Where does takeoff collide?

In W. W. Rostow's account of economic development, the take-off is the decisive stage at which an economy overcomes the old barriers to steady growth and enters a period of self-sustaining industrial expansion. Rostow's 1956 The Take-Off Into Self-Sustained Growth and his 1960 The Stages of Economic Growth placed it as the third of five stages, marked by a sharp rise in productive investment, one or more leading growth sectors, and institutions that let growth feed on itself. The referent is a phase in a country's long-run development, spanning decades, and it borrows the image of flight rather than describing an aircraft, a product's sales, or an AI system.[1][2]

Unit

A stage in a country's economic development, spanning roughly two to three decades

Recovery means

Not applicable

Measured by

Rostow read it from a rise in the rate of productive investment, a leading growth sector, and supporting institutions; the criteria are debated and hard to date precisely

In research on the diffusion of new products, the takeoff is the point at which sales of a really new product jump from a slow introductory trickle into rapid growth. Golder and Tellis's 1997 Will It Ever Fly? studied the histories of consumer durables and defined the takeoff as the first year in which the increase in sales over the previous year crosses a threshold, giving a datable turning point in the sales curve. The referent is a transition in a product's market, dated to a single year, and it is not an aircraft leaving a runway, a stage in a national economy, or a change in machine intelligence.[3]

Unit

A turning point in a product's sales curve, dated to the year sales accelerate

Recovery means

Not applicable

Measured by

Identified as the first year in which the proportional jump in sales over the prior year crosses a threshold that varies with the current sales base

In debates about advanced artificial intelligence, takeoff names the hypothesized transition from roughly human-level AI to systems far more capable than any human, driven by recursive self-improvement. I. J. Good's 1965 Speculations Concerning the First Ultraintelligent Machine described an intelligence explosion, in which a machine able to improve its own design would rapidly outrun its makers, and Bostrom's 2014 Superintelligence framed the open question of whether such a takeoff would be fast (a hard takeoff over hours or days) or slow (a soft takeoff over years or decades). The referent is a conjectured shift in machine capability, argued rather than observed, and it shares only the word with a runway, an economy, or a product's sales.[4][5]

Unit

A hypothesized transition in AI capability, its duration debated from hours to decades

Recovery means

Not applicable

Measured by

Not directly measured; it is argued from trends and models, and characterised as fast (hard) or slow (soft) by how quickly capability would rise once the transition began