AI开发：事实和数字

Ruth Douglas explores the application of AI in international development and the risks associated with this rapidly growing technology.

Artificial intelligence (AI) is already being applied to every sector of international development from agriculture, health and education, to urban planning. And there is a groundswell of interest in harnessing this technology for social good.

Farmers are able to mitigate crop failure thanks to AI-based smartphone apps, health authorities can predict the next disease outbreak, and the world’s学校正在映射确定资源中的差距。

政府和公司正试图利用AI技术来预测和监测地震，洪水和干旱等自然灾害，并有效地针对紧急响应。

Yet the economic gains from AI in the next decade are expected to be relatively low in developing countries. According to recent predictions, global GDP will be up to 14 per cent higher in 2030 as a result of the accelerating development and take-up of AI — boosting the world economy by US$15.7 trillion.

The figure comes from areport on the value of AIby accounting firm PricewaterhouseCoopers (PwC), which finds that China and North America are likely to see the biggest economic gains, accounting for 70 per cent of the global total. But although it predicts that all economies will benefit, developing countries will experience “more modest increases” due to lower rates of adoption of AI technologies.

National strategies

尽管近年来，全球北部的许多国家都采用了国家的国家战略，但全球南方国家却越来越慢。根据《国家AI策略》的概述，印度，肯尼亚和墨西哥是唯一建立此类框架的人Politics+AI.

肯尼亚政府创建了一个专案组in 2018 to focus on AI and distributed ledger technology such as blockchain, and set out a roadmap for promoting these technologies in the next five years.

India’s AI strategy专注于社会增长和包容性，并于2018年6月与#AIFORALL标签启动。它旨在为印第安人提供在该领域工作的技能，并投资于AI研究，并将印度制造的AI解决方案出口到其他发展中国家。

However, despite a lack of national impetus in many nations, AI applications already abound which have the potential to improve lives and livelihoods in the Global South.

AI for agriculture

Agriculture is one sector that is seeing an explosion of AI-based applications, which are aimed at helping farmers tackle crop disease, pests and climate-related problems.

A new global industry农业人工智能报告beplay下载官网西西软件says the market size will grow from US$522.6 million in 2017 to US$1,765.98 million by 2023. It cites population growth and subsequent rising demand for agricultural production as a major factor in that growth.

With the global population projected to grow from 7.6 billion in 2018 to more than 9.6 billion in 2050, according to UN figures, there will be a significant increase in the demand for food. Meanwhile, natural resources such as water and fertile agricultural land are coming under increasing pressure from climate change and overuse, fuelling the need for new technologies to improve climate resilience.

根据数字创新和技术可以成为解决方案的一部分。Food and Agriculture Organization (FAO).AI技术（例如机器学习和预测分析）使农民可以了解天气状况和土壤水分，作物疾病和害虫，以告知其决策。许多非政府组织，初创公司和多边机构致力于使发展中国家的小农户可以使用它们。

Connectivity challenge

But for many of these farmers, living in rural areas, there remains a major obstacle – connectivity. And the same is true for health workers and others on the front line of development.

On the one hand, mobile phones have become ubiquitous: a 2016 World Bank report titledDigital Dividends发现，在发展中国家，有十分之一的人拥有手机，数量稳步上升。Even among the bottom fifth of the population, nearly 70 per cent own one, it said.

But despite this, “the lives of the majority of the world’s people remain largely untouched by the digital revolution”, the report finds. Nearly 50 per cent of the world’s population has no access to the internet and in developing countries the figure is closer to 40 per cent, according to the UN’sBroadband Commission for Sustainable Development.

这是AI在全球南部面临的主要挑战之一。在许多领域，连接性和带宽仍然有限或无法承受，并且可能不足以用于基于AI的应用程序。

In low- and middle-income countries, 1GB of data costs more than five per cent of what people earn in a month, well over the “affordable threshold” of 1GB of data priced at 2 per cent or less of average income, according to the负担得起的互联网联盟. The UN’s Sustainable Development Goals laid out in September 2015 call for universal, affordable internet access by 2020.

在这些国家，通常缺乏数字素养以及开发和维护复杂的AI系统所需的技能。AI应用程序依赖大量数据，这可能并不总是可用。

What are the risks?

Added to this, there are concerns over potential inaccuracies of AI software. AI uses data gathered in a particular context. If it is applied elsewhere it might not be relevant or, if applied at a later point in time, may be out of date.

“ AI工具应始终按本地量身定制，以便相关的上下文和敏感性告知软件开发人员如何处理工具设计和评估，” USAID报告标题为Making AI Work for International Developmentrecommends. It calls for “deep local partnerships”, involving technology users, owners and regulators at the design stage, and consulting them early on in the process.

This report also points to one of the most frequently cited risks of AI, both globally and in a development context: the danger that existing societal biases and prejudices could be built into AI programmes. AI models rely on data and this data may already reflect such biases. Or indeed the programme may reflect the inherent biases of its designer. Either way, the result could be the further marginalisation of already vulnerable or excluded sections of the population.

最后，人们担心AI会从事人们的工作。但这是建立的吗？当然，随着流程变得自动化并简化了服务，AI有可能使许多工作变得多余。但是，人们普遍认为，这些损失将通过在该领域的新角色来弥补这些损失。

In a survey由毕马威（KPMG）和哈维·纳什（Harvey Nash）推出的108个国家/地区的技术业务领导者中，受访者通常表示，他们认为公司的劳动力中约有10％的劳动力将在五年内通过AI/Automation替换。但是，超过三分之二的人说，他们认为新工作将出现以弥补这一点。目前，反乌托邦的机器人概念仍然是一个遥远的威胁。