Thinking DEEP to ensure AI delivers the greatest impact
October 4, 2023
At this year’s UN General Assembly, world leaders discussed the importance of inclusive global governance for Artificial Intelligence (AI), which presents powerful opportunities for humanity. This is reiterated in the UN Secretary-General's statement, in which he noted that a new UN agency may be required to help the world manage it.
Across UNDP’s country and representative offices, we are hearing debates and curiosity about ChatGPT, Large Language Models, and the trillion dollar impact they could have.
UNDP aims to play a significant role in shaping the future of AI for the better.
What exactly is AI?
AI is technology that can ‘think’ like humans - allowing it to solve problems, recognize patterns and make decisions. It’s not new. Research began in the 1950s, and the mathematician Ada Lovelace was discussing whether ‘computers’ could create original work as far back as 1843 (the 8 is not a typo!). However, today is different. Since its public launch last year, ChatGPT, an open AI platform that draws on enormous volumes of data to create content, has recorded more than 100 million users.
But it’s also posing considerable challenges. Many training datasets reflect the context, terminology, and realities of the Global North. There are concerns that automation could lead to significant job losses. We are also seeing global ‘decoupling’ as countries develop different approaches and it has the potential to accelerate problems such as disinformation. One AI-generated image led to US$500bn in value being briefly wiped off the US stock market.
So, how can we leverage the potential of AI and mitigate its downsides? AI, if developed and deployed ethically and responsibly, could be a crucial tool for national and digital development. It could lead to new industries and opportunities, greater economic growth, and improved service delivery. But these benefits won’t happen by themselves.
AI is often discussed in the context of tools and technologies – from machine learning to large language models (LLMs). But fundamentally, it’s about people: from researchers and computer scientists exploring cutting-edge innovation, to policymakers grappling with the opportunities and challenges of AI, and the individuals and communities who will be using AI-driven products and services.
AI is fundamentally about people, and this focus is essential. Applications of AI need to be led by lives and livelihoods, and not just by data points and digital - as highlighted by Bangladesh’s focus on ensuring that AI drives both economic and social growth. Similarly, putting people at the centre of AI thinking, piloting and scaling is a crucial foundation of our AI approach at UNDP is approaching it through a ‘DEEP’ lens.
- Democratize and demystify AI. AI should be understandable for all, and this will be achieved by driving public awareness, knowledge, and understanding of AI technologies – including their benefits, risks and terminology. By fostering an environment where people are equipped with the knowledge and skills to understand AI, its uses and limits can be a truly whole-of-society discussion and exploration. As an example, the Government of Finland launched its free and open online course – the Elements of AI – in 2018, to improve the critical understanding of AI across the country’s population. This also highlighted the potential for a whole-of-society conversation: over a quarter of course participants are aged over 45. Through broadening the accessibility of AI in these kinds of ways, we could shape it to be more relevant and useful in different contexts.
- Empower people to use AI to improve their lives and livelihoods. AI should serve all people. It should add real and important value, and it should not be an abstract concept. This includes moving beyond awareness-raising and making AI products, tools and technology relevant to the contexts and priorities within countries – including supporting people to integrate AI into their employment, their businesses, and into their lives. For example, Colombia has crafted a National AI Strategy with the aim of leveraging artificial intelligence to empower its citizens and drive economic prosperity. Whilst Nigeria is engaging its overseas diaspora to guide the country in leveraging AI for national benefit, and India aims to train three million civil servants in exploring the potential role of AI for policy and service delivery.
- Explore and experiment with AI, in a safe way. AI is evolving rapidly, yet its benefits are still unclear in many settings. Despite this infancy, many AI innovators are moving fast to develop AI products and services that are founded on enormous amounts of personal data. This data needs to be used responsibly – and protected. We therefore need to shape inclusive and safe opportunities to explore and experiment with AI. This includes physical testbeds and regulatory sandboxes being explored in higher-income countries (although other countries are broadening some of their existing structures) that could identify and shape these protections. It also extends to building cultures of responsible experimentation (with feedback loops) to identify how AI can have the greatest positive impact. We also need to think about a longer-term perspective. For example, the Government of South Africa has established a Presidential Commission on the Fourth Industrial Revolution to explore how emerging technologies could shape the country’s future.
- Protect people from the negative impact of AI. AI is being explored in all aspects of our governments, economies and societies. However, AI needs to be people-centered: with the rights, safety, and consideration of people as essential foundations. It also needs to be used responsibly and ethically, which can pose wide-ranging challenges. It includes tackling underlying data bias, providing grievance redressal and accountability for any AI decision-making, and ensuring that AI does not entrench or exacerbate digital (or broader) divides. The National AI Policy of Rwanda includes six key elements including ensuring ‘trustworthy AI adoption in the public sector’ through building the skills and capacities of public servants and regulators. Whilst the African Union may yet be a potential source of ‘model’ legislation on AI and key foundations such as data protection. A number of other countries are also shaping policies on the ethical usage of AI.
Understanding what AI means for development
UNDP works in more than 170 countries and territories around the world. And many of these places are already actively engaging with AI. This global portfolio has reaffirmed the importance of what we call the 5 ‘Cs’.
First, the importance of context. Understanding the local realities and landscape where AI could add value (and where and how it could also have negative consequences).
It’s about collaboration and convening. No single country has ‘figured out’ AI; this is a global journey and conversation.
We are focusing on capacities and competencies. How do we improve the understanding, development, and application of AI across society – and build the skills needed for local research, development, and scaling?
And, increasingly, we’re seeing a sixth ‘C’: comparative advantage. How can countries and communities make AI work most effectively for their national development priorities and international standing – including in achieving the SDGs?
Across our global work, we’re also seeing three other trends.
First, AI is one tool in our global development – and digital development – toolkit. It is not a panacea, and it is not a replacement for human ingenuity. Instead, AI can augment, complement, and amplify human efforts. And people need to be at the centre of AI (and digital transformation more broadly). In recognizing this, we need to ensure that AI adds value, and that AI development and usage is founded on an inclusive and rights-based approach. We need to avoid technological ‘veneers’ that exist solely for technology’s sake. And we need to understand when AI may not be relevant, useful, or beneficial.
The digital manifestations of AI – from ChatGPT to the AI-image generation platform Midjourney – are a very small snapshot of the scale, potential and risks. They are only the very public face of a global AI ‘supply-chain’. And many of the aspects of this supply-chain are not technological. As the researcher Kate Crawford argues, AI is neither artificial nor intelligent. It is not artificial because it is made possible by vast amounts of natural resources, fuel, and physical labour. And it’s not intelligent because it is often relying on inadequate data and other sometimes weak foundations. To engage we need both computer science and social science.
Finally, AI is more than Silicon Valley. It’s about the Silicon Savannah in Kenya, where AI startups are approaching $100m in funding. It’s about an emerging regional approach from Latin America, to Thailand’s efforts to shape an AI ecosystem by 2027, and everywhere in between. Many of the exciting, important, and cutting-edge explorations of AI are being undertaken in ‘developing’ countries. Despite rhetoric focusing on a handful of companies and countries, there is an enormous opportunity for global discussion, collaboration, and innovation.
As we stand at the start of an era where technology and human endeavours could become more intertwined than ever before, it is crucial that we understand and leverage the power of AI in a responsible and equitable manner. AI could bring us one step closer to a more sustainable, inclusive and prosperous world. But it could also entrench existing digital and other divides.
We will only be able to gain the benefits, and tackle the downsides of AI, if we approach it through the DEEP lens. And if we work together. Get in touch if you would like to collaborate. Digital@undp.org