Spotlight: LatAm AI strategies require some coordination

Artificial intelligence technology will require coordinated international alignment and governance to deal with ethics, accountability and data processing aspects.

This was one of the key takeaways of the first Unesco Regional Forum on AI for Latin America and the Caribbean, held in São Paulo this week and co-organized by the Brazilian internet steering committee, CGI.

Some Latin American countries have been drafting national policies to assess AI risks and opportunities. But because data is often a transnational element and many aspects of AI permeate more than one market, harmonizing regional policies would be an appropriate move, Eclac's economic affairs officer, Valeria Jordan, said.

AI governance is at the top of policy-makers' agenda and around 30 countries have already put forward AI proposals.

The OECD recently launched its own set of AI principles and guidelines, focusing on ethics and non-biased algorithms.

Sarah Box, OECD senior counselor for science, technology and innovation said the group’s guidelines were adopted in May and that even eight non-OECD members are studying them.

“The goal is to ensure AI is innovative, trustworthy and respectful of human rights. We need a transparent and accountable approach to AI systems,” Box said at the event.

Meanwhile, the UN’s International Telecommunication Union (ITU) has since 2016 its “AI for Good” platform and summit. The latest G20 meeting also proposed principles to discuss the risk and opportunities of AI, making reference to the OECD document.

POLICIES TAKE SHAPE

Harmonizing public policies is also important because having different rules for each country could prove challenging to the operations of multinationals, which offer AI services in different parts of the globe.

The Brazilian minister of science, technology, innovation and communications Marcos Pontes - who at the event launched the public consultation for AI policy in Brazil - said he expects to extend cooperation with all neighboring countries.

Brazil is not alone is moving toward AI governance. Chile began working in January on its own AI policy. 

José Antonio Guridi, advisor to the Chilean minister of science, technology, knowledge and innovation, said a work plan is being followed and a public consultation will be held next year. The AI policy is expected for the first half of 2020.

"Automation should not be a problem to the labor market, but rather an opportunity," Guridi said, setting the tone of what should be one of Chile’s main AI policy concerns. He also said the government has already identified 40 artificial intelligence initiatives being worked out in ministries.

Concerned about data governance, Costa Rica created the first position of Chief Data Officer of a Latin American government.

According to Diego Montero, Costa Rica’s CDO and an advisor to the presidency, the agency aims to “instill a data-driven public policy-making culture.”

Montero added, Costa Rica has been able to gather socioeconomic data for about 80% of its citizens since 2015 and that the government is also considering an AI policy.

Canada is a leader in AI governance. The country’s pan-Canadian artificial intelligence strategy, or Cifar, became in 2017 the world's first AI strategy.

Jennifer May, Canada's ambassador to Brazil, said her country was experiencing a brain drain from researchers and scientists and the AI strategy turned out as a solution while also ensuring a set of principles for its use.

“By investing over Cdn$100mn (US$76mn) in a five-year program in partnership with the private sector we are now having a reverse brain drain and attracting foreign talents.”

Canada now has three AI centers: in Alberta, Montréal, and in Toronto. Employment in the tech sector went up 50% and investment 400% in the first year of the implementation of the AI strategy, she said.

EXCESSIVE REGULATION

A crucial part of AI governance is balancing ethics and the control of non-biased algorithms and a free enterprise economy. 

Some policy-makers in Europe believe it is be important to go beyond principles, by for example, empowering data protection agencies with computing auditing.

Wesley Vaz, a member of Brazil's federal audit court (TCU), thinks a public policy is necessary, but caution must be exercised to avoid “over-regulation.” In his opinion, an AI policy must be principles-based at best.

Still, one of the main concerns revolves around biased algorithms. This is because AI systems work in two ways: with probabilistic inputs, when algorithms go through the data and point to a trend; and deterministic, when they indicate a solution or a binary answer, such as yes or no.

“Because of these issues, AI systems must be able to explain very well, on backwards levels, the elements that led to a certain conclusion, deterministic or not,” said Fábio Rua, director at the Brazilian Association of Software Companies (ABES).

In the opinion of Carlos Affonso, director of the Rio-based Institute for Technology & Society (ITS), “we need to get out of the dichotomic approach of harms and benefits of AI, and the responsibilities, and first enlighten the end-user and the population about technology, what it does and how it works.’

In this regard, Edson Prestes, a member of the UN High-Level Panel on Digital Cooperation, said one of the panel's proposals is to reshape the internet governance forum so that it not only discusses the internet but also digital technologies, such as AI.

The US-China dispute has a clear intrinsic technological vector, and according to Prestes, the technology is not neutral but facilitates a “digital colonization” bias, so in this debate some countries are at risk of “becoming digital settlers.”

BRAZIL AI CENTERS AND CONSULTATION

At the event, minister Pontes signed an MoU with São Paulo research financing entity Fapesp for the creation of the first laboratory focused on the development of AI solutions. 

A total of eight such centers will be created in Brazil, four of which in partnership with Fapesp and focused on the priority areas set forth in Brazil’s national IoT plan.

Four other laboratories will focus on cybersecurity, defense, education, and management efficiency. His ministry also put to public consultation the country’s proposal for a national AI strategy. 

The objective is to collect inputs from society to draft a policy assessing risks and opportunities coming with the new technology. 

The proposed policy is inspired by European policies and the OECD’s guidelines, Pontes said. Telecom secretary Vitor Menezes said Brazil will have a "world standard" regulation.

The text is divided in six ”vertical” axes: education and training; workforce; research, development, innovation and entrepreneurship; government application; productive sector application; and public safety. 

And three “horizontal” axes include legislation, regulation and ethical use; international aspects; and AI governance.

 Photo credit: MCTIC