The Limited Times

Now you can see non-English news...

Artificial intelligence saves lives (and counts elephants)

2021-06-27T16:26:05.449Z


Machines that see, hear and learn alone are among us, sometimes inadvertently. They save time and are solving bigger problems, strategic for our future. This is a tour of some of his latest and surprising contributions.


Part of the article you are about to read has been typed. Since it is not signed by her, which could happen, what does the previous statement refer to? Because artificial intelligence (AI) has converted an entire hour of conversation into text in just a few minutes. The dream of any journalist raised in the tedious task of transcribing interviews for much longer than the meeting lasted. But this system that has recognized the sound of the motley human language, has digested it and has written a series of punctuated sentences with meaning is only a small star in the ubiquitous galaxy of that discipline that sees and hears without eyes and ears, that learns automatically. and that uses patterns similar to human neural networks to execute tasks as complex as winning Go,diagnose diseases with similar dexterity to that of doctors or search the vast jungle of the internet. It is said by a man who has worked all his life in this field, Ramon López de Mántaras, former director of the National Center for Artificial Intelligence: “AI is already providing very interesting results to the rest of the sciences; medicine, biology, chemistry, ecology… ”. By the way, the machine that translated this phrase turned "science" into "holm oaks." Nobody is perfect.the machine that translated this phrase turned "science" into "holm oaks." Nobody is perfect.the machine that translated this phrase turned "science" into "holm oaks." Nobody is perfect.

Engineer Nuria Oliver, co-founder of the European Laboratory for Learning and Intelligent Systems, sounds dramatic when she says: "I don't think we can survive as a species without it." IBM's research director, Ruchir Puri, assures that the only thing stopping the deployment of this “21st century electricity”, as defined by another eminent scientist, Andrew Ng, “is that society trusts the decisions made by the algorithm or the recommendations that it offers, if they are biased or not, or if the data with which it is fed were adequate ”. A hot topic, biases, but that's another story.

When we talk about AI we don't mean Hal, the conscious supercomputer from

2001: A Space Odyssey.

What surrounds us is a basic or "narrow" branch, not very intelligent, so to speak, focused on solving specific problems.

But enormously complex and important.

So, guided by these experts, let's see the last thing AI can do for us.

More drugs, please.

The contribution of AI to medicine is enormous, from the well-known diagnosis of diseases through image recognition - Google has just announced an application capable of identifying up to 238 skin conditions with an expertise similar to that of dermatologists - to predict cardiac arrest after learning 72 facts from 133,000 patients. Last year, $ 13.8 billion (about € 11.2 billion, almost five times more than the previous year) was invested in AI for drugs, drug discovery and cancer, according to Stanford University's annual artificial intelligence index. It is the field with the greatest economic injection. "The application of AI to medicine and transportation will change us as a society," says Puri."In the last year and a half, the discipline has accelerated drug discovery."

Against the coronavirus.

The sudden jolt of the pandemic needed quick responses. One initiative invited scientists to submit proposals for antivirals, and PostEra, a British AI start-up, evaluated at least 10,000 inputs with machine learning tools to analyze the ease with which the compounds could be manufactured and devised routes to synthesize them. “I did it in less than 48 hours. Human pharmacists would have taken between three and four weeks to carry out the same task ”, highlights the Stanford index. IBM also offered a similar tool to aid in drug research.

ValenciaIA4COVID, the team co-led by Nuria Oliver, created a system to predict the behavior of the epidemic in 236 countries and regions of the world within 180 days, applying different confinement and public health policies. “In different settings, for example the educational one, degrees of restrictions were combined, from teaching the classes remotely until they were all face-to-face, from university to primary school. The goal was to achieve the fewest number of infections with the highest economic activity possible. We share these simulations with the Generalitat Valenciana ”, he explains. They were the world winners of the international competition XPrize Pandemic Response Challenge.

Another question that plagues physicians in this crisis is how patients will evolve. A New York hospital fed a deep learning neural network with X-rays and clinical data from 3,661 patients to automatically predict a patient's risk of deterioration within 96 hours. The AI ​​system generated accurate predictions in real time, according to a paper published by the researchers in

Nature

this month. The Hospital Clínic de Barcelona and the Supercomputing Center of the Catalan capital are treating 300 types of data from more than 4,000 coronavirus patients with the same purpose. "And that it serves for other diseases," says the center's chief of medical informatics, Xavier Pastor.

How can I help you?

Not just at home (Alexa) or in your pocket (Siri).

We have a lot of conversations with machines, says IBM's head of research.

AI “is involved in customer service for almost every company around the world;

even when there is a person attending, an algorithm is behind guiding them.

Many times you end up talking to a virtual assistant equipped with conversational technologies.

It is the most widespread application ".

Save the animals.

African elephants face extinction from deforestation and poaching. Its population has decreased by 60% in the last 50 years. How to curb this disastrous trend? From the outset, counting the pachyderms, with the consequent risk of error and danger for humans. Now an algorithm accompanied by high-resolution imaging technology enables a satellite to scan large areas, up to 5,000 square kilometers, in a short time. The AI ​​developed by British researchers processes with "image recognition the presence of elephants in huge areas," says López de Mántaras. “It can be counted with much more precision, you can track in real time how their number is evolving. It gives an idea of ​​how versatile and horizontal AI is,useful for a multitude of completely different things ”, he emphasizes.

A medical robot created by the Chinese company Tami Robotics.Alex Plavevski (Efe) / EFE

A credit without being a customer.

There are 1.7 billion people in the world without a bank account, explains Oliver.

“But most have basic mobiles.

And the AI ​​can elaborate from the data that the cell phones draw on the user's behavior a risk model so that they can obtain a loan.

It is used in developing countries, including Vietnam, Bangladesh, Rwanda or Colombia ”.

This is how financial inclusion is achieved.

Keys of science.

If there is a crucial milestone - published at the end of last year - it is the one that solves one of the biggest problems in biology: predicting the shape of proteins in space, those structures that keep us alive. AlphaFold2, a neural network developed by DeepMind, a Google subsidiary, managed to successfully guess from the plane of the protein - its linear sequence of amino acids - how it is configured in three dimensions, "something that determines its function", explains López de Mántaras . "Biologists are delighted," he continues, "they say it will change medicine, research, everything." Misfolded proteins are behind ailments such as Alzheimer's, cystic fibrosis or mad cow disease. Unraveling this second code of life, carried out so far at tremendous cost and time,it is essential to tackle diseases and synthesize drugs. A similar advance is the resolution of a famous equation in quantum chemistry, that of Schrödinger, using a neural network capable of learning the complex shapes of how electrons are arranged around the nucleus. It will be very important to create new materials, believes the AI ​​expert.

See supernovae.

Professor López de Mántaras is fascinated by the window that AI opens to the universe. "With machine learning it has been possible to identify and classify supernovae, make a map of the universe, the galaxies, and also find exoplanets," he explains. There is a system that, going down to the pixel level, labels stars and galaxies in databases of astronomical observations. Another algorithm classifies supernovae without the traditional use of spectra based on their visible characteristics. And a team was able to confirm the existence of 50 planets outside the Solar System after training a neural network with other previously discovered exoplanets. It is usually done in an observational way and with computational analysis.

Feed us.

In an overpopulated world, which has lost half of its arable land in the last half century, ensuring food is key.

"AI is present in precision agriculture, automatic pest detection and crop optimization," explains Oliver.

Crops and soils are monitored with sensors and images taken with drones, and the impact of meteorological changes is predicted through machine learning.

What will the weather be like tomorrow?

Weather forecasting, says Professor López de Mántaras, “is based on very complicated mathematical models that require a lot of computing power in supercomputers.

With AI, predictions can already be made using a computer 7,000 times less powerful than those used now ”.

Source: elparis

All news articles on 2021-06-27

You may like

News/Politics 2024-03-23T00:35:34.817Z

Trends 24h

Latest

© Communities 2019 - Privacy

The information on this site is from external sources that are not under our control.
The inclusion of any links does not necessarily imply a recommendation or endorse the views expressed within them.