Catch the video of this essay here!

Artificial general intelligence.

It's seen as the holy grail of technology and computing. Some call it folklore, others see it as an impending reality.

Literally billions have been invested to try to achieve it.

And if realized, AGI can usher forth unprecedented change and fundamentally reshape how we interact with tech and one another.

Today we’re looking at OpenAI, the creators of Dall-E and ChatGPT and what they’ve put forth as their roadmap to achieving AGI.

What is AGI?

McKinsey defines artificial general intelligence as “AI that is as capable of learning intellectual tasks as humans are.”

Academics and practitioners are split on the timeline for when such technology will arrive. However, most agree that the impressive capabilities that we’ve seen in just the last few months alone from generative AI like Stable Diffusion and ChatGPT, signal that we’re entering a brave new world of artificial intelligence.

And one of the companies at the forefront of spearheading this is OpenAI.

OpenAI’s origin story

Open AI was originally founded in 2015 by Sam Altman and Elon Musk as a non-profit research lab. With an initial $1 billion in funding, Open AI’s initial vision was to conduct research openly and share their work and patents with collaborators around the world.

Musk even said that its core aim was to create open-source AI tech so that Google and other tech giants didn't have a monopoly over it.

In the beginning, OpenAI’s research was bent towards Reinforcement Learning and was marked by achievements like defeating human players in 5v5 competitive DOTA and building robots that could solve a Rubik’s cube with one hand.

While certainly interesting and important results in their own right, they lacked the gravity that would cause people to think that AGI was coming soon.

Enter GPT and Dall-E

However, all that changed with GPT2.

Announced in February 2019, GPT2 was a successor to the first Generative Pre-trained Transformer language model. This model in particular was so capable at generating believable text, OpenAI publicly cautioned that it was too dangerous to release. However, they soon changed course and not only open-sourced GPT2 but also became more comfortable releasing even more powerful language models like the infamous ChatGPT.

And not just in natural language, but OpenAI has caught the public’s attention with their image generation model Dall-E. Seen as a precursor to the open-source Stable Diffusion, Dall-E showed the world that you could generate impressive-looking images with just a prompt of text.

We see from a bird-eye view, all of these machine learning models and the features they offer are approaching what many of us consider signs of true intelligence. Knowing this, OpenAI wrote a charter detailing how they view their efforts going forward.

OpenAI’s roadmap towards AGI

Why does OpenAI even want to build AGI in the first place?

In their own words:

“AGI has the potential to give everyone incredible new capabilities; we can imagine a world where all of us have access to help with almost any cognitive task, providing a great force multiplier for human ingenuity and creativity.”

How they achieve this can broadly be summarized into:

1. Get AI into the hands of as many people as possible.

2. Create increasingly aligned models (like InstructGPT and ChatGPT) that better match human preferences and are more controllable.

3. Start a global conversation about governing AI systems, distribute their benefits fairly, and ensure equitable access.

However, realizing that achieving AGI would require significantly more human capital and compute resources, OpenAI shifted from a non-profit to a “capped-profit” structure. This effectively allowed it to raise funding from external partners and build products for financial gain. Their most recent tie-in with Microsoft has Microsoft becoming the “preferred partner” for commercializing their AI tech.

Longer term, OpenAI sees the development of the first AGI as just one step in a longer continuum of super-intelligence. The key will be how the once-nonprofit can prevent misalignment and misuse especially in the light of competing financial interests.

Closing thoughts

So, is AGI achievable? And is OpenAI on the right path to reaching it?

Or are they just cracking open Pandora’s box and will be caught off-guard by the consequences of their actions?

One thing’s for sure, the next few years will be one for the history books.