The productive sector is already using Artificial Intelligence (AI) as a mechanism to increase productivity
In 2025, services involving artificial intelligence should move US$ 36.8 billion, according to Intel estimates. Data show that in 2016, Google, Facebook and Microsoft invested about US$ 20 billion in the area. In Brazil, the numbers are more modest. This year, Sebrae, which produced a report on the subject, projects spending of US$ 182 million with AI.
Even with more timid investments, Brazil is the second country in the world that most uses Watson, IBM's artificial intelligence engine. “Brazilian companies have already understood that they are in a position to use AI to improve their business. It's not just the big ones, many startups use Watson”, highlights the tool's sales leader in Brazil, Roberto Celestino. About 20 types of industries use Watson today. “In the banking sector, Bradesco's BIA. In education, the Anhembi Morumbi universities are another example. Educational institutions use artificial intelligence for the relationship with the student or to help with the study journey”.
emotions
Watson is IBM's artificial intelligence engine launched in 2011. The computer faced the two biggest champions of Jeopardy (TV quiz show) and won the challenge.
The basic idea of introducing artificial intelligence is to help human beings in everyday tasks. “Watson, for example, is able to make better decisions than humans for productivity. He analyzes routine, has parameters and is not susceptible to emotional issues, at least not now. This facilitates better decision-making using artificial intelligence”, highlights the President of the Brazilian Industrial Development Agency (ABDI), Guto Ferreira.
Industry is one of the sectors that has stood out in the use of artificial intelligence. For Ferreira, the introduction of the tool is capable of increasing productivity.
The startup I.Systems, located in Campinas (SP), has been working with AI for the productive sector since 2007. They use a computer program to understand the routines of the industries and improve productivity, as one of the founders, Igor Santiago, explains. “Our system is configured, without interruption of production. After the configuration, the system continues to learn, and it adapts to generate even more savings”.
One of the main activities of the program is in boilers for generating energy in industries. In this field, the startup estimates that artificial intelligence generates savings of 2% to 10%. The boiler normally supplies energy to the plant's equipment. As there are many machines distributed over long distances, managing the amount of energy needed for each production moment is complex. “Our system is watching when the equipment is turned on and off. With that, we were able to increase the amount of energy generated by the boiler when we anticipate that a machine will start operating”.
One learning takes place with AI analyzing short-term data, another is done in the cloud – a place where content is stored on the internet –, when artificial intelligence gathers data from months or years to do this learning. “After learning from the production cycle itself, information is sent to the cloud. We apply algorithms to make improvements to large amounts of data,” explains Santiago.
Professor of software engineering at the University of Brasília (UnB) Sergio de Freitas explains that artificial intelligence works well in this type of competence because it achieves the same result as humans much faster. “Often, computers will provide information that only experts could reach. Through various ways she manages to verify what is correct, deviations from the correct. It's how a human being behaves, but at a higher speed.”
With this speed in mind, General Motors (GM) is implementing artificial intelligence for quality control at the Chevrolet plant in São Caetano do Sul (SP). The automaker's technology manager, Carlos Sakuramoto, explains that the AI will work on the car's paintwork review. “We identified that the perception of stamping was different among our inspectors. They undergo training so that everyone tries to see the same type of defect. The distortions are micrometric.”
As the automaker is global, there are inspectors around the world responsible for analyzing the stamping of the bodywork. According to Sakuramoto, there is no system that can assess human visual perception. “The customer, when he looks at a perfect print, something attracts his eye, even if he doesn't know what it is. That's why we want to standardize this assessment”.
Artificial intelligence was developed by startup Autaza. The technology-based company was created after GM looked to the university for intelligence to create a technology solution. “The three founding partners of Autaza were researchers on this project. We decided to undertake and take this application to other areas”, recalls Renan Padovani, one of the creators of the startup. Artificial intelligence works with a database. “The system has a neural network training. It learns the defects it should find from a defect database.” The system takes photos of the vehicle bodywork that are sent for analysis. The computer program is able to classify any problems.
Autaza estimates that there is a reduction of 60% in rework. “Often, the automaker's quality inspection errs the severity of the defect. In 60% of the cases they rate the problem as more serious. Rework becomes more labor intensive. With the correct classification of the error, the rework is less”, explains Padovani.
Another benefit is scanning throughout production, as GM's technology manager points out. “We will do a 100% analysis of production at a much faster rate. Previously, the inspection was done by sampling, withdrawing from production, or taking advantage of a stop. Now the system is continuous and in time, it guarantees production and quality gains”, points out Sakuramoto. Currently, the system is in the validation phase to gain global scale. After proven efficiency and safety, artificial intelligence for bodywork inspection should be installed in other plants by 2020.
The machines in power?
One of the main challenges of artificial intelligence is the security of the AI installation. GM is doing the tests in a closed system, the computers are not connected to the network. The strategy of breaking up content to ensure security is also used by IBM, as explained by Roberto Celestino. “Whenever someone creates an account on the IBM Cloud, that part belongs to that person. Everything created belongs to the person. I, as IBM, don't have access to that information unless you let me."
In addition to data security, another point of debate in artificial intelligence is the possibility of machine learning itself. Sci-fi movies like “The Terminator” and “The Matrix” painted apocalyptic futures involving AI. In the late 1980s blockbuster starring Arnold Schwarzenegger, Skynet – highly advanced artificial intelligence – evolved and saw humanity as a threat to its existence. The solution found by the machine was the annihilation of planet Earth. Is it possible for something like this to happen?
Experts differ on this possibility. Celestino points out that Watson, for example, only learns something it was programmed to learn. “Artificial intelligence has no will of its own. She must do what she was trained to do. In Watson's case, if I've trained to answer about a certain domain of knowledge, anything other than that, she won't be able to do”. The specialist also recalls that IBM and the other major developers have an ethical agreement. “The objective is that we can develop the technology thinking about the use in the most ethical way possible. We have to bring ease to the human being.”
UnB professor Sergio de Freitas recalls that this concern was already discussed in robotics. The development of robots must respect three laws: the machine cannot harm a human being, a robot must obey the orders given to it by human beings, and lastly, it must protect its own existence as long as such protection does not conflict. with the first or second premises. “I don't believe in this apocalyptic world. If we consider machine learning and it can learn from good intentions, I think it would learn good intentions too. Just like a child.”
ABDI's President, Guto Ferreira, already thinks differently. “Artificial intelligence already learns visually. So, as she sees aggressive attitudes, she may find that normal. Japan, with this in mind, makes household robot consumers sign a term. The idea is not to have aggressive attitudes around them”. But Ferreira also remembers that it can play a fundamental role in a competitive environment, where more and more productivity is sought.
Even with more timid investments, Brazil is the second country in the world that most uses Watson, IBM's artificial intelligence engine. “Brazilian companies have already understood that they are in a position to use AI to improve their business. It's not just the big ones, many startups use Watson”, highlights the tool's sales leader in Brazil, Roberto Celestino. About 20 types of industries use Watson today. “In the banking sector, Bradesco's BIA. In education, the Anhembi Morumbi universities are another example. Educational institutions use artificial intelligence for the relationship with the student or to help with the study journey”.
emotions
Watson is IBM's artificial intelligence engine launched in 2011. The computer faced the two biggest champions of Jeopardy (TV quiz show) and won the challenge.
The basic idea of introducing artificial intelligence is to help human beings in everyday tasks. “Watson, for example, is able to make better decisions than humans for productivity. He analyzes routine, has parameters and is not susceptible to emotional issues, at least not now. This facilitates better decision-making using artificial intelligence”, highlights the President of the Brazilian Industrial Development Agency (ABDI), Guto Ferreira.
Industry is one of the sectors that has stood out in the use of artificial intelligence. For Ferreira, the introduction of the tool is capable of increasing productivity.
The startup I.Systems, located in Campinas (SP), has been working with AI for the productive sector since 2007. They use a computer program to understand the routines of the industries and improve productivity, as one of the founders, Igor Santiago, explains. “Our system is configured, without interruption of production. After the configuration, the system continues to learn, and it adapts to generate even more savings”.
One of the main activities of the program is in boilers for generating energy in industries. In this field, the startup estimates that artificial intelligence generates savings of 2% to 10%. The boiler normally supplies energy to the plant's equipment. As there are many machines distributed over long distances, managing the amount of energy needed for each production moment is complex. “Our system is watching when the equipment is turned on and off. With that, we were able to increase the amount of energy generated by the boiler when we anticipate that a machine will start operating”.
One learning takes place with AI analyzing short-term data, another is done in the cloud – a place where content is stored on the internet –, when artificial intelligence gathers data from months or years to do this learning. “After learning from the production cycle itself, information is sent to the cloud. We apply algorithms to make improvements to large amounts of data,” explains Santiago.
Professor of software engineering at the University of Brasília (UnB) Sergio de Freitas explains that artificial intelligence works well in this type of competence because it achieves the same result as humans much faster. “Often, computers will provide information that only experts could reach. Through various ways she manages to verify what is correct, deviations from the correct. It's how a human being behaves, but at a higher speed.”
With this speed in mind, General Motors (GM) is implementing artificial intelligence for quality control at the Chevrolet plant in São Caetano do Sul (SP). The automaker's technology manager, Carlos Sakuramoto, explains that the AI will work on the car's paintwork review. “We identified that the perception of stamping was different among our inspectors. They undergo training so that everyone tries to see the same type of defect. The distortions are micrometric.”
As the automaker is global, there are inspectors around the world responsible for analyzing the stamping of the bodywork. According to Sakuramoto, there is no system that can assess human visual perception. “The customer, when he looks at a perfect print, something attracts his eye, even if he doesn't know what it is. That's why we want to standardize this assessment”.
Artificial intelligence was developed by startup Autaza. The technology-based company was created after GM looked to the university for intelligence to create a technology solution. “The three founding partners of Autaza were researchers on this project. We decided to undertake and take this application to other areas”, recalls Renan Padovani, one of the creators of the startup. Artificial intelligence works with a database. “The system has a neural network training. It learns the defects it should find from a defect database.” The system takes photos of the vehicle bodywork that are sent for analysis. The computer program is able to classify any problems.
Autaza estimates that there is a reduction of 60% in rework. “Often, the automaker's quality inspection errs the severity of the defect. In 60% of the cases they rate the problem as more serious. Rework becomes more labor intensive. With the correct classification of the error, the rework is less”, explains Padovani.
Another benefit is scanning throughout production, as GM's technology manager points out. “We will do a 100% analysis of production at a much faster rate. Previously, the inspection was done by sampling, withdrawing from production, or taking advantage of a stop. Now the system is continuous and in time, it guarantees production and quality gains”, points out Sakuramoto. Currently, the system is in the validation phase to gain global scale. After proven efficiency and safety, artificial intelligence for bodywork inspection should be installed in other plants by 2020.
The machines in power?
One of the main challenges of artificial intelligence is the security of the AI installation. GM is doing the tests in a closed system, the computers are not connected to the network. The strategy of breaking up content to ensure security is also used by IBM, as explained by Roberto Celestino. “Whenever someone creates an account on the IBM Cloud, that part belongs to that person. Everything created belongs to the person. I, as IBM, don't have access to that information unless you let me."
In addition to data security, another point of debate in artificial intelligence is the possibility of machine learning itself. Sci-fi movies like “The Terminator” and “The Matrix” painted apocalyptic futures involving AI. In the late 1980s blockbuster starring Arnold Schwarzenegger, Skynet – highly advanced artificial intelligence – evolved and saw humanity as a threat to its existence. The solution found by the machine was the annihilation of planet Earth. Is it possible for something like this to happen?
Experts differ on this possibility. Celestino points out that Watson, for example, only learns something it was programmed to learn. “Artificial intelligence has no will of its own. She must do what she was trained to do. In Watson's case, if I've trained to answer about a certain domain of knowledge, anything other than that, she won't be able to do”. The specialist also recalls that IBM and the other major developers have an ethical agreement. “The objective is that we can develop the technology thinking about the use in the most ethical way possible. We have to bring ease to the human being.”
UnB professor Sergio de Freitas recalls that this concern was already discussed in robotics. The development of robots must respect three laws: the machine cannot harm a human being, a robot must obey the orders given to it by human beings, and lastly, it must protect its own existence as long as such protection does not conflict. with the first or second premises. “I don't believe in this apocalyptic world. If we consider machine learning and it can learn from good intentions, I think it would learn good intentions too. Just like a child.”
ABDI's President, Guto Ferreira, already thinks differently. “Artificial intelligence already learns visually. So, as she sees aggressive attitudes, she may find that normal. Japan, with this in mind, makes household robot consumers sign a term. The idea is not to have aggressive attitudes around them”. But Ferreira also remembers that it can play a fundamental role in a competitive environment, where more and more productivity is sought.
EN 
PT 
