Category: Humanity

Potential Source of Harm: Humans Merge with AI

Updated January 18, 2024

 

Written in collaboration with Tony Czarnecki, Managing Partner of Sustensis 

 

Nature of Harm

This potential harm is associated with the possibility that humans will merge with AI – physically, mentally, behaviorally or otherwise. This is related to the risk of AI Domination as well as work on transhumanism.

 

It is difficult to predict how humans will merge with AI, because we are in very early days of this process. However, the process has already begun at least from a behavioral perspective. For example, a substantial and increasing number of humans are so reliant on AI tools in digital applications (often delivered via mobile devices) that it would be reasonable to treat such individuals as partially merged with AI. Work on more substantial integration of humans with AI and other digital tools is underway, for example:

  • China’s AI-Brain Project is directly focused on merger of humans with AI

  • Neuralink is building implantable brain-computer interfaces, which is entering a first clinical trial aimed at people with quadriplegia and amyotrphic lateral sclerosis.

 

There are significant potential benefits associated with merger of humans and AI, including medical benefits such as those initially pursued by Neuralink, more effective human operation of digital technologies, and many others; and as in other areas, we expect such that benefits will exceed harms. However, potential harms may range from damages from malfunctioning, infection or rejection of implants to assisting with AI domination of humans.

 

Regulatory and Governance Solutions

Existing regulatory and governance solutions for the risks associated with humans merging with AI are currently mostly not distinct from those associated with general regulation of AI (see Regulatory & Governance Solutions page). There are however specialized areas of regulation that are relevant to this risk – notably medical devices regulation as applied to solutions like Neuralink.

 

Ultimately, as AI solutions increasingly permeate society, more extensive regulation is likely to be needed. Initial explorations of such controls have been explored in books like:

 

Technical Solutions

For existing technical projects relating to human merger with AI, safety considerations are generally part of the project architecture. For example, the first policy recommendation in the discussion of the China AI-Brain project linked above is “Pay greater attention to AI safety”. Similarly, safety requirements are central to the regulatory approval process for Neuralink, although there has been controversy around Neuralink’s compliance with these requirements.

 

As for regulatory and governance solutions, more will be required over time. This is discussed in the books linked above, and in more technically-focused books such as:

 

Government and Private Entities

The entities taking responsibility for addressing this risk are mostly the same ones addressing general regulation of AI, as summarized on our Entities & Funding and Regulatory & Governance Solutions pages.

 

Medical device regulatory agencies such as the US Food and Drug Administration and the European Medicines Agency are developing programs specific to AI devices.