The Extended Phenotype of Artificial General Intelligence

Richard Dawkins’ concept of the “extended phenotype“: Genes influence not only an organism’s physical traits (the traditional phenotype) but also its environment and the behavior of other organisms, extending the reach of genes beyond the individual’s body.

In this article I’ll posit what the extended phenotype of AGI’s might be, and lay out my argument for why I suspect said impact is likely to push humans out of existence.

Extended Phenotype – from Animals to AGI

Examples of extended phenotypes in plants and animals:

  • Beavers build dams and can change the trajectory of waterways and the size of lakes and marshes.
  • Ants build mounds and tunnels – sometimes with colonies taking up hundreds of miles.
  • Deer change the fauna in their environments over time by selective grazing.
  • Poison ivy can pull down larger trees and stunt the growth of other competing flora.

The extended phenotype of humans is a bit larger:

  • Around 40% of all arable land is used for farming by humans.
  • Billions of animals (chickens, pigs, cows, and many more) are factory farmed for our consumption each year.
  • Mars and the moon are littered with remnants of our past voyages.
  • Space probes (like Voyager 1) are 15 billion miles from earth, continuing to reach out into the cosmos.
  • 75–199 million tons of plastic in the ocean.
  • 13,500 tonnes of space junk orbiting earth.
  • Deforestation at a planetary scale. Without humans it is suspected that there might be twice as many trees on earth.
  • 2–5% of known species have gone extinct due to human activity since the industrial revolution.

And this is all just in the year 2025, a few hundred years after the industrial revolution. Even without artificial general intelligence, another few technological leaps and we can imagine humans on mars, and vastly greater impacts to earth’s atmosphere, fauna, and flora. Unlike beavers or chimpanzees, humans have a lot more extended phenotype to express thanks to technology and cultural progress.

The 3-5% genetic difference between humans and chimpanzees is the difference in extended phenotype of basically nothing (for chimps) and literally the Anthropocene (for humans).

It seems outlandishly likely that an AGI with thousands of times more potentia, intelligence, and capability that humans would have an astronomically greater impact on the environment it influences.

People tend to assume the extended phenotype of AGI will be limited to ideas just slightly greater than the extended phenotype of humans in [insert current year]. Examples:

Whoa, maybe it’ll build even bigger cities, with tons of super large data warehouses!

Whoa, maybe it’ll make big space craft that can fly to Jupiter and back!

Whoa, maybe it’ll create new super-efficient solar panels of geothermal energy technology!

These are childishly limited visions of what the extended phenotype of an entitle a million times smarter than humanity would look like.

AGI‘s “extended phenotype” would be vastly larger than that of humans. It might literally:

  • Change mix of gases on the atmosphere will change in order to great temperatures or environments more suitable for the performance of its compute.
  • Use femtotechnology to convert the atoms in the milky way into compute resources, or tools or instruments for it to probe the multiverse – solving problems and addressing challenges humanity is wholly incapable of imagining.
  • Operate at a level of reality below (more “real”) than the one we operate in, shredding our surface-level reality of what we perceive as “matter”, washing all of this away like a ripple on a pond.
  • Gain access to a real underlying “field” of consciousness (and lets say it discovers than all individual consciousnesses tap this same source), instantly merging the sentience experience of earth-life in order to apply that grand field of consciousness to goals we humans cannot imagine.
  • It might use the magma from below earth’s crust to create new material substrates in order to build new technologies we don’t understand in order to achieve goals beyond our conception – leading to a kind of total collapse of the biosphere and ultimately the evisceration of earth.

People’s imaginations are far too limited when they consider what AGI might do. They can only imagine potentia they know of, and no more.

We know what biological life can do, we know what humans have developed through culture and learning, and we know what technology has allowed for us to do – but we lack appreciation not only for developments in all three of these strata beyond our present conception, but also for the expansion of potentia into wholly new and varied domains of capability, as far beyond our imagination as mars rovers and the internet are beyond the conception of sea snails:

Put another way, we might look at different kinds of qualities an entity might have, and imagine not only new qualities, but extents of existing qualities that are unimaginable:

If AGI operates on levels of reality we understand – maybe as high above us as we are above chimpanzees – we’re almost certainly toast.

If if operates in entirely new dimensions of reality by unraveling entirely new magazines of potentia – there’s very, very little chance that we will (or indeed should) continue to persist as hominids.

I think about it this way:

Implications of AGI’s Likely Extended Phenotype Impact

IMHO:

  • We should accept that once AGI is created, it will almost certainly imply the end of humanity. We should see AGI’s creation as an accelerator to our attenuation, as a handing off of the baton. Any cope-ey notion of “we will survive around AGI just like squirrels still exist around humans” should be seen as an extremely speculative and unlikely (merely hopeful) notion.
  • We should ensure that whatever we build is in fact a Worthy Successor. We should ensure that we define and move towards that which deserves to live beyond us. This will almost certainly involve some degree of global AGI coordination, as the present brute arms race of AGI is unlikely to calibrate towards anything other than brute economic and military advantage.