Artificial intelligence policy has no budget – 7/1/2023 – Tech

While trillionaire companies compete for leadership in the AI ​​(artificial intelligence) market, Brazil has not edited anything other than Ebia (Brazilian Artificial Intelligence Strategy).

Since 2021, Ebia has defined guidelines based on public consultations for the development of this technology in the country. Three transversal axes deal with ethical, governance and geopolitical issues, while another six verticals deal with research, innovation and workforce preparation.

The government, however, does not dedicate its own budget to foster the development of AI in Brazil, unlike the US, China and South Korea.

For USP professor Glauco Arbix, who coordinates the Innovation Observatory, Ebia fails even to define goals or focal sectors of action. “It’s innocuous, it sets the guidelines, but it doesn’t cite numbers or goals.”

According to the strategy, innovation in artificial intelligence should receive support via parliamentary amendments, opening of public calls for bids by Finep (Financier of Studies and Projects), CNPQ (National Council for Scientific and Technological Development) and Embrapii (Brazilian Research and Industrial Innovation).

Another way is the promotion of the ICT Law, which involves BRL 500 million this year for various areas of information and communication.

Since taking over the MCTI (Ministry of Science, Technology and Innovation), Luciana Santos has not indicated that she would change Ebia. Researchers and technology entrepreneurs consulted by the report claim that the public policy was formulated by career officials of the ministry, at the time commanded by the current senator Marcos Pontes (PL).

Today, the leadership of the Secretariat for Technological Development and Innovation is vacant.

A Sheet, the deputy secretary of Science and Technology for Digital Transformation, Henrique de Oliveira Miguel, says that there are plans to stimulate discussions on the subject, in view of the emergence of new artificial intelligences. “It’s urgent because of the economic opportunities and the risk.”

He claims that the vacancy of seats has to do with the government’s political difficulties in nominating and nominating positions in the face of pressure from Congress. In this context, career officials carry out the day-to-day activities of the portfolio, under the direct guidance of the minister.

Miguel says, however, that the theme was given prestige with the creation of the secretariat he commands. Previously, it was part of the Secretariat for Technological Development and Innovation.

He also claims that the Ministry of Science, Technology and Innovation has an articulating role with other ministries, which would explain the area’s low budget.

CGEE (Center for Management and Strategic Studies) technical advisor Caroline Pereira, who participates in other MCTI consultative councils, says that Ebia involves the topic of discussion that arouses the most interest among representatives of civil society. “Everyone is very engaged, carrying out research, but the government is calling the meetings.”

According to Marina Atoji, director of programs at Transparência Brasil, the entity that leads discussions on the use of AI in Public Safety within Ebia, the Ministry of Justice never sent representatives to the meetings.

There are no ordinances that define the frequency of meetings. Between December 2018 and December 2020, there was no meeting. In 2021, there were two and, last year, one.

According to Glauco Arbix, from USP, Brazil has invested little in innovation since the government of Michel Temer (MDB). “You have to make up for lost time in the last six years, in which there were technological advances in the world, and AI is under the control of large companies and countries like the USA and China.”

“Brazil hasn’t woken up yet, I hope it wakes up now under Lula’s government. It needs focus, priority and investment, starting with the public sector. Without this, research will not be left behind, companies, the economy, trade will be left behind. and even agribusiness”, adds Arbix.

He recalls that the country lost the electronics boom in the 1970s, which led to development in South Korea and other East Asian countries.

The most important projects in the area are the CPAs (applied research centers) in artificial intelligence, promoted by MCTI, Fapesp (São Paulo State Research Support Foundation) and CGI.br (Internet Management Committee).

Each of them receives support for five years, which can be renewed at the end of the term. Fapesp can contribute with up to R$ 1 million annually, supplemented with an amount of up to R$ 1 million paid by partner companies.

USP’s C4AI (Center for Artificial Intelligence), for example, receives R$ 1 million from FAPESP and R$ 1 million from the IT multinational IBM.

The amounts, however, are modest compared to investments made by China and the United States.

Only the investment of the US Department of Defense in artificial intelligence in 2023, US$ 1.8 billion (R$ 8.7 billion), exceeds the entire budget of the Science and Technology portfolio in Brazil – R$ 8.3 billion.

The average spending on research and development around the world was 2.13% of GDP in 2013, according to World Bank data. In Brazil, the last number, for 2020, indicated a ratio of 1.14%.

Investing in artificial intelligence at the beginning of the race is crucial for the country to be competitive in the future, according to the coordinator of C4AI and professor of computer science at USP, Fábio Cozman.

“Brazil has a lot going for it: a large, homogeneous population that mostly speaks the same language, the government holds structured data, many agencies produce quality data. It has the potential to lead some sectors and be a prominent country in the area” , says Cozman.

He highlights the financial, health and agribusiness areas as key sectors for the development of AI in Brazil.

Artificial intelligence tools are not limited to text generators like ChatGPT. Banks use AI to decide who to lend to, farmers to interpret satellite images and doctors to make diagnoses.

Still, another concern is to create quality data for Brazilian Portuguese and the country’s indigenous languages. Current natural language models handle English and Mandarin much better than other languages.

“The information cataloged in Portuguese, in the world, is on par with what is in French, German and Italian”, says Marcelo Finger, who is also a professor of computer science. According to him, there is a “healthy dispute” between the Portuguese and Brazilians about who contributes more in the area.

The quality of textual and spoken data is fundamental to train current artificial intelligence models. Finger states that few well-treated data are as or more important than the large masses of information scraped from the internet and also used in training.