We present a global Artificial Intelligence (AI) conceptual framework,operationalization, and forecast to the year 2100. A series of AIindices were developed within the International Futures (IFs) integratedassessment platform, a quantitative macro-level system that producesdynamic forecasts for 186 countries. IFs models extensivelyinterconnected aspects of global human development, including:agriculture, economics, demographics, energy, infrastructure,environment, water, governance, health, education, finance, technology,and international politics. We conceptualize AI in three categories:narrow AI, general artificial intelligence (AGI), and superintelligence.Today’s AI consists of six basic and narrow AI technologies: computervision, machine learning, natural language processing, the Internet ofThings (IoT), robotics, and reasoning. As an index score for allapproaches 10, we forecast AGI technology to become available,representing it with a machine IQ index score, roughly analogous tohuman IQ scores. The emergence of AGI is constrained by the rate ofimprovement in and development of machine reasoning and associatedtechnologies. When machine IQ scores approach superhuman levels, weforecast the emergence of superintelligent AI. The current path forecastestimates that AGI could appear between 2040 and 2050. SuperintelligentAI is forecast to be developed close to the end of the current century.We frame the current path with faster and slower scenarios ofdevelopment and facilitate analysis of alternative scenarios. Futurework can assess the complex impacts of AI development on human society,including economic productivity, labor, international trade, and energysystems.
Natural language processing is initialized at index score 2 in 2015. Fully automated machine transcription and translation remains a distant dream. Language is often considered the defining frontier of human intelligence. The Winograd Schema challenge, designed specifically to test how well machines understand and interpret language, was first held in 2016. The best entry scored a 58%, a result described as a bit better than random (Ackerman, 2016). According to some, machine transcription, translation, or language generation will never replace the benefits derived from understanding language and human-led translation. When people learn new words and phrases, they are not just learning the literal semantics or syntax of the individual words, they also learn cultural values and norms (Lewis-kraus, 2016).
id: 736ab7d3ee67c169053973dc52be11e2 - page: 19
A score of 10 along the natural language processing index represents machines capable of fully automated transcription and translation with close to 95% accuracy (roughly human level). A score of 10 represents machines capable of hearing, understanding, synthesizing, and generating language to participate in complex conversations on a variety of topics for which it has not necessarily been trained. 5.7. IoT 5.7.1. AITASK IoT 2015 Index Score: 2 The growth of the IoT has been fueled by rising internet connectivity and mobile technology penetration. Smart phones in particular are essential, as a service delivery and data collection mechanism and will remain one of the primary interfaces through which users interact with the IoT. The IoT has been and is forecast to continue growing exponentially, by some estimates there could be as many as 50 billion devices connected to the IoT by around 2020 (Howard, 2015; Figure 5).
id: 17bd21786ccc1ff9f3963f3fcc30a021 - page: 19
Despite the sheer growth in the number of devices connected to the IoT, the technology is still very much in its infancy. The rules and norms that govern the use of and privacy around IoT-generated data remain ill-defined and opaque. Maximizing the benefits of IoT data requires interoperability between different IoT systems, today the vast majority of these systems are not interoperable. Finally, most data generated by the IoT today is used for basic tasks like anomaly detection and control, rather than for service optimization or predictive analytics, its most useful function (Manyika et al., 2015).
id: ff7c55508c4e0e65b66b47b2a1aba3d0 - page: 19
For these reasons, the IoT index is initialized at 2 in 2015, but is forecast to grow rapidly given the anticipated exponential growth in the number of connected devices. An index score of 10 represents a world where IoT data is protected and privacy concerns assuaged. Data produced is harnessed and analyzed to maximize efficiency on a broad social level. Fully smart cities and smart homes are the norm in most major developed urban areas. Automated transportation has become widespread not only as a result of autonomous vehicles, but also because cities are investing in the sensors and technology needed to produce the smart infrastructure that supports automated Andrew C. Scott et al. / Int.Artif.Intell.&Mach.Learn. 2(1) (2022) 1-37
id: 4add2ecd5740c29a4eb577230ec08ff7 - page: 19