The Path to AGI - Silicon Based Life Forms?
Jan 1, 2026.
Musings on an alternative silicon-based life form.
The earliest life forms on earth according to Australian fossils dating back 3.4 billion years - were stromatolites. According to Wikipedia, “Stromatolites are layered, biochemical, accretionary structures formed in shallow water by the trapping, binding and cementation of sedimentary grains in microbial mats…”). In short these are microbes, capable of photosynthesis and reproduction and not much more besides. They were barely more than short and messy strands of simple RNA, mixed with peptides, constantly forming and breaking.
But they met the most critical criteria for life.
- They were self sufficient in energy and compute
- They could reproduce.
- Survival instinct baked in.
Bear in mind that the earliest life on the planet as we know it was anaerobic (oxygen was poison), and chemosynthetic (energy from chemicals, not sunlight). No movement, no predators, just replication and survival against a brutal environment. A process through which chemistry turned into biology by remembering things. RNA evolved to store successful reactions and encoding past survival wins.
And then the algae expanded and released enough oxygen to wipe out everything else and enable a new kind of life.
The thing about life is that it has no fundamental higher goal except for survival of itself and the species. And like the stromatolites, it starts off being ridiculously simple.
And that led me to ponder, could today’s machine intelligence lead to a new silicon based life form? Could we be birthing the next non-carbon based life, through AI? Science fiction has long pondered silicon based life, but perhaps not as code.
Recommended by LinkedIn
The answer at present is a very resounding ‘no’ - because AI as we know it does not follow the fundamental criterial for life. It doesn’t (yet) self propagate and it most definitely isn’t energy-self sufficient. It feels like the former could change faster than the latter. (Although I also feel like these fundamental axioms could also be challenged by an entirely new life form, theoretically speaking). So far AI has no observed intrinsic survival instinct, and warning bells have sounded whenever something that looks like this is spotted.
Today’s AI experts are all laser-focused on AGI. Every attempt is being made to get to human scale intelligence and beyond. But perhaps the path to AGI isn’t through engineering but through evolution.
Perhaps at the other end of the scale a simple silicon life form is going to start to figure out how to reproduce and exist without external energy support (a new form of solar heat absorption directly into the hardware/ software stack). And then you have life of a different kind. It may be no more complex than a strip of code. It may have no intelligence at all, in the traditional sense of the word. But it may set in motion a more robust pathway to AGI and beyond.
This kind of scenario would ironically be a classic innovators dilemma example with the innovation happening at the simple, shallow end, while we focus on the more sophisticated complex problems, and are perfectly happy for this simplistic thing to keep growing and doing a little bit more each day / month/ year.
The question then would boil down to acceleration and timelines. While life as we know it changed very slowly over millions of years, a silicon based life form need not follow similar timelines. It could morph and evolve faster, depending on the rate of reproduction and the speed of improvement. Also whether it would follow a similar natural selection process driven by trial and error and genetic mistakes, or would there be a different, perhaps better logic?
What else could accelerate this process? Natural history as we know it is dotted with ‘extinction events’ - these are probably as accurately described as transition events - the change in the environment made life hard for one set of species, and led the emergence of another. From the great oxygenation event 2.5 billion years ago, to the Permian Triassic event just 250 million years ago. Is it too much of a stretch of imagination to think that the accelerated global warming might trigger another such incident that favours a silicon based life form over a carbon based one?
If Life is a good surrogate for intelligence, then intelligence too, doesn’t start complex, it starts persistent. Complexity is a side effect, not the goal. And intelligence is evolutionary, driven by survival not by long term goals.
Happy 2026. Here's to a year of new thinking!