How I expect TAI to impact developing countries
| This is a Draft Amnesty Week draft. It may not be polished, up to my usual standards, fully thought through, or fully fact-checked. |
Commenting and feedback guidelines: I would welcome any feedback, particularly from those with experience working with developing countries. |
Summary
Transformative AI’s (TAI) impact on developing countries[1] is likely to look quite different from its impact on developed countries. Labour displacement will be slower, and there will be less fiscal pressures on developing governments. But impacts will be uneven, and countries that rely heavily on service exports and remittances are likely to lose out in the near term.
Global inequality is likely to increase, but this in itself is unlikely to have a large negative impact on global human welfare
TAI could produce global public goods, which would improve global welfare even if global coordination and altruism don’t increase
Liberal democracies are likely to weaken, which could have indirect negative impacts on people already living in highly unequal societies.
Overall, I am highly uncertain whether global human welfare after TAI would be better or worse than our current world. Since the factors above point in opposite directions, the net effect on global welfare will depend on the relative strength of each factor. Generally speaking, I expect that things are likely to go better for people in developing countries if developed countries (particularly the US) manage their domestic inequality pressures well.
Assumptions and caveats
Assumptions
The timeframe is roughly the next 10-20 years.
I assume we get TAI, which may or may not involve AGI (AI capable of handling all cognitive tasks at least as well as humans). I am not assuming ASI (an AI agent or agents with intelligence far exceeding humans).
Humans have roughly as much control over AI as they do now.
There isn’t a huge uptick in global coordination or altruism where the winners from TAI voluntarily share their wealth with the rest of the world.[2]
Caveats:
I am pretty uncertain about the views expressed in this post and am far more familiar with developed OECD economies. Part of why I am publishing this now, during Draft Amnesty week, is so that others can correct any misconceptions I might have.
There may be some factual inaccuracies in this post. Please flag as you see them.
TAI’s impact on developing countries is likely to look quite different
Less labour displacement
I can think of at least four reasons why labour displacement in developing countries is likely to be slower than in advanced economies:
Services that can be digitally delivered make up a much smaller share of developing countries’ economies (e.g. law, accounting, programming, clerical work) . Developing countries’ economies tend to skew more towards agriculture and in-person services (e.g. food services, haircuts, domestic work). There are some important exceptions, such as India and the Philippines, which earn substantial export revenue from clerical, IT and call centre services. However, even in India, services that could be delivered digitally make up a relatively small proportion of overall employment.[3]
AI models have mostly been trained on data in advanced economies. They are less likely to have local knowledge relevant to developing countries. I expect AI will be able to displace US lawyers more easily than they can displace Ethiopian lawyers because much less of the relevant data will have been digitised.
Poor telecommunications infrastructure and supply chains. This is also likely to slow diffusion.
Cost savings are lower. Even where AI can replace jobs in a developing country, diffusion is likely to be slower because wages are lower, so there are fewer cost savings to be gained. So even if we saw a breakthrough in robotics such that they could perform many in-person services, it seems unlikely that robots will be cost competitive in developing countries within the next 10-20 years.
But, unlike some commentators, I don’t see slower diffusion as being a bad thing because I expect widespread labour displacement to be very disruptive. Countries that experience slower diffusion can therefore watch and learn from countries where diffusion is more rapid.
Less government fiscal pressures
Developing countries’ tax bases tend to rely less on labour income than developed countries’. This is because many developing countries lack the administrative capacity required to effectively enforce income taxes, and have large, untaxed informal sectors. Instead, governments rely more on consumption taxes, tariffs and customs duties, corporate income taxes, and natural resource revenues (where available). This should make their tax bases more resilient to widespread labour displacement.
Developing countries also tend to have much lower government spending, with very limited social safety nets. This may make life harder for workers who are displaced, but it also means developing country governments will not experience the same fiscal pressures that developed country governments will.
AI could even strengthen government finances in some developing countries, as many countries face tax capacity and administration challenges that AI might be able to help with. While the diffusion of TAI in developed countries is generally led by the private sector, I expect diffusion in lower-income countries will be stronger initially in the public sector, often with support from non-profit or international organisations.
But impacts will be uneven
Globally, we can expect to see countries’ comparative advantages shift. This process will create some winners and losers, depending on how valuable their resources are to the AI supply chain. However, there could also be global bottlenecks apart from those in the AI supply chain.
For example, how is TAI likely to affect demand for saffron, Ethiopian coffee, or New Zealand lamb? Even if TAI might be able to make cheaper synthetic substitutes, those substitutes may not weaken demand for original products if consumers prefer the originals for whatever reason (e.g. distrust of substitute, status signalling, habitual consumption, or trade protectionism). So countries and firms involved in the production of such goods might still earn reasonable returns even if they don’t have a place in the “AI supply chain”.
Speaking very broadly, I expect:
Countries with abundant water, energy sources and cool climates will benefit.[4]
Countries that currently have a large share of service exports (e.g. India) will lose out.
Countries that rely heavily on remittances (e.g. Latin America and the Pacific) will also lose out. In a way, remittances can be seen as payment for “exports” of in-person services like construction or domestic work. Even if these jobs don’t get automated away, I expect competition for them will get fiercer as people who don’t hold capital run out of other ways to make money. (This risk may be mitigated if those richer countries manage their domestic inequality pressures well.)
Global inequality is likely to increase, but this won’t necessarily worsen global welfare
If TAI’s economic benefits are not evenly distributed across the world, global inequality is likely to increase. Most solutions I’ve heard suggested so far, such as a global UBI, require unprecedented levels of global coordination on tax and welfare policy. I do not expect this level of global coordination will happen in the foreseeable future. When TAI begins to displace workers at high rates, states will first look for ways to manage internal inequality and political stability pressures.[5] If that goes reasonably well, states may grow more benevolent in the longer term.
The good news is that rich countries getting richer doesn’t automatically make poor countries poorer as long as new wealth is being created—that is, if the pie is growing overall. However, poor countries can be worse off even if the overall pie is growing if the rich countries’ increased wealth makes it harder for poor countries to compete for a scarce supply of globally traded goods, such as food, energy, raw materials, minerals, etc.
Whether TAI will make poor countries worse off overall depends on how much it increases the supply of those resources vs how much harder it makes it for poor countries to compete for those resources. Again, the impacts on each country will be different, and will depend significantly on their natural resource endowments.
One area where increasing global inequality will likely hurt poorer countries is by further reducing their geopolitical power. However, I am uncertain how readily a country’s geopolitical power translates into welfare for its general population. Moreover, countries in the Global South already wield far less influence than their population share would suggest, so a further loss of relative power may not have large impacts on welfare for citizens (not justifying this in any way—my claim here is purely descriptive, not normative).
TAI could produce global public goods
Public goods are goods that are both non-rival and non-excludable:
non-rival means my enjoyment of that good doesn’t affect or diminish your enjoyment.
non-excludable means the good cannot be excluded from others (at least not without incurring a lot of costs).
Clean air is a classic example of a public good.
TAI will generate valuable new information, and information is naturally non-rival. Once generated, it can be shared almost costlessly. However, information is not a pure public good because it can be excluded using things like paywalls, intellectual property laws, and trade secrets. So while valuable new information generated by public institutions (e.g. universities, non-profits) are likely to be public goods from the outset, when private companies generate new information they may try to keep it secret in order to earn higher profits.
But keeping information secret is imperfect and often temporary—paywalls get bypassed, secrets leak, and IP laws expire or get breached (especially across borders). Others may also reverse-engineer the information or come across it independently. So I expect a lot of valuable information generated by TAI will seep into the public domain over time,[6] even if the people who generated that information do not wish to share it.
Global public goods will not reduce global inequality since, by their nature, they are available to all. But they can improve global welfare significantly. A few examples of public goods that TAI could produce or improve include: weather predictions; information to help with disaster response and aid; lower levels of carbon emissions; vaccine design; and repurposed generic drugs.
Of course, public goods will not automatically improve welfare if other bottlenecks exist. For example, reliable predictions of extreme weather events could save lives if they give people more time to prepare and evacuate. But if those people have nowhere to go and lack the resources to bunker down, improved predictions would do little for wellbeing.
Weakening influence of liberal democracies
One thing I’m pretty concerned about is countries becoming less liberal. While this isn’t itself an economic factor, it does have economic drivers and economic consequences. I think this could have a bigger negative global welfare impact than increasing global inequality per se.
Although liberal democracies can weaken even in the absence of TAI, I expect TAI to exacerbate this risk because:
If existing liberal democracies (especially the US) don’t manage their national inequality pressures well and become more extractive, they will be less likely to exert pressure on already extractive countries to become more inclusive.[7] One may reasonably question how effective this pressure to be inclusive currently is; but I do expect global welfare to take a hit if it disappears entirely.
TAI is likely to make immigration much more restrictive, at least in the near term. While this is already happening for various reasons, I expect TAI will accelerate these trends as the arguments for allowing in workers who are able and willing to do the jobs that locals can’t or won’t will disappear. This will close off an important development pathway.
Rentier states are states which get most of their revenue from rents (usually natural resources like oil and gas) rather than from taxing their residents’ labour. Such states already exist—examples include Saudi Arabia, Qatar, Russia, Iran and Venezuela. Life for people at the bottom of the pyramid can be truly awful. Some of the worst human rights abuses in these states are currently mitigated by the pressure exerted by powerful liberal democracies.[8] If the global influence wielded by liberal democracies shrinks, the pressure to contain human rights abuses could disappear altogether.
- ^
“Developing country” is an imprecise label, as there is a lot of diversity within this. But I think these labels are good enough for the purposes of this post, which attempts to describe high-level impacts.
- ^
There may be some win-win forms of strategic benefit sharing in the near term, but that only goes so far when one party is much more powerful than another. I am not assuming any higher levels of benefit sharing, driven by altruism rather than self-interest.
- ^
Such services make up a larger share of total exports than of employment, so TAI could have a significant impact on these countries’ relative wealth (which I discuss below). Here I am just talking about labour displacement.
- ^
Unless they get invaded for their land and resources. But seizing and keeping territory is hard, especially if the invader is on the other side of the world. So I expect most countries will reach negotiated settlements, though these deals could be quite lopsided.
- ^
I discuss in my first post how countries might manage the fiscal pressures from this.
- ^
Albeit not necessarily within my 10-20 year timeframe.
- ^
The extractive/inclusive dichotomy comes from Why Nations Fail. Broadly speaking, extractive institutions concentrate power and wealth in the hands of a few (usually by exploiting the many), while inclusive institutions distribute power widely.
- ^
For example, the public backlash during the 2022 FIFA World Cup led Qatar to make some positive changes to their labour laws (though enforcement is mixed).
Thanks for the post! This is a very valuable topic, and the development econ mainstream is totally lost on this question!
I agree with some of your points, but I think we need to distinguish very carefully between “developing countries”. All the factors you mention with regards to labour displacement (structure of the economy, data availability, telecommunication infra) are wildly different between, say, Togo, Brazil, and Indonesia. Same with private- vs. public sector diffusion; within “developing countries”, you’ve got countries with massive tech hubs and their own tech billionaires, and those where most people still don’t have electricity.
For me, the most important development question with regards to TAI (and the reason it’s important to distinguish) is the feasibility of the export-led development model. Generally, if countries manage to develop a high-value added export sector, they attract FDI, get foreign currency, climb up the value chain, and become richer. If they don’t, they stay poor. Except for the occasional country finding insane levels of natural resources, this is the only real way that countries have become rich over the last 100 years.
If we get safe, transformative AI, we can imagine that demand for imports massively rises in the West, and middle-income countries like China, Vietnam, and Indonesia with strong export sectors (and the infrastructure to build on their existing exports) are able to take advantage of this. As these countries already have good infrastructure (e.g. electrification, internet access, land and shipping transport) they can probably also benefit from AI & Robotics to develop “Industry 4.0” and make their export sector even more dominant.
I’d therefore estimate that a few of these “developing” countries with existing strong export sectors will catch-up and become rich relatively soon.
But what of the poorest countries?!
Most African countries with a GDP below, say $3000 are very low down the value chain in all sectors, with little but raw materials (e.g. coffee, cocoa, oil if they have it) as exports. They’re struggling to compete with Asian developing powerhouses, and they haven’t got the transport infrastructure, governance, or capital etc. to develop a quality export-led economy. In a world without TAI, as middle-income countries get rich, poor countries would develop the export industries and climb the ladder themselves, but this seems very unlikely with displacement of manufacturing labour by robotics.
My overall take is that (in an optimistic AI scenario) well-governed middle-income countries would probably end up more similar to rich countries. But we’d have really “kicked away the ladder” from the very poor countries.
Thank you! Yes, totally fair point. I am not trained in development economics so was very uncertain about this post, and expected there to be large differences between countries that I wouldn’t pick up. It’s disappointing to hear that the development econ mainstream has not been engaging with this topic.
I had in mind the lower-income countries (mostly in Africa) when writing most of this. Your point about how, without TAI, these countries might be able to develop export industries and climb the development ladder is an interesting one. I had thought of that briefly, but was unsure how likely that was to happen, given we haven’t seen any African country do it yet (to my knowledge). But perhaps it’s just something that takes time and can only really happen after the middle-income countries become rich.
I think it’s important to explore the potential impacts of TAI on low and middle income countries (LMICs), and there has been very little done, so thanks for doing this. I agree with most of what you say.
That could be true in the later stages of TAI with robotics, but in the earlier stages, I think that it will be more knowledge worker jobs that will be automated, and the demand for manual labor will increase dramatically, because the people making a lot of money from AI will want bigger houses, more roads, and more physical goods in general. Yes, there will be some unemployed knowledge workers who may want a manual job, but those who had some investments will likely be very rich. Also, in this scenario, government revenue would be large because they would be taxing the high wages of manual workers, and taxing massive corporate/capital gains. Some of the production of goods will be done in the HICs (perhaps increasing remittances), but I think a lot will be done in LMICs, which would a boon for those countries.
I think prices of those commodities would increase, but I think the worker wages would increase even more.
Again, I agree once you get huge amount of robotics, but in the first part of TAI, I think demand for physical workers in HICs will skyrocket, so they will want more immigration.
Thanks for the feedback and you raise some interesting points.
I’m not sure about this. I think we have different intuitions on how broad the class of “people making a lot of money from AI” vs how broad the class of knowledge workers losing out from TAI. My sense is that the former will be smaller than the latter, just based on data I’ve seen showing how little most people in developed countries have saved for retirement in the form of financial assets like shares (i.e. I haven’t researched this, and it likely depends on the country).
I guess this will also depend on the specific capabilities of TAI and how much labour it will displace.
I appreciate this challenge and it would be good to see some modelling on it (I might have a quick go at some stage).
Interesting point. Some of the goods you mention people getting rich from AI will want are bigger houses and roads. These are not globally traded so, to the extent TAI increases demand for construction work, I would expect this demand to increase in HICs rather than LMICs.
To the extent that people getting rich from AI want to increase their consumption, I expect much of that will be channelled through higher demand for in-person services—e.g. personal chef, gardener, masseusse, and (as you point out) construction.
Some increased consumption might also be channeled through physical goods, but physical goods are already incredibly cheap relative to incomes in HICs. The main bottlenecks to consumption are likely to be physical space (your house needs to be big enough for your toys) and leisure time. I expect TAI could remove the leisure time bottleneck for existing capitalholders, which could boost demand for leisure complements. But there is still only so much time in the day, and many higher-tech leisure goods like TVs, gaming rigs, bikes, cars, yachts don’t rely much on LMIC labour anyway. (Some things like clothing and tents may be more exceptions.)
I do agree with you that some LMICs could see a boon—but I don’t expect this effect to be widespread across LMICs in general.
I disagree with this because most governments tax labour more heavily than capital. So if all the income currently earned by labour just shifts to capital, all else equal tax revenues will decline. (At the same time government spending will likely have to increase to pay for increased unemployment. I go into this more here.)
If governments are able to shift their tax bases towards capital we might see tax revenues remain stable.[1] But there are major challenges to taxing capital: capital is mobile and more easily hidden, there are serious valuation issues when you try to tax unrealised gains[2], and it is distortionary because it is very hard to properly account for inflation. Plus there are a bunch of political challenges. All these challenges are why tax bases rely more heavily on labour to begin with.
I find it highly unlikely tax revenues would increase in the near term.
Most countries tax realised gains instead, but this leads to undertaxation because taxpayers can choose when to realise those gains, which is usually only when they need to fund personal consumption. There are also tax avoidance opportunities and distortionary lock-in effects that crop up when taxpayers can choose when to realise their gains—which is part of why capital is almost always taxed at flat rates rather than on a progressive schedule.
AI agrees with you on cars and yachts, but says the majority of TVs, gaming rigs, and bikes consumed in HICs are made in LMICs. Overall, I think I agree with you that the new demand created by new wealthy people would initially probably increase demand for manual laborers in HICs more than LMICs. However, if labor is the bottleneck, I think there will be a large incentive to shift production of goods to LMICs. I think this would not just be consumer goods, but even housing in the case of prefabricated panels (it only costs a few cents per kilogram to ship something across an ocean). I think it would also apply for high tech goods because the AI could tell the workers exactly what to do even if they are not very skilled. But I agree that this won’t necessarily work out for all LMICs (conflict zones, those that restrict trade, etc).
I think it’s best to discuss this in your other post, but for completeness, here is my comment from there:
Epoch estimates that wages will ~10x in the run up to full automation. Capital would rise much faster. If TAI is anything like this, governments won’t have to worry about revenue.
Fair enough. I think most of these are made in Asia and I do expect Asia (particularly China) to fare better than most other LMICs or developing countries.
Another thing I don’t see here is the importance of being in the supply chain of AI products. Countries with compute clusters whose compute they export will gain as AI becomes more ubiquitous, countries that are next importers of AI services with no stake in the chain will be harder hit. Even if AI is not substituting labor, but is only necessary productivity boost to keep up with the productivity of other workers, you end up paying a fee from each worker of yours for AI (mostly to developed countries). This may not be a big deal if open source models remain competitive, compute needs go down with algorithmic advancements, etc. but right now all countries are paying mostly US for AI use, and if it remains the case, US will benefit, as will Netherlands, Taiwan and others directly in the supply chain; others, less so.
Hi Dushan. I cover this at a high level under “But impacts will be uneven” heading. I agree with you that countries in the supply chain will benefit and others less so.