Situational Context#

In early 2023, the situation was different. AI had gone mainstream; it was the second time this had happened, but in a fundamentally different way.

The kind of popularity AI enjoyed throughout much of 2022 began to shift. The era of funny, distorted image generators with eight fingers per hand vanished faster than most had anticipated. The reason: MidJourney v4, which produced images of striking quality. At that point, it already took some training to tell certain images apart from real photographs.

People had begun to develop a certain respect for the technology, and yes—up until then, image generation was the only popular face of the revolution that was about to unfold.

November 2022 was historic. A completely new facet of AI saw the light of day: one that could write essays, poems, scripts, give advice, talk to you—entirely free, through a chat interface. Something never before seen, which caught off guard those who hadn’t been paying attention to the steady advances in NLP toward LLMs.

The surprise intensified just four months later, with the announcement of GPT-4 in March: an even larger, more powerful LLM—one that could see.

I still remember my astonishment when OpenAI’s CTO sketched a rough webpage, asked GPT-4 to write the code, and how that extremely basic HTML left me speechless and programmers everywhere holding their breath. That vision capability would only become available for us to try nearly six months later.

The leap was so sudden, and the iteration from GPT-3.5 Turbo to GPT-4 so quick, that many even believed GPT-5 would be released that same year, ignoring the fact that GPT-3 (the base model, not the chat version) had come out in 2020—not just a few months earlier.

In any case, the hype and expectations were higher than ever. You could hear people speculating that AGI would arrive in two years, that we’d have fully AI-generated movies personalized for each individual the following year, that all jobs would soon be automated and replaced, and so on.

The collective hysteria grew so large that an initiative was launched to pause large-scale AI experiments for six months, with the goal of conducting a deep study into their potential risks and socio-economic impact. And while at first glance this might have sounded like a hippie movement led by outsiders caught off guard by the technology, it was in fact supported and signed by highly influential figures in the field such as Elon Musk, Emad Mostaque, Yoshua Bengio, and Geoffrey Hinton, among others.

And it is amid all these events that our story begins.