Daniel Silveyra dot com

Since the deployment of the first impressive large language models, people have been flagging different ideas about how to best deal with the impact of accelerating technological change, specifically the massive loss of employment as AI and robotics improve.

A big question looming over all of this is who will get to decide how this tech gets deployed, which I think is really a question about the balance of state vs market approaches in light of a new technology. I think this most hinges on what AI does to economies of scale in production and other sources of market power, as well as national security and arms-race considerations. We don’t know what the answer is yet, but the weight of evidence seems to point towards concentration. I’d argue that even without AI, current levels of power concentration in the West are politically destabilizing and need to be addressed.

I’m more interested in the conversation about redistributing the gains because more specific things can be said here. Generally, I am for redistributing the gains from AI and other innovations to a broader base of society both for utilitarian reasons (the resources will create more wellbeing this way because of diminishing marginal returns to wealth) and political reasons (egalitarian values already held in many societies, compensating society for the information that is powering the technology, broadly shared progress fosters political stability). 

I’m concerned that redistributing gains through creating a new entitlement will solve the income inequality problem created by new technology without addressing the underlying power inequality problem. 

One issue is how secure the redistribution would be. If some later administration wants to remove the entitlement, it may do so and those affected only have their vote and ability to sue to exert influence. If times are bad, it is more likely that such an administration will exist and that elections and judicial systems will be less than perfect – see Weimar Germany and the introduction of Greek austerity in the 2010s. Admittedly, democracy would have to get very, very bad for this to matter: universal entitlements create powerful self-defending constituencies and are extremely hard to cut in democracies (see pension schemes, etc.). 

My main concern is that only giving people an entitlement – be it through monthly payments or through an equity share in some national trust – would not reverse the hollowing of social relationships coming from a small group of people owning most of the productive capacity in society. If control over how most citizens access to resources and services can be easily centralized then it can be easily captured and defended by a small set of people, leading to tyranny. We need to continue encouraging people to be involved in the creation of value for other people – yes, by giving them ownership in the capital, but also by ensuring that this ownership comes with the ability to influence how the capital is governed and the ability to continue creating relationships of value exchange. That means introducing different kinds of ownership structures (e.g. a-la worker’s cooperatives), investing in education so that technical knowledge is more broadly held, and encouraging more participation in commercial and civil society.

With sufficiently powerful AI models, and depending on the activity, it is easy to imagine companies without any staff outcompeting others. Embedding voices with voting rights – speaking on behalf of both the original owners of capital and on behalf of broader society – in all organisations might slow production in some instances, but it would ensure that the ability to meaningfully impact production decisions (e.g. major restructurings, plant closures, working conditions) is more broadly distributed even in an age where the tech invites centralization and scale. As with any other democratic mechanism, the quality of judgment brought by this increased participation will depend on increased public investment in education – we cannot afford a society of passive users who view the increasingly sophisticated technology as magic.

Who exactly gets to have a vote in which company could be determined by a qualified lottery such as how jury duty is handled, though this would have to be run by some centralized body (likely the state) who would also be at risk of capture if democratic institutions broke down. Being assigned as a “citizen shareholder” to different companies by lottery would inevitably create winners and losers; all citizens would have to be guaranteed a minimum return on their shares but would otherwise be incentivized to help the firm they’re a part of succeed. There might also be decentralized governance options that work here, though they would need to be tested since what we’ve seen of Decentralized Autonomous Organisations leaves a lot to be desired. In both cases, the dispersion of voting rights across every organisation would act as a counterbalance to centralization and tyranny. 

It’s clear that AI will mean that humans should further lean into our comparative advantage – activities which are improved by or necessarily involve human interaction. Yes, this depends on AI not displacing human interaction itself, but I strongly suspect people will want to pay for more human-based services as they get wealthier through price deflation of anything that can be mass-produced. This should lead to an explosion of new services, experiences, arts, and crafts that we offer each other. We can encourage this renaissance through policies that encourage people to participate such as subsidized education, coaching and training, and the creation of businesses and non-profit organisations. This can include a large increase in the number of non-profit grants and subsidized small business loans offered by a diverse set of mechanisms. 

These policy ideas are all based on the observation that it is vital that every person continues, whenever possible, to figure ways to be helpful and be recognized as useful by others, to form alliances anchored by mutual benefit. We want to redistribute the wealth from large gains in productivity not just by sharing cash but by encouraging people to gain skills and establish relationships with others. The more things we need from each other, the more bound to each other we will be. If we need services and goods and art and time and attention from each other, we are more likely to interact and see our commonalities, more likely to find conflict between us wasteful, and more likely to defend our common interests. The network is more resilient if the nodes are interconnected in multiple ways rather than individually connected to one main thread.