Self-learning comes before machine learning

The government must relax investment laws to attract overseas capital in the IT sector

NATURAL INTELLIGENCE: Software engineers at work on Thursday in the Naxal office of the company Wiseyak, which primarily exports AI healthcare products. Photo: SUMAN NEPALI

Nepal has seen a dramatic recent growth in its Information Technology (IT) sector with tech companies and freelancers building software products, as well as providing digital services for clients all over the world.

Nepal’s IT service export industry in just the first half of 2023 is estimated at $800 million, although some say this is an underestimation. In 2022, an Institute for Integrated Development Studies (IIDS) report valued IT exports at $515 million, a whopping 64.2% growth over the previous year.

Annual IT exports now surpasses Nepal’s direct earnings from the tourism industry. Besides assisting in bridging the balance of trade deficit, it also provides jobs to thousands of young engineers who may otherwise have emigrated.

Advances in Artificial Intelligence (AI) are already impacting programming and software, however this can also allow Nepal to leapfrog into a new age of innovation by building on its young talent pool.

“Nepal slept through the last tech revolution,” says Kiran Bhakta Joshi of Incessant Rain Studios, referring to the dot-com boom of the internet bubble between the mid 1990s to the early 2000s as everything started going digital and online. “AI provides us the chance to harness cutting-edge technology and shape our future.”

The growth of Nepal’s tech industry is happening in parallel with the boom in the global AI market with dozens of new apps coming up every day. Nepal should take advantage of both trends and go all in on AI.

The IT sector in Nepal directly and indirectly employs an estimated 1.5 million people already, and by declaring the next 10 years the Decade of Information Technology could give tech exports a boost.

Joshi is the founder of Incessant Rain Studios and has worked on some of the most iconic animated movies of all time like The Lion King and Beauty and The Beast. He recalls how, while working at Disney, digital animation replaced traditional drawing methods really rapidly.

Learning from this, Nepal’s planners and government should be taking proactive steps to help tech companies here succeed in the AI game. Nepal needs the infrastructure, human resources and foreign investment to take maximum advantage.

But, says AI scientist Suresh Manandhar: “Nepal is going to miss the AI surge. It is a policy issue. As long as there are policies strictly regulating what foreign investors can do with their profits, overseas funds in the tech sector will be low.”

Read also: IT is it, Sahina Shrestha

IT revenue Nepal NT

Besides barriers to more foreign investment in Nepal’s IT sector, there is also a skills gap. The 2022 IIDS report talked to Nepal-based IT companies about problems regarding outsourcing work, and 67% of them cited the lack of skill and competence, and 53% noted a lack of professionalism.

And that is an area in which Nepalis cannot get away with blaming politicians. A large part of the appeal of tech careers is the ability to self-learn, and courses, videos and books are free all over the internet. The IIDS report stated that a fifth of all IT sector workers did not have any degree.

This means Nepalis must first concentrate on self-learning before they can master machine learning and AI. Tech promoters say that there is not much we can expect from computer institutes or the government. 

Suresh Manandhar is an AI scientist with over 30 years of experience in research, teaching and startups. Manandhar taught at the University of York for 23 years, heading its Artificial Intelligence Research Group. 

He is currently head of Artificial Intelligence Research at Fusemachines and Honorary Chair in Artificial Intelligence at the Madan Bhandari University of Science and Technology (MBUST) in Chitlang near Kathmandu. 

Manandhar’s priority for now is with his startup software company, Wiseyak, a healthcare initiative that uses AI and machine learning to enhance medical care and treatment. It also provides intelligent chatbots for the Nepalese and International market, and is working on large language models of its own that are more bilingual in Nepali and English, with transcription and conversational capabilities. 

“We are one of few companies in Nepal working with our own AI models, most companies are using APIs,” Manandhar told us. While Wiseyak trains its own AI models, most other Nepali software companies simply integrate advanced AI tools like ChatGPT into the programs they build. 

Manandhar says that as long as Nepal’s foreign investment policies are restrictive with strict rules on what investors can do with their profits, the flow of overseas funds will be low. “Running Wiseyak has been very difficult, much more than I thought,” he admits.

With AI applications getting better at generating deep fake videos, Kiran Joshi at Incessant Rain Studios is aware of its impending impact on animation and his business. After all, his company has done well because Nepali animators are less expensive to hire than ones in Japan or the West.

Joshi reckons that given the rapid advances in AI capabilities, the entertainment industry has about two years before AI is able to produce content close to the quality Hollywood requires. 

Read also: Bringing Silicon Valley to Kathmandu Valley, Sonia Awale

IT company Nepal NT
Photo: MONIKA DEUPALA/NT ARCHIVE

But Hollywood is already panicking, with both writers and actors going on strike in 2023 demanding safeguards against losing their jobs to AI.

“The landscape is incredibly volatile. You have to lead, and you have to stay very agile,” says Joshi, who is taking as many steps as possible at Incessant Rain in Kathmandu to stay ahead of the curve – to tame AI as a tool, instead of being devoured by it.

Joshi’s studio has introduced AI tools into its workflow, and is working with a Hollywood startup to carry out the manual, repetitive parts of animation post-production. Incessant Rain’s artists and animators all dedicate some of their time to learn about AI tools in their departments, and share findings with the large staff.

Two years ago, Joshi established an academy to teach the fundamentals of animation and storytelling. Starting October, his team is adding a course that goes over the basics of programming AI.

Another successful Nepal-based company that focuses on AI is Fusemachines, founded in 2013 by Sameer Maskey, who has a PhD in AI and Machine Learning from Columbia University, where he now also teaches.

Maskey had a prolific research career while also building a startup that is now headquartered in New York, and became the first Nepali company to be listed on the NASDAQ to go public.

Fusemachines makes several AI products, including Xtract, which understands and extracts important information from documents. Another is Fraud Detection AI, which identifies and investigates scamming patterns. Yet another application forecasts demand and inventory, promising high precision.

The company also runs a fellowship program as part of its pledge to ‘Democratise AI’, which involves a scholarship for a six-month course that aims at ‘producing high-quality AI and data science talent’.

For Manandhar and others in Nepal’s IT sector, finding venture capital and other investors to keep their companies running is a major hurdle, and the government so far has not only been no help, but has proven to be a hindrance. 

Kathmandu University has an AI course, COMP 472, and the syllabus is outdated with vague course objectives promising students will ‘get the full picture of AI easily’ and even adds that if interested they ‘may go on to graduate school for further study’.

The government recently released a ‘Concept Paper’ about AI in which it mentioned Alan Turing and the ‘Lmitation Game’ (sic, typo in the report) in its executive summary. The document does identify problems such as a lack of skill and infrastructure and data security concerns.

But the government’s foremost priority seems to be to regulate AI, and the second seems to be to make sure developers know the rules and follow them. Increasing research comes third, but with no pathway to reach it.

AI is now designed for quality control at factories or used in spotting cancerous tissue, but even before scientists have decided how it will affect jobs and technology, the Nepal government already wants to regulate it. Those resources would be better spent in making computers and the internet more accessible to citizens in underserved schools.

“Forget dealing with the bureaucracy, we use redundancy to get around the challenges we face in Nepal,” says Joshi, who has taught himself accounting-for-risk and do-it-yourself lessons running his business in Nepal.

The IT business already deals with prolonged power and Internet outages, and as more and more young Nepali engineers leave the country, or are planning to leave, technical staff needs to be replaced with year-round training to deal with the turnover.

“People migrate mainly for two reasons. They need money, or their job bores them. So I give my team projects that are interesting to work on,” says Joshi, who saw a lull in business during the Hollywood strike.

More successful tech companies means more jobs for Nepalis who prefer to stay home, more revenue for the government, hence more reinvestment in the sector, and a way to break the overwhelming dependence on remittance.  

Read also: Natural Intelligence, Nepali Times

Vishad Raj Onta

writer