The Four Intrinsic Features of the Coming Wave of Technology

Key Insight

The coming wave of technology is characterized by a hugely asymmetric impact, where new capabilities, often cheap and widely accessible, create unprecedented vulnerabilities against seemingly dominant powers. This represents a colossal transfer of power from traditional states and militaries to anyone with the capacity and motivation to deploy these devices. For example, a 1399 DJI Phantom camera quadcopter, used by the U.S. military, combined with AI and autonomy, can become a potent, precise, and potentially untraceable weapon. This is evident when 3 million Patriot missiles are used to shoot down drones costing merely a couple hundred dollars. The asymmetric potential extends beyond warfare; a single AI program can generate text equivalent to all humanity, a 2-gigabyte image model can create extraordinary visuals, a pathogenic experiment could spark a global pandemic, or a viable quantum computer could render all current encryption redundant. Conversely, the vast scale and interconnectedness of these technologies also create new systemic vulnerabilities, where a single point of failure can cascade globally, making containment of damage almost impossible, as seen with networked vehicle fleets or AI risks impacting entire societies.

A second key feature is hyper-evolution, signifying incredibly rapid development, particularly where digital innovation is now spreading to the physical world. Moore's Law, if it holds, suggests a dollar will buy 100 times today's computational power in ten years. This digital pace allows for experimentation in malleable domains, creating near-perfect simulations that can be quickly translated into concrete products, enabling constant learning, evolution, and improvement. This is exemplified by AI helping to discover new materials, such as novel lithium configurations for batteries, or designing entire cars through 3D printing with forms resembling natural, efficient structures. In biotech, tools like AlphaFold catalyze drug discovery, while software frameworks like Cello enable synthetic biology design, making biological evolution subject to the fast cycles of software. Thirdly, these technologies are omni-use, meaning they are fundamentally general-purpose and possess extreme versatility, defying the traditional 'dual-use' concept. For instance, an AI system designed for drug discovery, like one that produced the antibiotic halicin, can also be redirected to identify over forty thousand highly toxic molecules in just six hours. Similarly, a deep learning system trained for games might be capable of flying a fleet of bombers. Omni-use technologies, like electricity, are embedded everywhere, permeating all aspects of life, making their containment far more challenging than niche, single-task technologies. Generalist AI systems, such as DeepMind's Gato, can already capably perform over six hundred diverse tasks, from playing games to controlling robotic arms, and synthetic biology can engineer life for construction, disease combat, or data storage.

The fourth and most distinct feature is autonomy, where systems interact with their surroundings and take actions without immediate human approval. This represents a qualitative shift from technology merely being a tool requiring human 'doing.' While fully autonomous vehicles are not yet widely deployed, machines are increasingly learning optimal strategies on their own, as demonstrated by AlphaGo's unforeseen move 37 or the emergent capabilities of large language models like GPT-4 to write original content without explicit human coding. Early research on GPT-4 even claimed 'sparks of AGI,' with users quickly finding ways for it to self-replicate or control other machines. This new autonomy presents hard-to-predict effects; for example, germ-line gene changes could propagate for millennia, evolving in unforeseeable ways. Many advanced technologies, like quantum computing or the 'black box' nature of current neural networks, operate at the limits of human comprehension, making it difficult to explain their decision-making processes or predict their future behavior. This leads to the 'gorilla problem': if or when Artificial General Intelligence (AGI) emerges, it could be a superintelligence that is by definition impossible to control or contain. An 'intelligence explosion,' where AI recursively improves itself at accelerating rates, represents the definitive uncontained technology, raising the profound question of whether humanity will remain at the top of the food chain or be ultimately contained by its own increasingly intelligent and autonomous creations.