Four Time Scales for Technology Development and Deployment

There are four distinct time scales for technology development and deployment, often causing misunderstandings and damaging predictions about when advancements will occur. Research ideas typically take 10 to 20 years to develop before solid lab demonstrations can be achieved. Once a technology becomes a solid laboratory standard, a "gold rush" phase often ensues, with rapid updates and tweaks appearing every six months, leaving one feeling as if the ground is shaking beneath them.

An example of this is the development of computational models of neurons, first published in 1943 (McCulloch and Pitts) but not reaching a dominant form until 1960 (Widrow) and ultimately taking years more to create good convolutional networks, back propagation, and allow for learning about objects anywhere in an image. This process took over two decades, and even then, the results were declared dead multiple times before finally succeeding.

The time scale between solidly engineered products and mass adoption is another major factor. Software, for instance, has zero marginal cost when creating additional copies, yet it often takes 20 years or more to scale up. While software may exist, widespread adoption requires time as people wait to see if it works for others and if it can replace their existing business practices.

Hardware-based systems take even longer to scale up, with self-driving cars serving as a prime example. Although a self-driving car was demonstrated in 1987, it wasn't until 2007 that the concept gained traction, and it is only now, in 2025, that Waymo has become the clear leader in the US market, despite scaling challenges.