It concentrates power in the hands of whoever owns the systems

The argument

The case against begins with ownership. Opponents argue that building a capable system requires computing resources, data and capital on a scale available only to a few firms and states, and that this is not an incidental feature of the present moment but a structural property of the technology. Where earlier tools could be made and improved by many hands, these cannot, so the capability accrues to whoever can afford to build it. The mechanism opponents describe is leverage. A technology that substitutes for human cognitive labour transfers the value of that labour from the many who performed it to the few who own the substitute, and does so across every sector at once rather than one industry at a time. Opponents argue that this is a transfer of bargaining power as much as of income: a workforce whose skills can be replicated loses the leverage that made wages and conditions negotiable, and a public whose information, hiring, credit and policing decisions run through systems it cannot inspect loses the leverage that made institutions answerable. Opponents argue that the political consequence is the serious one. Concentrated capability of this kind gives its holders the means to shape what people see, to model and anticipate public behaviour, and to supply the state with instruments of surveillance and persuasion that no previous government could afford. Proponents answer that this is a matter of regulation, and opponents reply that regulators are themselves dependent on the firms for expertise, access and personnel. On this account the technology is not good for humanity, because its direct effect is to move power away from most of humanity and towards a very small part of it.

Premises

Counter-arguments

Proponents reply that the premises describe a market structure rather than a technology, and that market structures change. Capability has repeatedly become cheaper and more widely available as methods mature and open alternatives appear, and competition law, procurement rules and public research exist precisely to address concentration. They add that the argument proves too much: electricity, telecommunications and pharmaceuticals all began as the property of a few and are not now judged bad for humanity, and that the response to concentrated ownership of something valuable is to contest the ownership rather than to wish the thing away.

Rejecting the premises

[Rejecting P1] Capability has repeatedly become cheaper and more widely reproducible as methods mature, so the concentration describes a market stage rather than a structural property. [Rejecting P2] Substitution for labour has historically shifted employment between sectors rather than eliminating bargaining power, and the premise assumes the terminal case. [Rejecting P3] The means described are the subject of existing surveillance and competition law, so the premise identifies a regulatory failure rather than a property of the technology. [Rejecting C] The premises indict the ownership of the technology, which is consistent with the technology itself being good for humanity under different ownership.