How AI Can Help Us Envision More Diverse Intelligence

Key Points:

• The notion of intelligence as a linear hierarchy with humans above animals and AI at the top is flawed and limiting. We need to move beyond the idea of an "intelligence staircase."
• Creating AI that simply mimics human intelligence is too narrow a focus. We should broaden our view to consider other forms and manifestations of intelligence beyond just replicating human cognition.
• The "morphospace of cognitive complexity" provides a useful framework for visualizing and exploring the vast landscape of possible intelligences across dimensions of computational complexity, autonomy, and social interaction.
• There is currently a gap between biological and artificial systems in terms of autonomy and social complexity. Biological evolution has driven organisms to increasing levels of independent action, adaptability, and social coordination that most current AI systems lack.
• Humans are cognitively distinct in our highly complex individual and social intelligence, and ability to flexibly shape our environment through culture and technology in an "extended mind." But AI is not bound by biological constraints and could potentially develop radically different forms of alien intelligence.

The notion of intelligence as a linear hierarchy, with humans perched above animals and AI at the summit, is an enduring but flawed vision that has captivated the AI community for far too long. This “intelligence staircase” metaphor simplistically ranks the complexity of cognitive abilities and fails to capture the true essence of intelligence in all its forms. AI luminary Yann LeCun has often spoken about the incremental evolution of machine intelligence, suggesting that before AI can rival human intellect, it must first navigate the more modest cognitive landscapes of a mouse or a rat, much like a child learning to make sense of the world. He envisions a gradual climb, with AI systems developing and refining their capabilities step by step, reminiscent of the natural progression seen in the animal kingdom.

The ladder metaphor of intelligence, while illustrative of cognitive advancements such as planning and problem-solving, might not serve us well in the grand scheme of AI development. Part of this myopia is the idea that the best use of AI is to create AGI. This pursuit is so narrowly focused on emulating and surpassing human intelligence that it’s no wonder people worry about a future where AI reigns supreme.

The bias of re-creating human intelligence is rooted in the very inception of AI. What else is the Turing Test if not an instantiation of machines that outwit humans at being human? Are we too fixated on creating machines that mirror human intelligence? Could this narrow focus limit the potential and scope of what AI can achieve?

How else might we think about intelligence and, by implication, other ways to place the idea of an artificial—or hybrid—intelligence?

One approach to understanding intelligence is imagining it as a vast landscape of potentialities—a multidimensional space where different forms of intelligence, both known and hypothetical, coexist. Technically, this is called a morphospace.

In this space, we find intelligences that have naturally evolved and are observable in the world around us. There’s a hypothetical space, full of possibilities of intelligences that either haven't been discovered, are yet to be created, or may not be feasible within the confines of natural laws. A void may indicate constraints or “forbidden evolutionary paths”. Morphospaces help us see the big picture: possible designs and the relationships between them.

In "Evolution of Brains and Computers: The Road Not Taken," Ricard Solé and Luís F. Seoane introduce a framework called the "morphospace of cognitive complexity." This concept is visualized in three dimensions: computational complexity, degree of autonomy, and interactions between agents.

• Computational complexity refers to the range and depth of tasks an agent can handle, like memory processing, learning, and decision-making.
• The degree of autonomy measures how independently an agent acts, specifically in creating and following their own operational rules. This aspect is key to adaptation.
• The third dimension, interaction, focuses on the collective or social aspects of intelligence, exploring how intelligence manifests beyond individual capability.

Artificial systems, barring robots, are primarily varied in terms of computational complexity. They typically lack the elements of autonomy and don't extend beyond their programmed algorithms. Conversely, living systems are more dynamic in the realm of autonomy, often occupying areas characterized by high levels of independent action and adaptability. This distinction highlights how artificial and natural systems vary in their manifestations of intelligence.

When viewed through a lens of cognitive complexity there’s a stark contrast between living and artificial systems. This gap becomes increasingly apparent as systems grow in social complexity and autonomy. In biological systems, evolution and natural selection favor autonomy. As organisms evolve, their internal computational abilities improve, enabling them to better predict and adapt to complex environments. This evolution reinforces their autonomy, propelling biological entities toward heightened complexity and self-sufficiency. This evolutionary trajectory underscores a key distinction between living and artificial systems. Presently, artificial systems demonstrate relatively limited autonomy and social interaction capabilities compared to biological entities.

In this model of advanced cognition, humans are distinct due to our unique social and cognitive complexity. Like social insects, humans have significantly impacted the planet, but our form of collective intelligence differs. Humans are not only individually complex but we also have the ability to shape our surroundings through an “extended mind”, adapting and adopting information, technology, and culture in highly flexible and transferable ways. This cultural evolution, surpassing genetic evolution, sets us apart from other living systems.

To reimagine our understanding of intelligence, we have to move past the notion of an "intelligence staircase." AI systems are not bound by biological constraints or evolutionary pathways, so present a vastly different landscape. Divergence from natural biology opens the door to the possibility of encountering autonomous, “alien” minds in artificial forms.

The fact we observe convergent designs in nature, such as the independent evolution of the eye, tells us that there are possible paradigms in intelligent systems. Evolution's emphasis on movement in a complex, uncertain world drove brain evolution, suggesting embodiment as a crucial aspect of intelligence. Likewise, emergence, particularly in forms of social coordination and cooperation, is integral. This evolutionary perspective implies that physical embodiment and emergent social dynamics are not just byproducts but essential components of what constitutes intelligence.

If we want truly novel intelligence, AI design has to transcend the handful of universal design principles operating in nature, notably the constraints of computational scale and speed. The burgeoning field of synthetic biology is already unveiling novel “life forms”, which weirdly shows the untapped potential of artificial intelligence. Additionally, the concept of brain-computer interfaces and enhancing human cognitive abilities, provides a glimpse into the possibilities of supercharged human intelligence. This suggests that the next frontier in AI might not just be about augmenting existing intelligence, but creating entirely new forms of it.

Leveraging AI to navigate the vast landscape of morphological possibilities allows us to step away from the notion of mimicking biological intelligence. The key to transformative innovation in AI lies in shedding biological limitations and embracing new constraints. These could be societal decisions on the extent of human activities replaced by machines, ethical choices about not inducing consciousness in machines, or other values-driven limitations. For instance, envision a scenario where thousands of tiny, self-organizing robots perpetually clean your apartment floor. It begs the question: is living in a spotlessly clean space, buzzing with robot activity, more desirable than accepting a bit of dust and dirt?

The idea of robots tirelessly keeping our floors spotless nudges us to think bigger about AI. It's not just about the efficiency they bring but is a reminder that we face a future where AI is simply everywhere, whether we like it or not. It might seem bizarre but perhaps the leap from clean floors to redefining intelligence isn't such a leap after all because it shows just how much we will need to make choices about what kind of intelligence we want to be surrounded by, for what purpose and at what cost.

AI will force us to broaden our view of intelligence. Moving past the idea of simply copying human intelligence, we're stepping into a world where AI could show us entirely new ways of thinking and solving problems. The real success in AI development will be in discovering forms of intelligence that go beyond anything we've known, transforming how we understand and interact with the world around us.

Full Citation: Solé R, Seoane LF. Evolution of Brains and Computers: The Roads Not Taken. Entropy (Basel). 2022 May 9;24(5):665. doi: 10.3390/e24050665. PMID: 35626550; PMCID: PMC9141356.