Questões de Inglês

Texto

 

AI Picks Up Racial and Gender Biases When Learning from What Humans Write

 

AI¹ picks up racial and gender biases_² when learning language from text, researchers say. Without any supervision, a machine learning algorithm learns to associate female names more with family words than career words, and black names as being more unpleasant than white names.

 

For a study published today in Science, researchers tested the bias of a common AI model, and then matched the results against a well-known psychological test that measures bias in humans. The team replicated in the algorithm all the psychological biases they tested, according to a study from co-author Aylin Caliskan, a post-doc at Princeton University. Because machine learning algorithms are so common, influencing everything from translation to scanning names on resumes, this research shows that the biases are pervasive, too.

 

An algorithm is a set of instructions that humans write to help computers learn. Think of it like a recipe, says Zachary Lipton, an AI researcher at UC San Diego who was not involved in the study. Because algorithms use existing materials — like books or text on the internet — it’s obvious that AI can pick up biases if the materials themselves are biased. (For example, Google Photos tagged black users as gorillas.) We’ve known for a while, for instance, that language algorithms learn to associate the word “man” with “professor” and the word “woman” with “assistant professor.” But this paper is interesting because it incorporates previous work done in psychology on human biases, Lipton says.

 

For today’s study, Caliskan’s team created a test that resembles the Implicit Association Test (IAT), which is commonly used in psychology to measure how biased people are (though there has been some controversy over its accuracy). In the IAT, subjects are presented with two images — say, a white man and a black man — and words like “pleasant” or “unpleasant.” The IAT calculates how quickly you match up “white man” and “pleasant” versus “black man” and “pleasant,” and vice versa. The idea is that the longer it takes you to match up two concepts, the more trouble you have associating them.

 

The test developed by the researchers also calculates bias, but instead of measuring “response time”, it measures the mathematical distance between two words. In other words, if there’s a bigger numerical distance between a black name and the concept of “pleasant” than a white name and “pleasant”, the model’s association between the two isn’t as strong. The further apart the words are, the less the algorithm associates them together.

 

Caliskan’s team then tested their method on one particular algorithm: Global Vectors for Word Representation (GLoVe) from Stanford University. GLoVe basically crawls the web to find data and learns associations between billions of words. The researchers found that, in GLoVe, female words are more associated with arts than with math or science, and black names are seen as more unpleasant than white names. That doesn’t mean there’s anything wrong with the AI system, per se, or how the AI is learning — there’s something wrong with the material.

 

¹AI: Artificial Intelligence

²bias: prejudice; preconception

Disponível em http://www.theverge.com/. Acesso em: 18/04/2017.

Why does Artificial Intelligence (AI) pick up biases?

Texto

 

AI Picks Up Racial and Gender Biases When Learning from What Humans Write

 

AI¹ picks up racial and gender biases_² when learning language from text, researchers say. Without any supervision, a machine learning algorithm learns to associate female names more with family words than career words, and black names as being more unpleasant than white names.

 

For a study published today in Science, researchers tested the bias of a common AI model, and then matched the results against a well-known psychological test that measures bias in humans. The team replicated in the algorithm all the psychological biases they tested, according to a study from co-author Aylin Caliskan, a post-doc at Princeton University. Because machine learning algorithms are so common, influencing everything from translation to scanning names on resumes, this research shows that the biases are pervasive, too.

 

An algorithm is a set of instructions that humans write to help computers learn. Think of it like a recipe, says Zachary Lipton, an AI researcher at UC San Diego who was not involved in the study. Because algorithms use existing materials — like books or text on the internet — it’s obvious that AI can pick up biases if the materials themselves are biased. (For example, Google Photos tagged black users as gorillas.) We’ve known for a while, for instance, that language algorithms learn to associate the word “man” with “professor” and the word “woman” with “assistant professor.” But this paper is interesting because it incorporates previous work done in psychology on human biases, Lipton says.

 

For today’s study, Caliskan’s team created a test that resembles the Implicit Association Test (IAT), which is commonly used in psychology to measure how biased people are (though there has been some controversy over its accuracy). In the IAT, subjects are presented with two images — say, a white man and a black man — and words like “pleasant” or “unpleasant.” The IAT calculates how quickly you match up “white man” and “pleasant” versus “black man” and “pleasant,” and vice versa. The idea is that the longer it takes you to match up two concepts, the more trouble you have associating them.

 

The test developed by the researchers also calculates bias, but instead of measuring “response time”, it measures the mathematical distance between two words. In other words, if there’s a bigger numerical distance between a black name and the concept of “pleasant” than a white name and “pleasant”, the model’s association between the two isn’t as strong. The further apart the words are, the less the algorithm associates them together.

 

Caliskan’s team then tested their method on one particular algorithm: Global Vectors for Word Representation (GLoVe) from Stanford University. GLoVe basically crawls the web to find data and learns associations between billions of words. The researchers found that, in GLoVe, female words are more associated with arts than with math or science, and black names are seen as more unpleasant than white names. That doesn’t mean there’s anything wrong with the AI system, per se, or how the AI is learning — there’s something wrong with the material.

 

¹AI: Artificial Intelligence

²bias: prejudice; preconception

Disponível em http://www.theverge.com/. Acesso em: 18/04/2017.

Assinale a alternativa que está de acordo com o texto.

Brazil stops demarcating land for indigenous people: ex-government agency official says

 

    Home to the world’s largest tropical forest, Brazil has lost about one fifth of the Amazon rainforest in the last 50 years, according to the World Wildlife Fund. Research by the U.S.-based World Resources Institute shows that deforestation rates on land formally owned by indigenous peoples are about 2.5 times lower than other areas since they are more likely to conserve the forest than other users.
    But politicians representing rural voters¹ oppose demarcating new territories for indigenous groups, saying the land ought to be used for farming or cattle ranching to boost economic growth in the recession-hit country. Marcio Santilli, former FUNAI president, said rural politicians were proposing new steps to the demarcation process, which would make it “virtually endless”. He also said they were also proposing that Congress, rather than FUNAI or the Justice Ministry, make decisions on which lands are demarcated to indigenous groups.
(Chris Arsenault. www.reuters.com, 11.01.2017. Adaptado.)
 

¹ Politicians representing rural voters: bancada ruralista no Congresso Nacional.

No trecho do segundo parágrafo “the land ought to be used for farming”, o termo em destaque pode ser substituído, sem alteração de sentido, por

Brazil stops demarcating land for indigenous people: ex-government agency official says

 

    Home to the world’s largest tropical forest, Brazil has lost about one fifth of the Amazon rainforest in the last 50 years, according to the World Wildlife Fund. Research by the U.S.-based World Resources Institute shows that deforestation rates on land formally owned by indigenous peoples are about 2.5 times lower than other areas since they are more likely to conserve the forest than other users.
    But politicians representing rural voters¹ oppose demarcating new territories for indigenous groups, saying the land ought to be used for farming or cattle ranching to boost economic growth in the recession-hit country. Marcio Santilli, former FUNAI president, said rural politicians were proposing new steps to the demarcation process, which would make it “virtually endless”. He also said they were also proposing that Congress, rather than FUNAI or the Justice Ministry, make decisions on which lands are demarcated to indigenous groups.
(Chris Arsenault. www.reuters.com, 11.01.2017. Adaptado.)
 

¹ Politicians representing rural voters: bancada ruralista no Congresso Nacional.

No trecho do primeiro parágrafo “they are more likely to conserve the forest”, a expressão em destaque tem sentido equivalente, em português, a

TEXTO:
 

How Smartphones Help Farmers in Rural India
    In India, small farmers cultivate 50 percent of the
land, but they are often held back by inefficient methods.
The goal of the development project I work on is to help
spread information on agriculture among farmers through
[5] information technology. In 2009, I traveled to Devarahati,
a village three hours from Bangalore in order to better
understand the situation.
    At first glance, Devarahati’s residents seem to use
little technology. In fact, few things in this poor community
[10] remind me of life in the 21^st century. No toilets are
available except for a foot-wide (0.3 meter) hole in the
ground. Drinking water comes from a well. However, the
pump for the well can only be used during a six-hour
period each day when electricity is available.
[15]     When one looks again, however, one sees signs of
20^th century developments. A few homes that look like
they were constructed a thousand years ago have satellite
dishes. Plastic garbage covers the ground, and in the
distance, two cell phone towers mark the age of mobile
[20] communication.
    Mobile phones allow residents of rural India to
communicate with their families in cities and obtain
information on market prices. They also provide
unexpected side benefits. As I am walking around the
[25] village, my translator Suma points out farmers wearing
earphones. According to her, they “don’t even have a SIM
card but just use their phones as music players.”
    Suma seems upset that people in her home
community use technology for such everyday activities
[30] as entertainment, but I’m thrilled. It is clear that people
are spending money on technology, charging their phones
when power is available, and using phones not designed
for illiterate people.
    For our project, the mobile phone is the most
[35] promising way to spread information throughout the
farming community. The application we have developed
for touch-screen phones lets literate and illiterate farmers
share information about prices, seeds, fertilizers, and
pesticides. Touch-screen technology combined with
[40] sound and video enables illiterate people to use digital
information.
    In a trial this summer, we hope to learn if and how
farmers in Devarahati will use this new technology in their
decision-making. For farmers who do not own TVs, the
[45] mobile phone will probably become a source of
entertainment and serve as a flashlight during power cuts.

However, we hope that farmers will adopt agricultural
innovations if trusted peers have had good experiences
with them. Our goal is to use the word-of-mouth approach
[50] that locals trust, rather than coming into communities
and telling them what to do. It’s just that we’re using
technology to make word-of-mouth bigger and better.
Disponível em: <http://voices.nationalgeographic.com/2012/06/05/mobilelearning- how-smartphones-help-illiterate-farmers-in-rural-india/>. Acesso em: 21 jul. 2016.

Based on the text, fill in the parentheses with T (True) or F (False).

(  ) “constructed” (l. 17) is closest in meaning to appeared.
(  ) “thrilled” (l. 30) is closest in meaning to excited.
(  ) “promising” (l. 35) is closest in meaning to hopeless.
(  ) “approach” (l. 49) is closest in meaning to method.

The correct sequence, from top to bottom, is

TEXTO:
 

How Smartphones Help Farmers in Rural India
    In India, small farmers cultivate 50 percent of the
land, but they are often held back by inefficient methods.
The goal of the development project I work on is to help
spread information on agriculture among farmers through
[5] information technology. In 2009, I traveled to Devarahati,
a village three hours from Bangalore in order to better
understand the situation.
    At first glance, Devarahati’s residents seem to use
little technology. In fact, few things in this poor community
[10] remind me of life in the 21^st century. No toilets are
available except for a foot-wide (0.3 meter) hole in the
ground. Drinking water comes from a well. However, the
pump for the well can only be used during a six-hour
period each day when electricity is available.
[15]     When one looks again, however, one sees signs of
20^th century developments. A few homes that look like
they were constructed a thousand years ago have satellite
dishes. Plastic garbage covers the ground, and in the
distance, two cell phone towers mark the age of mobile
[20] communication.
    Mobile phones allow residents of rural India to
communicate with their families in cities and obtain
information on market prices. They also provide
unexpected side benefits. As I am walking around the
[25] village, my translator Suma points out farmers wearing
earphones. According to her, they “don’t even have a SIM
card but just use their phones as music players.”
    Suma seems upset that people in her home
community use technology for such everyday activities
[30] as entertainment, but I’m thrilled. It is clear that people
are spending money on technology, charging their phones
when power is available, and using phones not designed
for illiterate people.
    For our project, the mobile phone is the most
[35] promising way to spread information throughout the
farming community. The application we have developed
for touch-screen phones lets literate and illiterate farmers
share information about prices, seeds, fertilizers, and
pesticides. Touch-screen technology combined with
[40] sound and video enables illiterate people to use digital
information.
    In a trial this summer, we hope to learn if and how
farmers in Devarahati will use this new technology in their
decision-making. For farmers who do not own TVs, the
[45] mobile phone will probably become a source of
entertainment and serve as a flashlight during power cuts.

However, we hope that farmers will adopt agricultural
innovations if trusted peers have had good experiences
with them. Our goal is to use the word-of-mouth approach
[50] that locals trust, rather than coming into communities
and telling them what to do. It’s just that we’re using
technology to make word-of-mouth bigger and better.
Disponível em: <http://voices.nationalgeographic.com/2012/06/05/mobilelearning- how-smartphones-help-illiterate-farmers-in-rural-india/>. Acesso em: 21 jul. 2016.

Some farmers are unable to benefit from the information about market prices that is now available through text messages because