INTELLIGENT MACHINE TECHNOLOGY AND PRODUCTIVITY GROWTH

Abstract

This paper provides preliminary estimates of the productivity impact of intelligent machine technology (IMT) and the rate of return to IMT research and development (R&D) over the next two decades. The paper adapts economists’ traditional productivity growth model to enable the use of industrial experts’ forecasts of a few key parameters of the model to form the estimates of productivity growth and rate of return. Respondents – from a sample of firms operating in IMT development and applications in the automotive, aerospace, and capital construction industries – anticipate that IMT will generate substantial productivity growth over the next two decades, and the estimated social rates of return to IMT R&D are substantial.

Keywords:

• Intelligent machine technology
• Productivity growth
• Research and development

JEL Classification(s) :

• O300
• O400

Acknowledgements

We thank Albert N. Link and two anonymous referees for numerous helpful comments that have been incorporated into the final version of the paper.

Notes

¹We use the non-technical, descriptive term ‘human-like’ behavior; for both technical and philosophical reasons, ‘human-level’ behavior might be more apt. IMT is computational technology that senses its environment and adjusts its behavior based on modeling of interaction with that environment and evaluation to achieve goals. It can be encapsulated in a computer program, an intelligent sensor, or a robot. Examples of IMT include machine systems such as computer-aided design technologies, computer numerically controlled machine tools, computer controlled inspection systems, enterprise integration information systems, just-in-time production scheduling and inventory control technologies, internet technologies that enable out-sourcing to the most efficient suppliers, and multi-spectral measurement systems for construction site metrology and other applications.

²See Link and Siegel (2007, p. 97) for definition and discussion of enabling technologies.

³Link and Siegel (2003, p. 63, p. 78) define infrastructure technology and explain its role in innovation.

⁴Link and Siegel (2007, p. 126) provide a window to the literature about radical technologies; IMT is arguably a special case of such dramatic departures in process technology – special because of its pervasiveness, yet radical in the potential for transforming competition in individual industries.

⁵For example, see the description of IMT increasing the productivity of German automobile manufacturer BMW AG in Gumbel (2007).

⁶IMT development entails multidisciplinary research requiring sound technology policy. IMT is an emerging system technology, both technically – requiring the integration of technologies governing, as explained by Albus and Meystel (2001, pp. 17–18), sensory processing, world modeling, value judgment, and behavior generation – and institutionally – requiring the collaboration and integration of public, quasi-public, and private sector institutions specializing in basic scientific research, generic technology development, and infratechnology development. See Albus and Meystel (2001, pp. 145–152) regarding complexity of IMT and Tassey (1997) regarding complexity of technology and demands of multidisciplinary research. See Tassey (2005) regarding interaction of institutions in a successful national innovation system.

⁷See US Domestic Policy Council (2006). The goals for the American competitiveness initiative discussed there include capability and capacity in nano-manufacturing, intelligent manufacturing capabilities, and related sensor and detection capabilities.

⁸Computer programs not only play chess, they handle airline reservations, manage financial transactions, control inventories, verify customer identifications, dispense bank drafts, and track package shipments. Respected technologists believe the realization of ‘digital people’ and ‘engineered minds’ (autonomous machines) is palpably close (Albus and Meystel, 2001, p. 146).

⁹Solow's (1957) seminal paper focused on a much more inclusive residual – namely, the residual productivity growth rate after the growth rates in capital and labor were controlled. See Link and Siegel (2003, pp. 27–28) for a discussion of Solow's residual.

¹⁰A complete list of the technology specialists with whom we consulted and their affiliations as well as the detailed future scenarios are provided in Leech et al. (2006).

¹¹We have then a very specialized portion of the residual in the original Solow (1957) formulation. Solow's residual reflected the growth rate in output not explained by the growth rates in labor and capital. That residual was classically attributed to an exogenous rate of output growth and the rate of growth in the stock of knowledge. Here we have focused on the residual attributable solely to new IMT.

¹²The 14 companies with which we discussed IMT and to which we are grateful for the insights provided by their representatives are Boeing Company, Bechtel Corporation, CH2M Hill, CNH Global, Daimler-Chrysler Company, FANUC Robotics of America, Ford Motor Company, Foster-Miller, Inc., iRobot Corporation, John Deere, Mag IAS, Okuma, Northrop Grumman Corporation, and Toyota Corporation.

¹³We chose experts and firms in industry by using patent data to identify individuals and companies with a considerable stock of applied knowledge about IMT and also by consulting with a group of distinguished technology specialists from NIST, industry associations, and academe.

¹⁴Even infrastructure technology support from government and cooperative R&D in the industry are held constant at their accustomed levels in recent years. When respondents provided their estimates that have been used with the model to make predictions about IMT-induced productivity gains and about IMT R&D rates of return, the respondents were asked to assume that industry and government activities such as cooperative R&D and government support with infrastructure technology continue in the accustomed way. Estimates about quality multiples and computational capability and so forth are provided for the upcoming two decades and productivity growth rates and rates of return on investment are then derived.

¹⁵For technical details about why we anticipate a downward bias in the estimated rate of return when, as in the present case, we are estimating a social rate of return, see Scherer (1984, p. 283, endnote 7).

¹⁶The patent proportions were provided by 1790 Analytics Inc.

¹⁷Some studies have accounted for R&D embodied in purchased inputs. A prominent example is Scherer (1984, chapter 3 and chapter 15). For a review and further development of the idea that benefits of R&D done outside the using industry affect R&D rates of return, see Scott (1993, chapter 9).

¹⁸We gain insight about the uncertainty of future advances in IMT from the range of the respondents’ responses about γ, α, and β and hence the range for predicted IMT-induced productivity growth.

¹⁹Although a number of the corporations surveyed are multinational, the survey is focused on the US; future surveys should broaden the sample.