20 June 2023
By Annys Rogerson
The government of the Republic of Tajikistan has set the goal that 1% of its GDP will come from AI-related economic activity by 2026. This is an ambitious goal as Tajikistan is predominantly a primary sector economy, with agriculture responsible for 60% of employment. However, as the country begins to industrialise, its policymakers are looking one step further, towards AI. Why? Azizjon Azimi, Head of Tajikistan’s AI Council and founder of startup zypl.ai, suggests that supporting AI development in the country will help Tajikistan avoid the middle income trap, where a country progresses to middle income status but fails to transition to a higher income economy. Therefore, while the Ministry of Industry and New Technologies focuses on supporting industrialisation, the recently formed AI Council is working to demonstrate how AI can contribute to the country’s long term development.
At this first stage, the AI Council has prioritised demonstrating why AI is something the Tajik population and its government should care about. This means showing how AI can be deployed usefully right now. To do so, the AI Council has been working with public and private partners to find and test use cases across several sectors of the economy. These include:
AI is now used to underwrite 1 ⁄ 4 of all loans in the country, as a result of 8 financial institutions adopting software developed by the Tajik startup zypl.ai.
Now, the National Bank of Tajikistan is looking to progress digital banking applications further by using AI for remote identification. The aim is to increase the number of people who can apply for loans or open bank accounts among those who are not able to attend a bank in person.
AI is now supporting career counselling for high school students in Tajikistan. For the first time this Summer, students will complete a questionnaire as part of a national exam at the end of high school and their answers will be fed into an AI algorithm that provides career recommendations. This project is part of a collaboration between UNICEF, the NGO TajRupt, the Ministry of Education, the Ministry of Labor, and the AI Council and will be rolled out at the national scale.
Another collaborator with the government is the UNDP, who recently co-delivered the ‘Upskill 1.0 project’ for University teachers. This saw 50 university professors complete a one-year programme on ML teaching.
A third project within education is piloting the use of facial recognition technologies as an alternative to access cards or fobs to enter school buildings. The project is currently being piloted in 3 schools. The government has worked with their private partner to protect student privacy by ensuring both that the ML algorithm is run locally within the school, and the servers store the data internally.
In the telecoms industry, companies are using AI to predict which of their customers are likely to switch providers and increase customer retention. Significant competition in the industry has led to a high number of customers switching providers, posing a challenge to telecom companies. In response, two major telecom companies built AI models in-house with talent from the AI Council’s AI Academy. The companies are now using these models to predict customer churn and offer bonuses to customers with a high likelihood of switching providers.
The government has been an important player in getting these use cases off the ground; either as a buyer or builder of AI. The government’s involvement in each of the cases was guided by its national AI strategy, officially approved in 2022 with the support of Mr Sherali Kabir, Minister of Industry and new technologies, and the efforts of the AI Council.
Beyond demonstrating current use cases of AI, the government is working to create the conditions for AI companies to grow in the country. Their work appears to centre around two aims (1) encouraging companies to set up in Tajikistan by defining their competitive advantage in AI and (2) adapting the legislative and regulatory frameworks so that AI applications can scale and be deployed responsibly.
Since economic activity is largely not digitised in Tajikistan, companies lack the data needed to build AI. However, Azizjon Azimi thinks they’ve found a route around that problem; synthetic data. Synthetic data is data generated using a sample of real data and a model trained to produce more data with the same characteristics and structure of the sample. This technique gives teams a route for generating more, usable data based on a small sample. Therefore, Azizjon Azimi suggests that synthetic data makes it possible to build models faster, without the need to first take more basic digitisation steps.
Synthetic data is at the core of many of the AI projects in Tajikistan, and the country’s AI Academy is focused on upskilling students in how to use this technique. The bigger aspiration is for models to be piloted in Tajikistan, proved useful, and then be shipped abroad and scaled using real data. Azizjon Azimi suggests that, in this sense, we can think of Tajikistan as the “sandbox capital of the world”.
The AI Council is working to create a more start-up-friendly legislative environment by setting the basic preconditions for startups to set up and grow. This includes encouraging the government to pass venture capital legislation to enable Tajik startups to raise capital in the country. Azizjon Azimi points out that, currently, this is not possible. A company looking to raise VC has to open an entity abroad and raise money there before bringing the capital to Tajikistan.
In Tajikistan, regulations around AI are the responsibility of existing sector regulators. Azizjon Azimi notes that the government is choosing not to create a new regulatory body for AI in order to avoid over-bureaucratising the sector. As things stand, an organisation looking to develop an AI application submits a proposal to the relevant sector regulator, who then has the remit to accept or deny the proposal. While excessive bureaucracy is a worthy concern and a sector-based regulatory approach may be appropriate, the approach needs to be accompanied by proper training and guidance for existing regulators who may not be familiar with the unique risks of AI. One way to do this could be for the AI Council to play a coordinating role in upskilling regulators.
In addition, since there is no AI-specific regulation, approval of projects is left to the judgement of regulators. Without formal regulations or mandatory guidance, citizens cannot be reassured that only responsible uses of AI are approved and must rely on the regulator to make the right call in each case. Take the example of deploying facial recognition in schools. It is positive to see that the government took action to ensure the protection and privacy of student data. However, citizens need certainty that, firstly, these concerns will be addressed in all cases, and secondly, other concerns will be considered too, such as whether facial recognition could disadvantage some students if the algorithm does not have a high enough accuracy for all groups in the population. It would be positive to see upcoming legislation on AI, the timetable for which is laid out in the national AI strategy, addressing the need for standardised regulations or guidance across regulators.
Another key part of the AI Council’s work is driving regional coordination on AI. In the Government AI Readiness Index, Tajikistan is behind in the region in terms of the maturity of its technology sector. However, Tajik startup zypl.ai’s recent win at a regional Google Accelerator competition and the company’s expansion into Kazakhstan and the MENA region has shown the country’s ability to produce competitive companies. The company’s expansion also demonstrates that the benefits of AI development are reaped by the whole region.
With this in mind, Tajikistan’s AI Council and their peers in neighbouring countries are working on a common document to harmonise regulations and principles on AI, including an approach to AI ethics, across the region. Once the document has been agreed at this level, the next steps are for it to gain ministerial approval and, lastly, to be agreed by the C5, a government forum of 5 central asian countries (Tajikistan, Kazakhstan, the Kyrgyz Republic, Turkmenistan, and Uzbekistan).
This coordination is expected to promote investment and allow AI workers to move between countries. It offers an important study for other regions looking to understand the benefits of coordination, and could be a useful demonstration of the power of forums like the C5 to take impactful measures. Looking ahead, this coordination provides an important opportunity for supporting more success stories of using AI to serve the needs of people in the region.
Should we be interviewing you next? Contact us at email@example.com