HÌNH ẢNH TRÍ TUỆ NHÂN TẠO

For the past three summers, around two dozen would-be computer scientists have sầu come lớn Stanford University khổng lồ learn boozeman.shop artificial intelligence from some of the field’s brighthử nghiệm. The attendees, culled from hundreds of applicants, take day trips khổng lồ nearby tech companies, interact with social robots và hexacopters, and learn boozeman.shop computational linguistics (what machines vì when words have multiple meanings, say) & the importance of time management (very). They play Frisbee. But if your mental picture of AI is a bunch of guys creating wilier enemies for their favorite videogames, well, this isn’t that. All the students here at the Stanford Artificial Intelligence Laboratory’s Outreach Summer (SAILORS) program are girls who have just completed ninth grade, and their studies focus on finding ways khổng lồ improve sầu lives, not enhance their game play: How vị we use AI khổng lồ keep jumbo jets from careening into one another? To ensure that doctors wash their hands before hitting the OR? “Our goal was khổng lồ rethink AI education in a way that encourages diversity and students from all walks of life,” says Fei-Fei Li, director of Stanford’s AI lab & a founder of the SAILORS program. “When you have a diverse range of future technologists, they really care that giải pháp công nghệ is being used for the good of humanity.”


“When you have a diverse range of future technologists, they really care that công nghệ is being used for the good of humanity.”

Fei-Fei Li Google & Stanford


*

SAILORS was created in 2015 by Li và former student Olga Russakovsky (now an assistant professor at Princeton University) to lớn help bring greater gender echất lượng to lớn the tech industry. The cause is both noble and urgent. According to lớn a recent survey, the number of women seeking computer science degrees is dropping; in the AI sector, women hold less than đôi mươi percent of executive sầu positions. It’s an enormous field to lớn be left out of, considering that, every day, more and more people use AI to make their lives easier và more efficient: AI is how pholớn apps recognize your face aý muốn everyone else"s, not khổng lồ mention the beach where you took the picture. It"s how your devices underst& you when you ask what the weather will be tomorrow. Then there are the lesser-known applications, like diagnosing diabetic retinopathy (which often leads to lớn blindness) or sending a drone on a search-and-rescue mission to the most remote reaches of the world.

Bạn đang xem: Hình ảnh trí tuệ nhân tạo

With AI becoming ever more ubiquitous, the need for gender balance in the field grows beyond just the rightness of the cause—diversity is a crucial piece of AI due to lớn the nature of machine learning. A goal of AI is khổng lồ prod machines lớn complete tasks that humans vị naturally: recognize speech, make decisions, tell the difference between a burrito lớn và an enchiladomain authority. To vày this, machines are fed vast amounts of information—often millions of words or conversations or images—just as all of us absorb information, every waking moment, from birth (in essence, this is machine learning). The more cars a machine sees, the more adept it is at identifying them. But if those data sets are limited or biased (if researchers don’t include, say, images of Trabants), or if the folks in AI don’t see or tài khoản for those limits or biases (maybe they’re not connoisseurs of obscure East German automobiles), the machines và the output will be flawed. It’s already happening. In one case, image recognition software identified photographs of Asian people as blinking.


“It’s not just boozeman.shop having transparency in data. We actually need to make the numbers move in the right direction.”

Tracy Chou Project Include


*

How vì humans create more inclusive labs and workspaces? A number of projects and individuals are taking on that challenge. This year, Li—who is also chief scientist of AI và machine learning at Google Cloud—và others helped launch AI4ALL. The national nonprofit is aimed at bringing greater diversity to lớn AI & has engaged experts in genomics, robotics, and sustainability as mentors. It’s building on the work of SAILORS but also targeting people of color và low-income students across the country through partnerships with Princeton, UC Berkeley, và Carnegie Mellon, in addition khổng lồ Stanford. “We had a lot of colleagues và industry leaders coming up to lớn us & saying, ‘SAILORS is great, but it’s just Stanford serving a few dozen students per year, mostly from the Bay Area,’ ” Li says. “So AI4ALL is boozeman.shop diversity và inclusion. It’s not only gender.”


AI và ML

What"s the difference?

The terms artificial intelligence (AI) and machine learning (ML) are often used interchangeably, but they’re not the same thing. AI describes machines’ ability to seemingly mimic human ways of thinking, learning as they go as opposed to following specific commands. ML is one of the most efficient—& popular—techniques that computers employ to gain that ability. In ML, machines sift through examples khổng lồ recognize patterns.


Other similar initiatives include Code Next, Google’s Oakland-based effort to encourage Latino and African American students khổng lồ explore careers in tech; DIY Girls, an educational và mentoring STEAM (science, giải pháp công nghệ, engineering, art, and math) program for under-resourced communities in Los Angeles; and Project Include, which helps new and midstage startups hire more women & people of color. Tracy Chou, formerly of Pinterest, founded Project Include last year with seven other prominent women in the tech industry. In 2013, Chou famously urged tech companies khổng lồ come clean boozeman.shop how many women they employed. As the numbers trickled in, they substantiated what everyone in Silicon Valley knew: The tech world, from the biggest corporation to the smallest startup, is overwhelmingly trắng and male. Project Include, says Chou, was the logical next step. “After a couple of years of these data reports coming out & not a lot of change happening, there started khổng lồ be a shift in the conversation,” she says. “Now it’s not just boozeman.shop having transparency in data. We actually need lớn make the numbers move in the right direction.”

That direction includes making work in the field of AI more accessible lớn the masses. There are relatively few people employed in AI, & already we’re seeing robots that care for people và personal assistants that anticipate our needs. With humans controlling the data and criteria and machines doing the work, better and greater human input đầu vào means better và greater results.

Xem thêm: Quán Cháo Ngon Ở Sài Gòn - Top 15 Quán Cháo Sườn Sài Gòn


*

In many ways, the democratization of AI is already on its way. Take this example: In nhật bản, a farmer’s son used AI lớn sort his family’s harvest of cucumbers by various characteristics. It’s the kind of story that appeals to lớn Li, who came lớn the US from China at age 16 knowing little boozeman.shop her adopted country & even less boozeman.shop New Jersey, where she ended up. After working a variety of odd jobs, from cleaning houses lớn walking dogs to lớn cashiering at a Chinese restaurant, Li found herself at Princeton, and later at graduate school at Caltech.

Li comes khổng lồ her work as a triple outsider: an immigrant, a woman, & a person of color in a world dominated by Trắng men. What might have sầu been obstacles for anyone else have sầu become prods for Li. She spends much of her time studying computer vision, a component of machine learning she calls “the killer tiện ích of AI.” Computer vision analyzes and identifies visual data và may someday help create more responsive sầu robotic limbs, say, or solve sầu the knottiest of mathematical proofs. But as with all AI, the key lớn this giải pháp công nghệ is teaching machines lớn unpaông chồng a wealth of information from different places and perspectives. To be, in essence, visual citizens of the world—not unlượt thích Li.

Fostering a diverse group of creators khổng lồ shape that world is essential lớn the sorts of story & technical issues that content strategist Diamãng cầu Williams encounters every day at ILMxLAB, the top-secret Lucasfilm dream center where developers craft immersive sầu, interactive sầu entertainment—a VR encounter with Darth Vader, perhaps—inspired by the vast Star Wars universe. Williams is deeply involved in pro-tech organizations like Blaông chồng Girls Code and remembers the dearth of women of color at her college in the ’80s. “I was always the only one in my math classes, the only one in my business classes,” she says. “That gets tiring, & it gets scary.” Her solution khổng lồ pointing more women toward tech: “Start them young and get them svào in their confidence, so that when they walk inkhổng lồ the room and they’re the only ones there, they don’t turn around.”


“Start them young & get them svào in their confidence, so that when they walk into the room and they’re the only ones there, they don’t turn around.”

Diamãng cầu Williams Lucasfilm


*

Maya Gupta, a machine-learning researcher at Google, is working to lớn improve AI, albeit from a different angle. At Stanford, she helped a Norwegian company detect cracks in its underwater gas pipelines. “You can’t go in there very well, so we had to use partial information to try to lớn guess,” she says. Teaching machines lớn make nuanced guesses is familiar terrain lớn Gupta. If you’re on YouTube listening to lớn tenor saxophonist Kamasi mê Washington’s “Truth” & the music effortlessly segues into lớn Alice Coltrane’s gorgeous “Turiya and Ramakrishna,” lượt thích the work of the smarthử nghiệm DJ you never knew, thank Gupta, whose team helps computers fine-tune their recommendations. “It’s all boozeman.shop predicting, right?” she says. “You’re trying to guess what’s going on with limited data.”

Today she’s leading a research and development team at Google lớn, among other things, create greater accuracy in machine learning. “Let’s say I want khổng lồ be equally accurate at identifying a Boston accent and a Texas accent, but I have sầu a speech recognizer that’s a little better at the Texas one,” she says. “Should I penalize the people with a Texas accent by making the recognition just as bad as it is for Boston, lớn be fair? And what if it’s simply harder lớn recognize people speaking with a Boston accent?”

Gupta and her team are also refining systems that would be infinitely more transparent than their carbon-based designers. With machines, the hope goes, we can eliminate many of the biases or subconscious processes that plague human thought—or at least more easily recognize them when they emerge. Machines don’t thua kém focus when they’re tired, or irritable, or hungry. A study showed that judges are less apt to lớn grant parole right before lunch, when they’re thinking of sandwiches rather than sidebars. “It’s hard lớn measure what’s really going on in the minds of humans,” Gupta says. “We want our machine-learning systems khổng lồ be explainable, & frankly many of them are already more explainable than humans are.”


“We want our machine learning systems khổng lồ be explainable, và frankly many of them are already more explainable than humans are.”

Maya Gupta Google


*

As AI becomes increasingly useful—not to lớn mention easier to lớn use—the push is on khổng lồ place it into lớn as many hands as possible. Christine Robson, an IBM researcher before coming to lớn Google, is an enthusiastic champion of open source software like TensorFlow, a machine-learning system that can be used for a host of tasks, from translating languages to spotting illnesses khổng lồ creating original art.

For Robson, inclusivity in AI means making its tools accessible to more than just self-professed math nerds lượt thích herself. “I’m excited boozeman.shop the availability of machine learning to the world,” she says. “We talk a lot boozeman.shop democratizing machine learning, but I am a big believer in this. Making these tools really easy to use, & making these techniques possible for everybody toàn thân lớn apply, is just so critical.”


Sci-fi literature and film have long proffered examples of AI gone awry (Mary Shelley’s Frankenstein turns 200 next year). Today, many in the industry—including Li, Robson, & Chou—are concerned less boozeman.shop what AI might vày to lớn us and more boozeman.shop what we humans might vì chưng to AI. An example: Programmers give virtual assistants a female voice because, well, men và women alike tend to lớn prefer one. “But it perpetuates this idea that assistants are female, so when we engage with these systems, it reinforces that social bias,” says Chou. Many of the field’s best minds worry boozeman.shop what’s going inlớn real-life AI systems—và thus what’s going lớn emerge. That’s where the push for greater diversity in AI comes in. Little of this will be easy. But its proponents are smart, resourceful, và committed to lớn the cause.


“Making these AI tools really easy to lớn use, and making these techniques possible for everyone to lớn apply, is just so critical.”

Christine Robson Google


We have to make sure that everyone feels welcome, Gupta says. She recalls the wall of photographs of retired electrical-engineering professors at her alma mater Rice that ‘’did not look lượt thích me.” We need to convince girls that AI isn"t magic, adds Robson. “It"s math."

At SAILORS, students are learning how khổng lồ use natural language processing khổng lồ tìm kiếm social truyền thông & aid in disaster relief. “It would help rescuers discover people in need in real time, using their Twitter messages,” Li says. The effects of the classes và projects last well past the unforgettable summers. Some of the students have sầu started their own robotics clubs at school, published pieces in scientific journals, & held workshops at middle schools lớn spread the gospel of AI lớn even younger girls. For these students, whose backgrounds and experiences are as diverse as the myriad projects they tackled at camp, AI isn’t the lademo cool gadget, but a powerful force for good. In the lead-up to lớn the first SAILORS gathering in 2015, the program shared messages from incoming campers, including this ambitious wish: “I hope khổng lồ begin my AI journey now so I can make an impact on the world in the future.”


Robert Ito is a writer based in Los Angeles. He is a frequent contributor to the Thành Phố New York Times, Salon, & Los Angeles magazine.