Defending democracy against downside artificial intelligence risks*
ROBERT MICHAELS CEO, consulting risk assessment/risk management scientist, RAM TRAC Corporation, Schenectady, New York, USA
In “Managing extreme AI risks amid rapid progress,” Bengio et al. (1; Science, 24 May) highlight downside artificial intelligence risks that must be managed effectively and urgently if posited AI benefits are to be realized safely, especially given rapid progress in AI development. In our current state of inadequate preparedness, they warn, malicious actors could deliberately embed undesirable goals; even well-meaning
AI developers inadvertently might embed unintended goals. We therefore must manage risks potentially posed by intended and by unintended consequences of AI development and use.
This accurately conveys the broad AI downside risk management problem. Its proposed solutions, by comparison, are narrower and, I think, inadequately responsive to the nature, scale, and immediacy of the problem. In particular, Bengio et al. primarily address AI developers and users who reasonably might be expected to
comply with protective legal and regulatory risk management controls. In contrast, AI risks needing management most urgently would seem to involve malefactors domestically and abroad who might be expected to flout such controls, currently with impunity.
Warning that “large-scale cybercrime, social manipulation, and other harms could escalate rapidly,” Bengio et al. propose “reorienting technical R&D” and
enacting “government measures.” Further R&D, they reasonably suggest, is needed to create safe AI products. True, but R&D already has created unsafe AI products, such as facial recognition and deepfake software by China. Proposing further R&D is great, to the point of becoming a mantra, but closing a barn door after its horses already have escaped is not great.
A global suite of actions to manage downside AI risks is necessary as Bengio, et al. laudably propose, but nonetheless insufficient as long as their focus remains limited
to AI development and user communities. Missing but essential is a broader focus, specifically including potential victims of AI abuse, that is, all of us. What is missing is a proposal for immediate and massive expansion of civics education, including
education to better prepare people to function defensively and effectively in the nowpervasive digital social media world.
AI developments in China cited above exemplify our vulnerability, as deepfakes appear poised to be deployed toward interfering with elections and undermining democracy in Hong Kong (3), and facial recognition software already has been deployed to oppress the Uyghur population in China (2). Malefactors exist globally,
at the national, subnational, and even at the individual level. Likewise, their reach is global, as they often employ international social media platforms. Civics education urgently must warn and defend by establishing a robust civics education presence in every nook and cranny of this brave new digital world.
Civics lessons must be communicated by trusted messengers, including messengers matched to specific local audiences. For example, clinicians may be
enlisted to address disinformation about politically sensitive vaccine issues to their patients, such as via online patient portals. Ultimately, civics outreach urgently must combat facile complacency and credulity, for example by instilling healthy skepticism of ‘conspiracy theories’ and other assertions that are unsupported by credible evidence. Thus civics lessons must outline civic responsibilities for adoption as prerequisites to belief formation and action taking, including voting.
__________________ *Dr. Michaels is a risk assessment/risk management scientist, and author of “Civics and Science: Contemporary Issues for Civil Democracy”
Suggested citation: Michaels, Robert A. Defending democracy against downside artificial intelligence
risks. Response to Bengio et al., Science, 384(6698):842-5, 24 May 2024, eLetters, online 18 June 2024. __________________
References and Notes
1. Bengio et al. Managing extreme AI risks amid rapid progress. Science,
384(6698):842-5, 24 May 2024.
2. Feldstein, Steven. China’s high-tech surveillance drives oppression of Uyghurs. Bulletin of the Atomic Scientists, 27 October 2022.
3. Lee, Sze-Fung. Deepfakes with Chinese characteristics: PRC influence operations in 2024. Jamestown Foundation, China Brief, 24(7): 29 March 2024.
Publication citation
Michaels, RA. Defending democracy against downside artificial intelligence risks. Response to Bengio et al., Science, 384(6698):842-5, 24 May 2024, eLetters, https://www.researchgate.net/publication/381698382_Michaels_Robert_A_Defending_democracy_against_downside_artificial_intelligence_risks_Science_eLetter_18_June_2024, online 18 June 2024;