Development of artificial goods is concentrated in a few countries and countries in the Global South are largely under-represented, industry leaders and experts said on Tuesday.

While speaking on the first day at Global Partnership on Artificial Intelligence 2023, UNESCO Assistant Director-General (ADG) for Social and Human Sciences, Gabriela Ramos said UNESCO works with 193 countries, but the scenario is not good. The three-day GPAI 2023, New Delhi witnessed some of the prominent minds speaking their thoughts and opinions on AI, including Prime Minister Narendra Modi.

"I can tell you that the image is not as happy as it looks because the truth is that these capacities, the amazing transformative power of AI, is still highly concentrated in a few hands. Few countries, there has been a 28-fold increase in investments in AI for the last decade, but this has accrued to the US, China, and some other parts of the world. It's highly, highly concentrated in terms of the development," Ramos said.

AI software needs high performance computing infrastructure to read the software and process various parameters based on data used for getting desired output.

According to Top 500, top performing supercomputers are located in US, Europe, Japan, China’s, South Korea, Saudi Arabia where purchasing power parity is high. China has 162 supercomputers, followed by US which has 127 supercomputers and India has 3 supercomputers, as per Top 500 list issued in November 2022.

Indian government has announced plans to scale up GPU based compute infrastructure.

Minister of state for electronics and IT Rajeev Chandrasekhar has said that an announcement related to setting up significant compute infrastructure will made in January.

Ramos said that there are many sectors that can be improved using AI.

"I have to say that we really need to tackle the question of high concentration in terms of how these things are developed. Because at the end, this is what led to the high concentration at the top, leading to the discriminatory patterns that we see at the bottom," she said.

Ramos said that the Global South is not being represented in the data just by the mere fact that half of the world doesn't have access to a stable internet.

“So, I think this question is not easy to solve. We really need to do a huge effort to bring these capacities to the developed developing world and then to ensure that we have not only the possibility of the global south, to take advantage of the technologies but to be able for them to develop their own systems,” she added.

“If you're a government in the Global South, you don't want to have a platform, a private platform running your health systems or your education system, or your food and agriculture just doesn't work like that, because there are questions of security,” she said.

At present demand for GPUs (Graphics Processing Units) over CPUs (central processing units) based servers has increased because of requirement of high-speed computing requirement of AI.

NVIDIA currently dominates GPU market with about 88 per cent market share.

NVIDIA’s AI Technology Centers, Director, John Ashley said that NVIDIA has been focusing for a long time on democratizing AI.

“We want to make the software, the hardware, the solution to the systems, available to as many people to as broad an ecosystem as possible,” Ashley said.

He said NVIDIA spends a lot of time optimizing so that you can do more with less, right, less energy and less time.

When asked about AI development with respect to AI, Ashley said addressing the Global South is a huge challenge.

“It's one that our global public sector team and our admissions team takes very seriously. If you look at the best sort of proxy for where computers are distributed, supercomputing list. There's only 30 some countries on that list, which means there's 150 some countries that are not. Those countries are disproportionately in the Global South,” he said.

He said there's going to be a need to reduce a certain amount of the entry barrier so that it's easier for people to adopt models.

“We can share the computer. There are countries across the global south that have got the education systems and the resources to deliver the talent. The challenge then becomes data. You can have the talent be removed from the computer, but you can't have the data removed from the computer,” Ashley said.

He said NVIDIA is working through issues with technical solutions, federated learning, privacy, encryption, and also with political solutions, treaties etc. 

OpenAI, Global Affairs, Vice President, Anna Makanju said that AI can generate positive change.

She said that as these systems become more powerful, they can unlock even more personalized energy, faster travel life-saving medical treatments.

Google, APAC Google Research and AI Partnerships, Head, Harsh Dhand leads APAC level initiative and has the data generated collected curated in the region all across. He said that company is building data sets for large language modules that can be used by stakeholders in AI ecosystem. 

"India being an epicenter all the way from Southeast Asia, we are making sure that we collect the data from the region which is trained. We are ensuring that regional large language models not just for Google, but for everyone in the ecosystem--for open AI, startups," Dhand said.