Four Time Scales for Technology Development and Deployment Identified

Introduction to Technology Time Scales

The author identifies four distinct time scales that influence the development and deployment of new technologies. A common pitfall is misinterpreting these scales, which often results in incorrect and potentially damaging predictions about when a technology will achieve its full potential.

New Research Ideas: A Long Incubation

The first time scale involves the incubation of new research ideas, which typically requires 10 to 20 years to progress from initial concepts to robust laboratory demonstrations. Some foundational steps can take even longer, with many false starts and difficult challenges requiring decades to overcome. Once a technology is established in the lab, a 'gold rush' phase often ensues, characterized by rapid, incremental advancements.

The AI Development Timeline

The evolution of artificial intelligence serves as a key example. The first computational models of neurons appeared in 1943, but it took until 1960 for a dominant variety of linear threshold neurons to emerge. Further work led to convolutional networks and backpropagation, and then another two decades until 2012 when 'deep' learning structures allowed trained neural networks to surpass conventional vision algorithms. Large Language Models (LLMs) followed another decade later, marking a 60-year journey from initial research to widespread impact, despite being declared unfeasible multiple times.

Hype Generation and Its Cycles

The second time scale is characterized by intense hype cycles. During these periods, a technology rapidly transitions from being known only to a small group of specialists to appearing daily in the business press. Companies and researchers often re-market their existing work to align with the current trend. An example is the rapid rise of 'AI agents,' which quickly became a prominent advertising theme in San Francisco, only to be replaced by new hype as cycles progress.