In a world captivated by AI hype, Kevin Kelly's recent reflections cut through the noise with a sobering truth: artificial intelligence is not a single, monolithic entity hurtling toward dominance, but a sprawling galaxy of diverse, alien minds we're only beginning to map. Originally penned for Wired Mid-East and still resonant today, Kelly's memo dismantles common misconceptions, urging technologists to focus on three critical insights that redefine our approach to AI development and integration.

Article illustration 1

AI Is a Spectrum of Alien Intelligences, Not a Singular Force

Kelly stresses that we must discuss AIs in the plural. There is no 'The AI'—instead, we have a burgeoning ecosystem of models, each with unique traits. Large language models (LLMs) vary based on training data, yielding divergent responses, while non-LLM AIs, like those in autonomous vehicles, serve entirely different functions. 'Our own brains are a society of different kinds of cognition,' Kelly notes, highlighting that future AIs will be complex systems integrating dozens of specialized intelligences, from memory to pattern recognition. Crucially, none will think like humans; they are inherently alien, solving problems in ways we can't anticipate. This diversity isn't a flaw—it's AI's superpower, enabling solutions to 'wicked problems' that stump human minds alone. Intelligence, Kelly argues, isn't a ladder but a multidimensional 'radiation,' with human cognition occupying just a tiny niche in a vast galaxy of possible minds. For developers, this means building tools that leverage this heterogeneity, moving beyond one-size-fits-all frameworks to create adaptable, specialized AI ensembles.

Spatial Intelligence, Not Chatbots, Is AI's Real Game-Changer

While LLMs grab headlines for answering questions, Kelly contends that spatial intelligence—AI's ability to simulate, render, and interact with the 3D world—holds far greater transformative potential. This capability underpins everything from robotics to augmented reality (AR), requiring AIs to master real-world physics, cause-and-effect logic, and volumetric reasoning. 'Without ubiquitous spatial AI, there is no metaverse,' Kelly asserts, pointing to emerging tech like AI-generated video and 3D environments that could turn solo creators into one-person studios for games or films. Training on massive datasets, such as Tesla's billion-hour driving videos, these models learn physical laws and cascading forces, essential for humanoid robots or AR overlays via smart glasses. For engineers, this signals a shift: invest in physics-informed AI and simulation tools, as spatial prowess will drive the next wave of innovation in automation, manufacturing, and immersive tech far more than textual interactions ever could.

We Have Time: Slow Adoption Debunks Doomer Hysteria

Kelly dispels urgency around AI's rise, emphasizing that adoption is inherently slow and evidence shows minimal harm so far. Despite massive investments, only players like Nvidia and data centers see real profits, while AI service costs remain high and unsustainable for complex tasks. Organizations can't simply 'plug in' AIs; they require fundamental redesigns, akin to how electrification reshaped offices into skyscrapers. Small companies lead adoption due to agility, but full integration into society will take 5–10 years, even without further advances. Crucially, Kelly refutes 'doomer' narratives of exponential intelligence growth or mass job losses: 'The total number of people who have lost their jobs to AI as of 2024 is just several hundred employees,' he states, citing translators and help-desk roles. Intelligence gains are sluggish, hampered by poor metrics and exponentially rising training costs. Tech leaders should prioritize evidence-based policies, recognizing AI's benefits outweigh current risks, and use this window to foster ethical, adaptable frameworks for AI's alien minds.

As Kelly reminds us, today's AIs emerged from unintended surprises—like reasoning abilities in translation algorithms—underscoring that we're still uncovering intelligence's essence. The path forward isn't fear, but curiosity: embrace AI's diversity, champion spatial innovation, and build deliberately, knowing we're early explorers in a universe of possibilities.

Source: Kevin Kelly, 'Artificial Intelligences, So Far,' Wired Mid-East (2024), available at https://kk.org/thetechnium/artificial-intelligences-so-far/