What artificial intelligence generates fascination also generates fear. For Daniela Rus, who works with this technology, the verdict depends on us humans.
“It’s a tool that isn’t inherently good or bad, it’s what you decide to do with it. Like any other tool — like a knife, fire — it can be used in different ways, and we need to make sure it’s used for positive reasons. “, says Rus, professor of electrical engineering and computer science at the Massachusetts Institute of Technology (MIT), in the United States, as well as director of the Computer Science and Artificial Intelligence Laboratory (CSAIL).
“The more people participate in artificial intelligence, the more it democratizes, the more it can flourish.”
Rus acknowledges that the artificial intelligence research community “doesn’t understand every aspect of the technology, but is working hard to gain a deeper understanding and learn about its potential uses and obstacles.”
And he says that there are many who are dedicated to the development of tools against misinformation and against other things that “can go wrong with artificial intelligence”.
Rus feels optimistic. “(Artificial intelligence) can empower us in many ways: it can improve our memory, calculation speed, predictability, it can help us judge situations to make better decisions.”
Next, Daniela Rus and another technology enthusiast, engineer Mihaela van der Schaar, point out how artificial intelligence is already positively affecting our lives – from health to transportation.
“There’s a lot of debate about how artificial intelligence (AI) can replace humans, but I’m very interested in building an AI that doesn’t replace humans, but can make us, regardless of age, smarter. An AI for human empowerment “, points out Mihaela van der Schaar, professor of artificial intelligence and medicine at the University of Cambridge, England.
1. Diagnosis and treatment of diseases
The automatic learning (machine learning), which is a subfield of artificial intelligence, consists of using multiple data sources to map complex patterns.
“This is important, for example, so that personalized medicine allows the early diagnosis of diseases”, explains Mihaela van der Schaar, who directs a laboratory with her last name at the University of Cambridge.
She points out that machine learning has been used to analyze images, discover patterns and thus diagnose cancer, for example.
But the technology can be used long before a diagnosis is needed.
“Automatic learning is very good at identifying risk factors”, says the scientist.
And it is also useful afterwards, in the treatment of various diseases.
“Artificial intelligence can help us learn not just from different patients, but from different responses to various drugs.”
“Doing this mentally is very difficult for doctors. Even very smart doctors can’t integrate such diverse data sources, and they don’t have enough information in their hands.”
The doctor says that, in a joint effort with colleagues from the United States, United Kingdom and Holland, they realized that artificial intelligence can identify up to 24 hours before doctors that a patient will need to be admitted to an intensive care unit (ICU).
Another example of the use of this technology in medicine is that of Jean Tyler, 75 years old. She participated in a study called Colo-Detect, in the UK, which uses artificial intelligence to detect bowel cancer.
The technology marks a green box on the screen when it discovers areas that may be of concern to the specialist performing the colonoscopy.
These are areas that “the human eye can often miss”, according to representatives of the University of Newcastle, one of the institutions involved in the project.
Artificial intelligence detected in Tyler’s colonoscopy several polyps and an area with cancer. After undergoing surgery, the elderly woman had remission of the disease.
2. Find routes and market niches
“If you use Waze, Google Maps or any other navigation application, you are already using artificial intelligence, because all the statistics and forecasts to move from one place to another use this technology”, says engineer Daniela Rus.
Google has become one of the companies that most uses this tool.
Whether driving or walking, Google Maps helps users find a certain place “thanks to a system that learned to read street names and addresses from more than 1 billion Street View images”, explains the company in its blog .
In addition to transport, artificial intelligence is helping us in areas of the labor market, explains Rus.
“With artificial intelligence, we can get statistics on what customers are saying on social media, what they need and also what is needed in the supply chain.”
With a better understanding of consumers, it is possible to offer more personalized services, he argues. In addition, Rus points out that this distribution of tasks to machines does not despise human capabilities.
“With artificial intelligence robotics and machine learning technologies, we can delegate some routine tasks to machines. It really increases our efficiency and productivity, as well as allowing us to focus on more cognitive aspects like critical thinking and creativity.”
If you’ve used Google Translate, you’ve also used artificial intelligence. The tool was trained with content taken from the internet itself and is able to perform visual recognition of characters.
Rus explains that translation technologies are part of what is known as natural language processing or extensive language models.
This type of technology covers “a lot of data” – in text and not.
To help us better understand how it works, the specialist invites us to think in images.
If you want to have an artificial intelligence system that automatically recognizes the objects around you, such as a cell phone, a bookcase or a chair, you need to give the tool many examples of them.
With these registered examples – and at different angles – the automatic learning model is trained to recognize these objects.
In the case of generating and translating texts, what technology does is analyze examples of previous texts and their ways of ordering words, in order to later create new texts based on that.
“So, large language models now allow us to look at longer and longer strings of text, to generate increasingly complex strings.”