Machine Learning and Gene Editing at the Helm of a Societal Evolution

As the impact of machine learning (ML) and gene editing (GE) expands, forward looking policy is needed to mitigate risks and leverage opportunities. The two technologies have increasing significance, the complexity of which magnifies when they integrate. Consideration of technology advancements and policies in different geographic regions, and involvement of multiple organisations, further confound this complexity. Thus, this study explores the technological and policy implications of the intersection of ML and GE, with a focus on the United States (US), the United Kingdom (UK), China and the European Union (EU). Analysis of technical and policy developments over time and an assessment of their current state have informed policy recommendations that can help manage beneficial use of technologies and their convergence. The proposed approach can be applied to a variety of technologies and sectors. This study is intended for policymakers to prompt reflection and consideration of how to approach the convergence of the two technologies most effectively. Technical experts and practitioners may also find it valuable as a resource when considering the type of information and policy stakeholders to engage with on technological development.

Recommendations: Implement Nimble Policy, Focus on Data, and Incentivise Collaboration

• Policymakers should analyse the trajectory of both policy and technology development concurrently in multiple countries, to foster better understanding and planning of international cooperation and/or competition.
• To accommodate the fast pace of technology advances and the uncertainty with international relationships, policy must be anticipatory, participatory, and nimble and follow a policy lifecycle, oscillating between policy approaches, to mirror technology maturity levels.
• State governments and scientific communities should incentivise international collaboration and coordination by publicising potential national/international stakeholders; leveraging existing international brokers, which have a history of independently setting policy where there previously has been none; and encouraging the technical community to communicate more frequently to non-technical audiences.
• Governments and international brokers should develop and use international standards to foster international agreements.
• National policymakers should create frameworks and opportunities to support more public education and deliberative dialogue.
• Governments should develop centralised workforce development plans that target the interface of ML and GE and all levels of education.
• Governments and national policymakers should adopt both upstream (prior to the application of the technology) and downstream (pertaining to applications) regulation.
• Policymakers should focus on regulating the accessibility and distribution of underlying data.
• Governments should establish a knowledge bank about biosecurity measures, technology standards and frameworks.

Motivation: Integrated Technologies Require Proactive Management to Leverage Benefits and Mitigate Risks

The integration of artificial intelligence (AI) and biotechnology, while in its infancy, presents significant opportunities and risks, and proactive policy is needed to manage these emerging technologies. While AI continues to have significant and broad impact, its relevance and complexity magnify when integrated with other emerging technologies. The confluence of AI with GE in particular can foster substantial benefits as well as daunting risks that range from lack of ethical considerations to national security. These complex technologies have implications for multiple sectors, ranging from agriculture and medicine to economic competition and national security. And this complexity expands with the number of organisations, government departments and countries involved in collaboration and/or competition.

Application of ML (as one aspect of AI) to GE and its underpinning bioinformatics platforms will catapult the revolutionary potential of GE from “hypothetical” to imminent. This poses specific risks like potential weaponisation and bioterrorism and opportunities like improved health and wellbeing. Given the pace of technology advancement and convergence, there is an impetus to track and assess advanced technologies while increasing the focus on policy development and societal debate. This combined field has not yet been adequately studied from a policy perspective.

It is critical for policymakers to take stock of advancements and assess where the combined technology could progress. Furthermore, the public requires improved understanding of the state of the art of ML and GE capabilities to comprehend societal implications and to contribute to policy discussions. However, the policy frameworks and parameters that exist today may no longer be fit for purpose.

Approach: Literature Review and Historical Analysis to Inform a Table-Top Game

Our study entailed a landscape analysis which led to a futures assessment to identify prevalent risks and opportunities. We explored the current state of ML and GE technologies and policies, used historical analysis to project potential future risks and opportunities, and surfaced risks and opportunities of technology alongside potential policy interventions with a future focussed table-top exercise.

The landscape assessment consisted of software-assisted horizon scanning to summarise the state of the art of ML and GE capabilities as well as the integrated applications of these capabilities. We categorised these capabilities and applications based on their technology readiness levels (TRLs), their potential impact and the current barriers to further progress. We complemented this analysis with a desk-based review of the key policies that predate and/or follow GE and ML advancement, to assess their interplay and connectedness. We supplemented this assessment with interviews of subject matter experts on risks and opportunities. This analysis resulted in timelines of primary policy and technical developments across the United States, the United Kingdom, China and the European Union. These timelines were in turn used to extract past trends, extrapolate potential future trends, and compare policies and technologies between regions.

The futures assessment built on outputs from the landscape assessment and provided a deeper analysis of international relationships and more extensive policy actions. This led to the identification of primary drivers of change with regards to the convergence of these technologies, based on proposed future scenarios looking towards 2045. The scenarios were used in a discursive seminar game to develop potential policy actions to minimise harm and maximise opportunities across the United States, the United Kingdom, China and the European Union.

Results: Significant Advances with a Need to Manage the Convergence of Technologies and Assimilate Cultures

• ML is accelerating advances in biology, primarily by enabling faster processes with efficiencies.
• The integration of GE and ML has substantial practical implications, but much of the underlying technology still requires development.
• One of the most significant risks with these technologies is their dual-use nature—the capability for improving lives while simultaneously being used to create bioweapons, deadly compounds, malware and misinformation.
• Technology is advancing faster than associated policies, with little to no policy development at the intersection of ML and GE.
• Technology and policy developments are often interconnected across the global stage, highlighting the need for international policies and for supranational organisations.
• There are significant differences in the progress of technology and policy development for AI and GE. The domino effect of national AI plans across the international stage highlights the reactionary nature of recent policy actions regarding AI, which are underpinned by geopolitics rather than technological progress. Alternatively, GE involves constant iteration of technology and policy development adopting the precautionary approach. Furthermore, while key GE milestones in policy spread out over time and focused on regulation, AI and ML landmark policies are concentrated in a few clusters with past policies focused on innovation and current topical activity focused on regulation.
• ML and GE are set to revolutionise multiple sectors, but public engagement and perception are crucial to consider in future policymaking.
• International brokers can help fill a vacuum of agile and responsive policymakers.
• The culture gap between the ML and GE communities must be bridged to enact policies that address both communities and their concerns.
• Education and engagement of both the public and policymakers are crucial to policymaking but must be undertaken with a focus on the applications rather than on debating the technical aspects.
• Managing access to data could be central to effective policy development but related political and ethical issues must be addressed.
• An approach to policymaking is needed that is both reactive to unanticipated changes, and proactive with respect to anticipated risks and benefits.

The figure shows two overlapped circles.

The left circle is labeled Gene editing technologies. It contains the following items: Embryo model system, Precision editing, Knowledge bases and 'omics' libraries, and Genome-Wide Association Studies.

The right circle is labeled Artificial intelligence technologies. It contains the following items: Deep learning models, Large language models, and Artificial general intelligence.

The overlapping area in the middle contains the following items: Genome engineering, Predictive genome-phenome, and Protein folding and modeling. An area below the circles that expands from the center overlapping area is labeled Unlocked capabilities and contains the following items (each with an illustrative icon): Predictive power (with a silhouette of a head that contains a gear that is sorting items), Understanding of molecular models and systems (with a magnifying glass), Amplification of scale and increased rapidity (with an unlabeled speedometer with the pointer in the upper range), Improved data utility (with an upwards graph in a gear), and Improved targeting (with an arrow in the center of a target).