What does Artificial Intelligence mean for the future of poor countries?

September 22, 2017

     By Duncan Green     

What do the multiple overlapping new technologies currently breaking in tsunamis over the world’s economies and societies mean for the future of low and middle income countries (LMICs)? Last week I went along to a seminar (Chatham House Rule, so no names) on this topic, hoping for some interesting, preferably optimistic ideas and examples. I came away deeply, deeply worried. Houston we have a very substantial problem.

The conversation took place in a hipster warren (lot of beards per square metre) in London’s ‘silicon roundabout’ (yes, that’s British humour for you). The skype crashed during the first presentation from East Africa and we had to give up on it (but here’s what they probably would have said). Turned out to be an apt metaphor for what followed.

What technologies are we talking about? The OECD has kindly pulled them together on this mind boggling chart.

The tech revolution will affect LMICs on at least four levels:

  • Jobs and the productive economy
  • Private consumption of goods and services
  • Government service delivery
  • Rights and Voice

I found different balances of optimism/despair on the four. At the optimistic end, there are lots of apps and potential leapfrogging (eg diagnosing disease or mobile banking) which show real promise in bringing goods and services to hitherto excluded consumers. E-government could be a lot more responsive and less corrupt than the analogue variety.

So much for the upside, which was completely eclipsed (for me at least, we had a few techno-optimists in the room, exhibiting what one speaker called ‘the TED talk view of the world’) by the dark side. On Rights and Voice, state surveillance, private control of data, and general invasion/manipulation of our lives is becoming apparent much faster than the potential positives of cyber-activism.

But what really worried me was the economic impact of the new technologies. In particular, their impact on the way countries have historically moved from poverty to prosperity. An excellent paper by Shahid Yusuf of CGD and George Washington University sets out a gory prognosis that AI will kick away the ladder of development, under the striking subhead ‘will there be new jobs or no jobs?’:

‘There are those who believe that technological change will be a gradual process as in the past allowing institutions and labor markets ample time to adjust. The optimists are of the view that although automation, AI, and other technologies that are in the wings will in time eliminate a swathe of jobs, the pace of automation is likely to be slow and new occupations will appear to take the place of the ones that disappear as was the case in the past—assuming of course that economic growth is robust. Past industrial revolutions gave rise to similar fears that proved to be unfounded.

Could it be different this time around? Those who subscribe to the precautionary principle maintain that the pace of technological change has quickened making it harder for markets and institutions to make timely adjustments. The better-paid jobs that might materialize will require higher levels of cognitive, non-cognitive, and technical skills that are not plentiful in emerging economies and will accumulate slowly. Many emerging economies are currently coping with a large overhang of unemployed and underemployed workers and urgently need to ramp up employment in manufacturing and tradable services.’

His conclusion? ‘Export led growth of the sort that created the East Asian legend of the 20th century could be a thing of the past….. This includes industrialized economies that have relied on the exports of manufactures and participation in global value chains—i.e. the Malaysias, Polands, and Thailands—as well as other economies—the Pakistans, Egypts, and Honduras—that are at earlier stages of industrialization.’

So the most successful ladder of development of the last 70 years – low skill, labour intensive industrialization – looks like to be kicked away. 5 million women in Bangladesh currently working in the a garments industry that, for all its flaws, has transformed their lives (and accounts for 80% of Bangladesh’s exports), will probably be replaced by robots, located nearer the consumer markets (no need to make T shirts in Bangladesh if cheap labour is no longer necessary).

Yusuf thinks some countries can find alternative paths to development through natural resources, services, tourism or migrant remittances. For the rest, the outlook is grim.

The usual policy responses seem terribly inadequate to threats on this scale: A few cool apps really aren’t going to provide jobs for that number of people; trying to retrain huge populations of low skilled workers for high tech jobs, even if education systems are up to it, is running up a down escalator – robots and AI are likely to upskill faster than people.

Another commonly proffered response is a Universal Basic Income – crudely, the state taxes the robots, redistributes the income to the people, and we will all have to learn to enjoy our leisure. But it is easy to imagine other, more dystopian, political settlements between elites and robots. Jobless growth and rising inequality can lead to reforming governments, populist manipulators or repressive crackdowns (aided by new surveillance powers – according to last week’s Economist, China used facial-recognition technology to catch 25 criminal suspects at a beer festival, one of whom had been on the run for a decade).

Discussions about this topic usually divide between the optimists, who cite the Luddites and compare fears of technology to Malthusian anxiety about mass starvation through population growth, and people who say that, as with Climate Change, this time is different. After last week, I’m definitely closer to the second camp, even compared to this post from March. Thoughts from a slightly more optimistic angle here.

September 22, 2017
Duncan Green