Baby AGI and Developmental Cognitive Science

The only planet we live on - the Earth - is vulnerable to cosmic events (near-earth objects? near-earth supernova?) which can, quite literally, freeze or fry human life. This calls for a need to colonize beyond Earth, and research and development into space exploration and habitation.

But there is another parallel direction for research. Our biological bodies are ill-suited for living beyond Earth, especially for interplanetary travel. But while artificial limbs are important elements for overcoming our biology bodies, there is yet another component that is as crucial if not more… our minds. I'm interested in the creation of an artificial mind.

A separate motivation for this also stems from mere curiosity. Sitting behind those darting eyes lies one of the most mysterious systems in the universe – 100 billion neurons, each connected to almost 10 thousand others – all made of the same carbon we obtain when we burn sugar on a stove. How can something as nonliving as a piece of rock be made of the same basic elements that make up something as lively as you or me? Perhaps, along with the mysteries of the cosmos, mysteries of our minds will continue to be amongst the longest unsolved puzzles of humankind.

Of course, once Artificial General Intelligence is attained, speculation has it that it will soon explode into a singularity, for better or worse. Hopefully, it proves helpful in finding solutions to cancer, physics, inequity, and sustainable progress, amongst other things. (Or, maybe, the Covid19 pandemic taught us some form of equity already?)

But where does one begin taking a crack at AGI? Maybe one way is to look inward at the only example we have of sufficiently general intelligence – ourselves. But clearly, no direct intervention puts our minds into their adult forms. Plus, studying the adult mind is like studying the artifacts left behind by the sociocultural environment the individual grew up in. Doesn't seem like a great way to understand where nature ended and nurture began.

At the AGI 2021 conference, I was pleasantly surprised to be affirmed by this primitive conclusion. Bootstrapping was an important theme throughout the conference: We should not aim to create Artificial General Intelligence. We should aim at creating a (minimal) system that could give rise to Artificial General Intelligence. One may argue that this itself is sufficiently general intelligence.

Clearly, such a system already exists: the human child. If the reader is a younger me, pondering over what intelligence is, how our brain-minds do what they do, then know that Childhood Cognitive Development and Bootstrapping could be a valuable perspective to consider. I was pleased to hear that Michael Tomasello, one of the leading figures in First Language Acquisition, also pointed towards studying Phylogeny and Ontogeny to understand human cognition better in his Rumelhart Prize Presentation at the CogSci 2022 Conference.