close

Insider

[Podcast] The role of Artificial Intelligence in the fight against Covid-19

Written by Alana Team
on August 26, 2020

In the first four episodes of the Inside Alana Podcast,  we explored the interdisciplinary power of Artificial Intelligence. Today, we will understand more about the AI development scenario worldwide, and how it can be applied to different areas. We will give special emphasis to the health sector, where AI has been used to combat the great evil of today: Covid-19.

In the fifth episode of the podcast, in addition to the participation of Marcellus Amadeus, CTO of Alana AI, we invited the researcher Dr. Alexandre Chiavegatto, director of the Big Data and Predictive Analysis in Health Laboratory (Labdaps) at FSP / USP. 

He also coordinates the IACOV-BR (Artificial Intelligence for Covid-19) network, which is developing Machine Learning algorithms for the diagnosis and prognosis of the disease in the country.

 

The topics covered in the chat were:

    1. The current stage of development of artificial intelligence
    2. The impact of AI on society
    3. Artificial Intelligence in Health
    4. Machine Learning
    5. Future of AI

What is the current stage of the development of artificial intelligence?

Within the field of artificial intelligence studies,  some research sub-areas are being developed in parallel. Alana AI's CTO, Marcellus, for example, is an expert in the language field, while Alexandre develops studies in the health field.

For both, the evolution of technology is notable in recent years. In July 2020, for example, the language area had a special milestone: the launch of the GPT-3 program, created by OpenAI, which has 175 billion deep learning parameters.

Although it only works in English, the new program is capable of generating excellent quality texts, writing poetry, creating websites, and HTML codes autonomously.

However, in the podcast, Marcellus and Alexandre contextualized the high cost for the development of AI, which causes governments and the private sector to invest billions to position themselves in this market.

“We need to talk about the cost of developing AI. On the one hand, there are data libraries and open source knowledge, through which anyone with time and access to the internet can learn about machine learning techniques. On the other hand, programs like GPT require high investment ”, analyzed Marcellus.

He highlighted estimates that indicate that millions of dollars have been spent for the development of the GPT-3. The high cost, in a way, means that few companies can reach extraordinary levels of technology - which makes the market even more concentrated.

The Race for Artificial Intelligence

The search for a Strong AI and the dispute over the development of AI applications are comparable to the space race in the Cold War, in the opinion of experts. Artificial intelligence has applications in several areas that are essential for the social and economic development of countries, and that is why great economic powers want to consolidate themselves in this market.

The United States is still the leader, as it continues to generate significant knowledge for the area. To give you an idea, a Stanford University proposal for the government of the country proposes to invest at least US $120 billion in the American AI ecosystem in the next ten years.

On the other hand, the USA has faced pressure regarding the impact of AI technologies on automated decisions, with implications related to human rights, for example. 

China, in turn, has set a goal of 5 years to become a leader in the field of artificial intelligence, in addition to investing heavily in research and training of professionals in the field.

There are also major investments by governments and companies in Canada and England, which hosts one of the world's leading AI companies, Deepmind (creator of the famous AlphaGo, algorithm, which defeated world champions in the Chinese game Go).

imagem_1_eng-min-3

The interesting thing is that even small advances in the area already have great impacts on the market, with consequences for companies in the sector, governments, and the population.

The impact of artificial intelligence on society

It is difficult to estimate the exact number of researchers and specialists in the field of artificial intelligence, but there has been a growth in events, conferences, courses, and investments dedicated to the area in recent years.

There is great enthusiasm for the subject today, which is evident by the promises and investments of the private sector, by the performance of universities and the search for professionals for courses in the area. MIT, for example, decided to launch a specific undergraduate course for the discipline. 

The trend is that the number of people working in the segment will increase more and more, which will generate more knowledge and dissemination of information.

Another important aspect of the advancement of this technology is the regulation of artificial intelligence, given emerging laws related to data protection and application of AI algorithms around the world. These regulations tend to directly affect the pace of investment and development of AI tools.

Artificial Intelligence and Health

One of the benefits of AI for society can be in the application in basic services, as in the health sector. In this segment, it can be used at different times:

  • Screening patients
  • Setting the service priority
  • Decisions about diagnoses and treatments
  • Health risk assessment
  • Determination of relevant factors for a patient's discharge

However, the use of patient data still generates discussions between scientists and the medical community, since health data is considered sensitive information. Although there is a lot of information available in hospital databases, for example, there are several legal restrictions on sharing and using this data. 

There are still other ethical considerations and precautions that need to be taken into account when it comes to the use of health data in AI tools. Among the main uncertainties are:

  • Algorithms can make wrong decisions, depending on the context and quality of the data
  • Results can be skewed since algorithms can reproduce prejudiced human patternsimagem_2_eng-min-2

IACOV-BR (Artificial Intelligence for Covid-19 in Brazil)

In the context of the pandemic, one of the examples cited by Marcellus and Alexandre during the podcast was the case of Hospital Albert Einstein, in São Paulo, which made publicly available anonymous data on patients with Covid-19. The goal is that scientists from around the world can take advantage of them in the development of diagnostic solutions for the disease.

Professor Alexandre even had early access to the data, in partnership with the hospital. In the episode, he tells about his role in the IACOV-BR initiative, which uses machine learning to diagnose cases of Covid-19. 

The project is part of the Faculty of Public Health, University of São Paulo (USP), and includes the participation of research groups from the following universities:

  • Federal University of Pelotas (UFPel)
  • State University of Paraíba (UEPB)
  • Oswaldo Cruz Foundation of Paraná (FIOCRUZ-PR)
  • Federal University of Goiás (UFG)

The network's objective is to develop and validate machine learning algorithms for the diagnosis and prevention of Covid-19. The process takes place in four stages:imagem_3_eng-min-1

In order to reach a conclusion, the algorithm continuously learns from the blood data of several patients and then defines characteristics and patterns between them.

IACOV-BR's contribution is interesting because the algorithm can determine, much more quickly, the probability that the patient contracted Covid-19.

“When the patient arrives at the hospital with symptoms and takes the RT-PCR test, the test result can take days or even weeks to be ready. We found that just with blood test data, which is ready within an hour, the algorithm can tell the risk and probability of the patient having Covid-19, and then it is easier for doctors to make decisions about isolation, tests, and preventive measures ”, contextualizes Alexandre.

If the patient is confirmed to be infected, the machine learning system can still weigh, according to the individual characteristics of the patient, factors such as:

  • Need for ICU admission
  • Need for mechanical ventilation
  • Probability of death according to risk factors

According to the USP researcher, the algorithm's success rate has been demonstrated to be high, and researchers continue to prepare scientific articles on the project.

The importance of context in health algorithms

As in the area of ​​natural language processing (NLP), context is essential for the proper functioning of health-oriented artificial intelligence algorithms. 

“Something we are going to discover with the network (IACOV) is the diversity of data that we need to develop good health algorithms. It is something we do not yet know. If we develop an algorithm to predict heart attacks in São Paulo, will it work in New York? Probably not, because the realities of the population are very different ”, ponders Alexandre.

Machine Learning

Throughout the chat, it was noted that there is a difference in the working models and construction of algorithms for the development of an AI, depending on the objective of the scientists and the area of application. 

Marcellus, who works in the development of artificial intelligence for Alana AI's marketing products, bets on the creation of machine learning algorithms with deep learning techniques -  which require little or no human supervision. 

For health systems, Alexandre highlighted the need to fully understand the decision of the algorithm, as this area deals with sensitive information and is, literally, vital. In these cases, he recommends combining machine learning with other more classic techniques to help scientists justify decision-making.

imagem_4_eng-min-1

Future of AI

The history of artificial intelligence shows that this is an area in constant evolution and that it can (and should) be used to positively impact society. 

Projects such as IACOV-BR make clear the importance of collaboration between hospitals by health agents and researchers in the area.

The more the technology evolves, the more techniques will emerge and can be incorporated into existing artificial intelligence systems. In other words, AI has an exponential growth path and can bring more and more benefits to companies and individuals.

To learn more about AI, listen to this, and the other episodes of the Inside Alana Podcast on the platform of your choice. And stay tuned in the work done by the IACOV-BR network in combating Covid-19.

linha-min-1

 

personagens escrevendo

 



assinatura_teamalana

 

Você também pode gostar:

Insider

[Podcast] The challenges of artificial intelligence in communication

For many, artificial intelligence exists only in films or conspiracy theories about the end of the world. Although it is...

Insider

Alana AI expands operations in Latin America

Founded in 2015, Alana AI is opening offices in Mexico and Puerto Rico. Today, the artificial intelligence company alrea...

Insider

What changes with the General Data Protection Law (GDPL) in Brazil?

Did you know that the General Data Protection Law was enacted in September this year and will take effect from the last ...

Receba insights sobre
Inteligência Artificial por e-mail,
na medida certa.