Ben Goertzel, Ph.D.¹ and Allan Combs, Ph.D.²,
¹Novamente LLC; ², California Institute of Integral Studies

1. Introduction

How might intelligences adapted for vastly different environments differ from humans in their internal processes of cognition, perception, personal identity, communication, and memory? What might be the mental characteristics of an intelligent life form which evolved in an environment radically different from Earth?

To explore these questions requires at least a rough characterization of general intelligence beyond the human scope. Psychologists define human general intelligence using IQ tests and related instruments (Ravenscroft, 2004). However, to think carefully about minds very different than human ones, requires a more fundamental understanding of the nature of general intelligence. Many attempts to characterize general intelligence have been made; Legg and Hutter (2006) review over 70! Our preferred characterization of intelligence is: the capability of a system to choose actions maximizing its goal-achievement, based on its perceptions and memories, and making reasonably efficient use of its computational resources. By this definition intelligence is then understood as mental activity which enables the organism to effectively seek and accomplish a variety of complex goals in a variety of complex environments.

It is difficult to say anything nontrivial about general intelligence in general. Marcus Hutter (2005) has demonstrated that a very simple algorithm called AIXI^tl can demonstrate arbitrarily high levels of general intelligence if given sufficiently immense computational resources. This is interesting because it shows that general intelligence is basically a problem of computational efficiency. Due to the harsh reality of computational resource restrictions, real-world general intelligences are necessarily biased to particular classes of environments. Human intelligence is geared toward the physical, social and linguistic environments in which humanity evolved; and one would expect intelligences of radically nonhuman systems to be highly biased toward their particular classes of environments in which they evolved.

Here we present a careful but somewhat speculative analysis of how differences in the “naive physics” presented to minds by different environments, might shape and effect the cognitive and social structures of these minds. We illustrate our general ideas with some specific speculations about minds in complex fluid environments; i.e. creatures which evolve on planets consisting entirely of water.

2. Minsky’s “Universal” Ingredients of Intelligence

As a prelude to our analysis, we recollect Minsky’s (1985) bold “anthropic intelligence” hypothesis. He claimed that any truly intelligent being, at least any one that we are able to communicate with, must think and act in a fashion not unlike ourselves. More specifically, he posited that any intelligent alien would share with ourselves certain mental ingredients including.

SUBGOALS ------------ to break hard problems into simpler ones.

SUB-OBJECTS -------- to make descriptions based on parts and relations.

CAUSE-SYMBOLS---- to explain and understand how things change.

MEMORIES ------------ to accumulate experience about similar problems.

ECONOMICS ---------- to efficiently allocate scarce resources.

PLANNING ------------- to organize work, before filling in details.

SELF-AWARENESS--- to provide for the problem-solver's own welfare.

We will take issue with the particulars of this hypothesis below, in our discussion of minds in complex fluid environments. But even if Minsky’s observation were correct, it’s not clear how useful it would be. His ingredients are so abstract that they might apply to any of a wide variety of environments (“worlds”), so that the actual form of a communication -- the physical form of information exchange, metaphors utilized, etc. – might vary extremely widely.

3. Naive Physics

An important notion to introduce here is that of “naive physics” – that is, the physics of an environment as it is presented to an embodied mind in its everyday interactions with the world (a common notion in Artificial Intelligence (AI) theory; see e.g. Hayes, 1995).

Specific aspects of naive physics related to objects in our own everyday environment, for example, include (but are not limited to):

Specific aspects of naive physics related to temporality and causality, in our own everyday environment, include:

Our own minds – and the minds of other animals sharing our everyday environment – are strongly tied to these “naive physics” patterns, which have to do with the ways our brains and bodies are structured, the way our environment is structured, and the emergent patterns spanning our brains, bodies and environment.

4. Minds in Complex Fluid Environments: Water Worlds

As a thought-experiment regarding intelligences adapted to fundamentally alien environments, it's interesting to ask: What differences might characterize the mind of an intelligent agent evolved in a “complex fluid environment” -- meaning, a portion of the universe dominated by liquids or plasmas widely varying in viscosity, density, Reynolds Number and so forth? To illustrate the complexity of fluid behavior we refer the interested reader to a few sample videos depicting phenomena such as turbulent water flow, fluid vortices, and moving fluid patterns vaguely reminiscent of life forms (Videos 1 and 2).

In a complex, fluid environment, one may suspect that – on average, compared to the situation in worlds dominated by solid objects -- each event would be strongly correlated with multiple possible future and past events.

Hence, one may suspect that minds adapted to this sort of environment might not develop a “folk psychology / folk physics” notion of causality in the same way that humans have (see Gottfredson, 1998) for a review of various aspects of folk psychology including this one).

Note that, as we have discussed extensively in another publication (Goertzel, 2006), causality is not necessarily a fundamental concept in any human scientific theory. Consider quantum physics and the Heisenberg Uncertainty Principle, where there are an infinite number of possibilities and outcomes, and where knowing a particle's location cancels out the ability to know its momentum. Instead of causality we have indeterminacy. Causality, at the level of an individual's personal experience, is based on an intuitive grasp of reality, and is therefore a concept we have created to help conceptualize our everyday world. It is a “naive physics” notion not a “scientific physics” notion.

Causality, however, is arguably one of the roots of the notion of the folk-psychology notion of “will” as in "willing" something to happen and thus causing it to happen. The concept that “agent A wills X” is tied in with the notion that “some internal event E1 within A's mind, causes some internal event E2 in A's mind, body or surroundings to happen” (Goertzel, 2006). Of course there is much more to will than causality – but without our conception or feeling of causality, then the notion of will is likely to cease to exist.

Evolving in a water world, however, might dramatically alter all notions of causality or will. This is an important consideration for "water worlds" which may be amenable to life, have been discovered even in our own solar system, e.g. Encedalus and Europa (Chela-Flores 2010). Outside our solar system these include a super Earth (planet GJ 1214b) orbiting a low mass star with a hydrogen atmosphere but whose composition appears to be primarily of water (Charbonneau et al., 2009). Planet GJ 1214b orbits within the classic habitable zone in which complex life may evolve (Lal 2010). Therefore, considered as a thought experiment, we might ask what type of life and what type of intelligence and mind might evolve in a world which consists entirely of water.

A world which is ocean would be a world dominated by fluid dynamics and turbulence, and characterized by changing ocean current and columns, deep versus surface environments, and a hierarchy of eddies. Turbulence is characterized by chaotic, stochastic property changes, rapidly changing diffusion rates, convections ranging from high to low momentum, rapidly alternating velocities, and the development of vortices appear with interact with each other and which deflect and attach themselves to a solid surface, such as living object, which are then sucked inside.

Evolving and developing inside such a chaotic world would directly impact one's notions of causality and will. Without causality and will, what kind of psychology might exist? One possible alternative could be a psychology built around the notion of “flow,” so that for instance

The naïve physics of agents in fluid environments would involve many different kinds of flow, including various sorts of turbulent and laminar flow, potentially different sorts of plasmatic state, etc. If these agents often encounter turbulent regimes, then just as we have a well-developed naïve physics for relations between objects, they might have a well-developed naïve physics for the transition from orderly to chaotic/turbulent flow. “Routes to chaos” and “routes to complexity” as studied in dynamical systems theory might be intuitive and immediate to these agents.

Figure: Fluid-like” visualization of period-doubling bifurcation route to chaos.
From http://physics.mercer.edu/hpage/pcompr.gif.

Rather than focusing on building items from components, these organisms would likely focus on creating self-organized patterns via exerting pressures of various sorts from various directions. Understanding of routes to chaos and complexity would be important here. Shaping would likely be more important than building. Physical tools might be less important than patterned sequences of contextually appropriate shaping activity.

Our language, itself, is based on decomposition of meaning into combinations of discrete objects; but this is not the only possible way to communicate. One can imagine a language (an organized system of patterns intended to convey meanings) composed of continuous flows and changes in fluid state. In this sort of language, there might be no atomic component remaining persistent across different instances of communication – that is, nothing similar to a word, character or phoneme. Rather, the persistent patterns across different instances of communication with related meaning might be higher-level – say, statistical structural patterns in a strange attractor emergent from a borderline-turbulent fluid state.

Consider for example, the language of intelligent creatures who swim the oceans of Earth, i.e. whales and dolphins. Although lacking vocal cords, dolphins emply the sphincter muscles within the blow holes to communicate by producing a complicated system of whistles, squeaks, moans, trills and clicks of varying frequencies. The clicks are also a device for echolocation, or sonar, the clicking sounds bouncing off objects with the echoing sound waves sensed by the dolphin's bulbous forehead and lower jaw. These communicative mechanisms enables them to determine the distance, size and shape of objects (Berta et al., 2005).

Likewise, whales employ a variety of frequencies, with some species singing hour-long songs which can be heard hundreds of miles away. Like dolphins they also employ echolocation (Berta et al., 2005).

Whales and dolphins are mammals. It is the consensus view that whales may have lived on land and then returned to the sea. However, what of creatures who evolved on a world without land, and where the pressures and weights experienced at the bottom of the ocean make any experience with solids an impossibility?

In a water world of turbulence would alien life forms develop morality, a sense of individuality? Would they even evolve hard bodies and bones?

Consider for instance the case of intelligent agents comprising viscous fluid or plasma regions, floating around in some less viscous substrate. These agents might then be able to merge with each other, or temporarily open their boundaries to each other to share internals to a limited extent. Also, in this state, flow patterns in one agent might be able to “seed” similar flow patterns in another, allowing transmittal of “thought patterns” in a direct way not easily possible for agents with solid boundaries around their brains.

If this sort of communication were possible, the notion of an individual mind as an atomic entity might be much less significant, and the “phenomenal self” on which human psychology is based might take a very different form, or not exist in any recognizable way.

Overall, what this thought experiment suggests is that, in the case of intelligent agents evolved in a complex fluid environment, very familiar qualities such as cause, will, self, and compositional language might not be present.

Returning to Minsky's enumeration above, we find that many of his posited universals could plausibly be quite different for minds evolved in complex fluid environments (the most suspect ones are underlined):

SUBGOALS ----------- to break hard problems into simpler ones.

SUB-OBJECTS ------ to make descriptions based on parts and relations.

CAUSE-SYMBOLS--- to explain and understand how things change.

MEMORIES ---------- to accumulate experience about similar problems.

ECONOMICS -------- to efficiently allocate scarce resources.

PLANNING ----------- to organize work, before filling in details.

SELF-AWARENESS-- to provide for the problem-solver's own welfare.

The notions of subgoal and subobject seem closely tied to the solid-object environment and might not play any significant role in the architecture or psychology of intelligences evolved for complex fluid environments. The notion of cause-symbols might be replaced with something quite different based on notions of flow and turbulence rather than causation. And the rigid notion of an individual self, strictly separate from other selves, might be replaced by something quite different in a mind adapted to a complex fluid environment.

5. Communicating with Intelligences in Profoundly Alien Environments

How then might we communicate with intelligences existing in a complex fluid environment, in a world of water, or another environment radically different from our own?

Applying a general pattern recognition algorithm to detect the patterns constituting their minds and modes of communications would likely be infeasible. However, an interactive approach might be more successful. One potentially valuable conceptual strategy would be to:

1. Use a pattern recognition algorithm to recognize simple regularities in the environment

2. Project identical copies, and novel variations, of these regularities into the environment and observe what other regularities occur as a response

3. Bias one's pattern recognition algorithm to find new (maybe more complex) regularities, that are similar to the regularities that have already obtained responses

4. Then return to step 2 above with these new patterns, and also:

Note that the “return to step 2” portion of Step 4 is essentially monologue, whereas the other portions of Step 4 are attempts at dialogue. Of course this sort of generic process is quite crude and is merely intended as a suggestion for how to get started with probing for intelligent agents in a very unfamiliar environment. If interesting patterns are discovered, then after that the interaction and discovery process is likely to go in unexpected directions.

For example, if one were interacting with a complex fluid environment suspected to contain intelligent agents, one might proceed as follows:

6. Conclusion

At the present stage in the development of the theory of intelligence, it is difficult for us to know which aspects of human intelligence reflect general properties of “intelligence of autonomous agents under realistic computational resource restrictions." Rather, we must consider which properties reflect adaptations to the particular environments in which we evolved, and those quirks of human nature arising via self-organization and genetic drift. Here we have presented some ideas in this direction, looking at the contrast between our solid dominated environment and complex fluid environments, and thinking about the differences in cognition that might ensue. The conclusions, while speculative, are quite dramatic: a fairly straightforward analysis suggests that constructs very near and dear to human experience, such as will, cause, self, and language might in significant measure reflect the fact that we evolved in a complex physical-acquatic environment and have spent most of our lives on solid ground.

By contrast, intelligent extraterrestrial life may evolve on water worlds, gas-planets, and super- and miniature-Earth-like worlds. Moreover, some evolved life forms may not be carbon-based, or may possess a genome radically different from those of Earth (José et al., 2010; Naganuma and Sekine, 2010; Schulze-Makuch, 2010; Sharov, 2010) such that physically, they may not even be recognized as alive, much less intelligent.

Hundreds of planets have been discovered orbiting distant solar systems, many in what are called "habitable zones" where life may proliferate (Lal 2010). These include worlds of water, super gas giants, and super-Earths. In fact, water worlds are not limited to distant solar systems, but include Encedalus and Europa both of which are suspected to harbor warm-water oceans beneath their icy shells, and within which may dwell life (Chela-Flores 2010).

Thus it is imperative that we must consider the likelihood that intelligent extraterrestrial life may have evolved differently from intelligent life on Earth. In the search for extraterrestrial life we must be careful not to employ a notion of general intelligence that is too strongly implicitly biased to the particular environments for which our own intelligence is adapted.