AI solves every existential risk—larger capacity for problem-solving

It is suggested that AGI systems could revolutionize humanity. Projected benefits include curing disease, revolutionizing the nature of work, and solving complex environmental issues such as food security, oceanic degradation, and even global warming. In prospect, the effect on humankind promises to be even greater than both the industrial and digital revolutions combined. However, it is widely acknowledged that failure to implement appropriate controls and constraints could lead to catastrophic consequences (Amodei et al., 2016; Bostrom, 2014; 2017; Brundage et al., 2018; Omohundro, 2014; Steinhardt, 2015). It has been argued, for example, that untrammeled and uncontrolled AGI could even pose an existential threat to humanity (Bostrom, 2014; Hancock, 2017).

As the discipline that is focussed on enhancing human wellbeing, HFE clearly has an important and even determining role in the design, implementation, and operation of AGI systems. Despite this, there has been little discussion as to how HFE can and should contribute. This is reflected in a disturbing lacuna of HFE work in this area. Also, given the fact that questions are being raised regarding the suitability of HFE methods for today's complex systems (e.g., Salmon et al., 2017; Walker et al., 2017), it is important to question whether HFE is sufficiently equipped to contribute effectively to the design of systems that are first-of-their-kind, and necessarily nonhuman in nature. Context specific and context relevant theoretical and methodological development may be required for the HFE toolkit to be suitable for such applications.

In this article, we offer an agenda for HFE and its purposed-directed impacts on AGI. We discuss the role that HFE must adopt to ensure that the far-reaching benefits of AGI are realized without problematic threat to society. We seek to achieve this by examining current state-of-the-art HFE methods, and distinguishing their potential in the design, implementation, and operation of a prospective AGI system, as recently described by Tegmark (2017). This “ethnographic science fiction” approach is required as AGI systems do not yet exist, but the potential benefits and risks are so significant that work is required immediately. Further, such an approach is an acknowledged avenue for discussing future global issues where uncertainty exists (e.g., Raven, 2017). This study therefore acts to set a HFE agenda framed within in an “envisioned world” perspective. In doing so, we identify key areas where developments and extensions to HFE methods are required. We articulate a research agenda which describes the work required to situate HFE within wider multi-disciplinary efforts aimed at creating safe, efficient, effective, and controllable AGI systems.

2 UNDERSTANDING AGI

The term “Artificial Intelligence” was first coined in the middle of the 1950s by John McCarthy, an American scientist working at Dartmouth College. The formal field of AI was established soon after. Hard upon the intervening decades of research and development, ANI systems are now well established. Such systems possess intelligence in relation to specific tasks and remain constrained to their particular domain of operation. Widely known examples include Facebook's facial recognition system, Apple's personal assistant “Siri,” and Tesla's self-driving vehicles (Kaplan & Haenlein, 2018). In contrast, AGI systems will almost certainly be more broadly focussed and will equal or exceed human intelligence in wide swathes of cognitive capacities (Everitt et al., 2018; Gurkaynak et al., 2016). AGIs are expected to be able to plan, reason, make decisions and solve problems autonomously; even for tasks that they were not initially designed to address (Kaplan & Haenlein, 2018). A summary of the key differences between ANI and AGI systems is presented in Table 1.

[table 1 omitted]

2.1 The benefits and risks associated with AGI

AGI is a dual use technology in that it will be used both for good and bad. First and foremost, if AGI realizes its potential and surpasses human intelligence, there is no doubt that it could bring significant benefits to humanity (Bostrom, 2014; Yudkowsky, 2008; 2012). Postulated benefits relate mainly to systems which exceed human intelligence and develop a capacity to respond to the panoply of issues that threaten either human health and wellbeing, the earth, or our future existence globally. These include climate change and environmental degradation, overpopulation, pandemics, food and water security, misuse of the internet and social media, terrorism, cybercrime, nuclear warfare, inequality, antimicrobial resistance, and instability in the world's economy. In addition, it has also been suggested that AGI will help with the onslaught of forthcoming new and emergent issues such as automation replacing human work, the genetic modification of humans, an ageing population, and otherworld settling (FLI, 2018).