, it was clear that ChatGPT was a big, big deal. By December 5, it had over a million users, which according to CEO Sam Altman took GPT-3 nearly 24 months to get to and DALL-E over 2 months.

From a VentureBeat story I wrote that day:

The ‘interactive, conversational model,’ based on the company’s GPT-3.5 text-generator, certainly has the tech world in full swoon mode. Aaron Levie, CEO of Box, tweeted that “ChatGPT is one of those rare moments in technology where you see a glimmer of how everything is going to be different going forward.” Y Combinator cofounder Paul Graham tweeted that “clearly something big is happening.” Alberto Romero, author of The Algorithmic Bridge, calls it “by far, the best chatbot in the world.” And even Elon Musk weighed in, tweeting that ChatGPT is “scary good. We are not far from dangerously strong AI.”

There was criticism, too, particularly in introducing the idea of “hallucinations”:

Like other generative large language models, ChatGPT makes up facts. Some call it “hallucination” or “stochastic parroting,” but these models are trained to predict the next word for a given input, not whether a fact is correct or not.

Some have noted that what sets ChatGPT apart is that it is so darn good at making its hallucinations sound reasonable.

Technology analyst Benedict Evans, for example, asked ChatGPT to “write a bio for Benedict Evans.” The result, he tweeted, was “plausible, almost entirely untrue.”

More troubling is the fact that there are obviously an untold number of queries where the user would only know if the answer was untrue if they already knew the answer to the posed question.

That’s what Arvind Narayanan, a computer science professor at Princeton, pointed out in a tweet: “People are excited about using ChatGPT for learning. It’s often very good. But the danger is that you can’t tell when it’s wrong unless you already know the answer. I tried some basic information security questions. In most cases the answers sounded plausible but were in fact BS.”

Three years and what feels like a thousand updates and products later, ChatGPT has over 800 million active users. For many ChatGPT is to AI as Kleenex is to tissues.

Sometimes I wonder why I didn’t sense what was coming in that first email from Elie. Maybe I was still too new on the AI beat. Or maybe that’s just how it works, even when you are so deeply into a daily beat like I am with AI—back then with VentureBeat and now at Fortune.

Sometimes you don’t see what’s coming until you watch the world react.