Milan (AsiaNews/Agencies) – The Philippine Department of Agriculture and the Philippine Space Agency (PhilSA) are to launch a partnership to promote the use of artificial intelligence (AI) in support of agriculture and food security.
According to reports last week in the Philippine media, under a five-year agreement, satellite data processed via AI will be used to monitor crop conditions, estimate harvests and map areas affected by floods and droughts, identifying the most vulnerable zones.
Francisco Tiu Laurel Jr., Minister of Agriculture, added that these technologies could also be used to identify areas where excess water can be collected and stored, so that it is available during the dry season. “Some of the water that we currently regard as a hazard could become an agricultural resource,” he said.
Tools of this kind have also been introduced in Indonesia, Malaysia and Vietnam, demonstrating how, in South-East Asia too, the use of AI is shifting from predominantly individual applications to practical uses in key sectors. In this region, however, one of the issues causing the greatest concern is accessibility.
The fear is that this technology may prove less accessible or less effective for communities speaking local languages, for trading partners deemed less strategic, and for poorer countries, which may lack the financial resources needed to access efficient models.
Regional interest – involving governments, health services, universities and citizens – is therefore not solely about which of the major powers will develop the most advanced model.
From a practical point of view, in fact, even the most sophisticated system risks failing to produce positive results if people are unable to use it, for example because it is not available in local languages, such as Tagalog in the Philippines or Bahasa Indonesia.
The problem is particularly acute in a region characterised by significant linguistic diversity, where many local languages are under-represented in the data used to train the models.
A system that works well in English but produces results that are difficult to understand in other languages risks being unusable in sectors, such as healthcare or public administration, where linguistic accuracy is paramount.
According to Zenobia Chan, an associate professor at Georgetown University, the main concerns regarding AI in South-East Asia relate to accessibility, which manifests itself on at least three levels.
The first is – precisely – the linguistic one. A second limitation, however, is economic: for many public administrations and small businesses, access to these models depends on their cost, with the risk that a model that is too sophisticated may prove beyond their means.
Finally, there are political risks: national approaches could increasingly influence the distribution of and access to the most advanced systems, and states that are considered neither threats nor priority strategic markets might be the last to obtain the most sophisticated technologies, or receive versions with limited functionality.
To address these issues, several ASEAN countries have developed – or are seeking to develop – their own AI language models.
Indeed, as Chan points out, what is at stake in the ‘AI race’ is not so much the attempt to achieve full autonomy – a goal considered unrealistic for most of these countries, and which would probably not even constitute an efficient use of public resources.
The challenge rather concerns the region’s ability to secure ‘resilient access’ to these models, meaning access that is practically available both now and in the long term.
According to Chan, here too it is essential to combine technologies from different ecosystems, so as to preserve credible alternatives and maintain room for manoeuvre should access conditions change, whilst avoiding, however, the creation of merely more complex forms of dependency.
The risk associated with the worsening of global inequalities, therefore, does not simply concern whether or not one has formal access to AI. It also concerns the ability to use it in a practical and concrete way, in accordance with the specific needs of each individual country and the various linguistic communities within it.
The model is available but incapable of adequately understanding the language of its users risks, in fact, widening – rather than narrowing – the gaps that already exist.
