[ad_1]
Among the richest and most powerful companies in the world, Google, Facebook, Amazon, Microsoft, and Apple have done the basics of the AI business. Advances in the last decade, especially in the so-called AI technique deep learning, have given permission to control user behavior; recommend news, information and products; and especially targeted with ads. Last year, Google’s advertising apparatus generated more than $ 140 billion in revenue. Facebook generated $ 84 billion.
Companies have made huge investments in technology that has brought them so much wealth. Google’s main company, Alphabet, bought a London-based AI lab DeepMind In 2014 he spent $ 600 million and spends hundreds of millions a year to support his research. Microsoft signed a $ 1 billion deal with OpenAI in 2019 for the rights to market its algorithms.
At the same time, technology giants have become major investors in university-based AI research, and have a major impact on scientific priorities. Over the years, more and more ambitious scientists have gone on to work for technology giants or have taken on dual affiliation. From 2018 to 2019, 58% of the most cited articles in AI’s two major conferences were affiliated with a technology giant that had at least one author, up from just 11% a decade earlier, according to research by researchers. AI Radical Network, A group that wants to question the power dynamics in AI.
The problem is that AI’s corporate agenda is based on techniques that have commercial potential, leaving aside research that can largely address challenges such as economic inequality and climate change. In fact, these challenges have worsened. The drive to automate tasks has cost jobs and increased tedious labor, such as clearing data and moderating content. The impetus for the creation of ever larger models has exploded in AI’s energy consumption. Deep learning has created a culture where our data is constantly crawled, often without permission, to train products like facial recognition systems. And recommendation algorithms have exacerbated political polarization, and large language models have not cleaned up misinformation.
Gebru and the growing movements of scholars with ideas want to change this situation. Over the past five years, they have tried to change the industry’s priorities only by enriching technology companies by expanding those involved in technology development. Their goal is not only to mitigate the damage caused by existing systems but also to create a new, fairer and more democratic AI.
“Hello from Timnit”
In December 2015, Gebru sat down to write an open letter. After earning his doctorate from Stanford, he attended a conference on the Neural Information Processing Systems system, the largest annual AI research meeting. With more than 3,700 researchers there, Gebruk counted only five who were black.
With a small meeting on an academic niche topic, NeurIPS (as it is now known) was fast becoming the biggest bonanza of AI work of the year. The richest companies in the world came to show off the demos, have weird parties, and write hard checks to the weirdest people in Silicon Valley: skilled AI researchers.
That year Elon Musk came to announce a nonprofit project OpenAI. He, Sam Altman, president of the Y Combinator, and Peter Thiel, co-founder of PayPal, put up $ 1 billion to solve what they believed was an existential problem: a superintelligence solution that could one day take over the world. Their solution: build even better superintelligence. Of the 14 advisors or technical teams he anointed, 11 were white men.
RICARDO SANTOS | COURTESY PHOTO
While Musk was being lionized, Gebru was dealing with humiliation and harassment. At the conference party, he was rounded off by a group drunk on Google Research T-shirts and received unwanted hugs, a kiss on the cheek and a photo.
Gebru wrote a scathing critique of what he observed: the show, the cult worship of famous AI, and most of all, the utter homogeneity. The culture of this boy’s club he wrote drove talented women out of the field. Moreover, the whole community was moving towards a narrow dangerous view of artificial intelligence and its impact on the world.
Google has already rolled out a computerized visual algorithm that classifies blacks as gorillas, he noted. And the growing sophistication of unmanned drones was putting U.S. weapons on the path to a deadly autonomous weapon. But Musk made no mention of these issues in some future theoretical scenarios for AI to stop the world from taking over. “We don’t need to project into the future to see the potential detrimental effects of AI,” Gebru wrote. “It’s already happening.”
Gebru never published his reflection. But he realized that something had to change. On January 28, 2016, he sent an email with the topic “Hello from Timnit” to five other black AI researchers. “I’ve always been sad about AI’s lack of color,” he wrote. “But now I’ve seen 5 of you 🙂 and I thought if we started a black on the AI team or at least got to know each other it would be cool.”
The email sparked discussion. What was it about being black that informed their research? For Gebro his work was a very product of identity; for others it was not so. After the meeting they agreed: if AI was to play a greater role in society, they needed more black researchers. Otherwise, the field would create a weaker science and its detrimental consequences could be exacerbated.
Profit-driven agenda
As Black AIn as it began to strengthen, AI was taking its commercial step. In 2016, technology giants spent approximately $ 20 billion to $ 30 billion developing technology, according to the McKinsey Global Institute.
With corporate investment heated, the pitch tilted. Thousands more researchers began learning AI, but mostly wanted to work on deep learning algorithms, such as those behind large language models. “As a young doctoral student who wants to get a job in a technology company, you realize that technology companies have deep learning,” says Suresh Venkatasubramania, a computer science professor currently working in the White House Science and Technology Policy Office. . “So you take all the research into deep learning. The next PhD student who comes next looks around and says, ‘Everyone is doing deep learning.’ I should probably do it too. ‘
But deep learning is not the only technique in this area. Prior to its rise, there was a different view of AI known as symbolic reasoning. While in-depth learning uses massive amounts of data to teach algorithms about meaningful relationships in information, symbolic reasoning is focused on explicitly codifying knowledge and logic based on human knowledge.
Some researchers believe that these techniques should be combined. The hybrid approach would make AI more efficient in the use of data and energy, and would give it the ability to know and reason with an expert, as well as the ability to update with new information. But companies have little incentive to explore alternative approaches when the safest way to maximize profits is to build larger models.
[ad_2]
Source link
