In the fight to recapture artificial intelligence from the control of large tech companies

Among the richest and most powerful companies in the world, Google, Facebook, Amazon, Microsoft, and Apple have made artificial intelligence a core part of their business.The progress of the past ten years, especially in what is called Deep learning, Allowing them to monitor user behavior; recommend news, information and products to them; most importantly, target them with advertisements. Last year, Google’s advertising equipment generated more than $140 billion in revenue. Facebook created $84 billion.

These companies have invested heavily in technologies that have brought them such a huge wealth.Google’s parent company Alphabet acquires an artificial intelligence laboratory in London Deep thinking In 2014, he spent 600 million U.S. dollars and spent hundreds of millions of U.S. dollars each year to support its research. Microsoft signed a $1 billion agreement with OpenAI in 2019 to obtain commercial rights to its algorithms.

At the same time, technology giants have become large investors in university artificial intelligence research, seriously affecting their scientific focus. Over the years, more and more ambitious scientists have turned to work full-time for tech giants or adopt dual affiliations. A study by researchers showed that from 2018 to 2019, 58% of the most cited papers at the first two artificial intelligence conferences had at least one author belonging to a technology giant, compared to only a decade ago. Is 11%. Radical Artificial Intelligence Network, A group seeking to challenge the power dynamics of artificial intelligence.

The problem is that AI’s corporate agenda focuses on technologies with commercial potential, while largely ignoring research that helps to address challenges such as economic inequality and climate change. In fact, it makes these challenges worse. Driven by automated tasks, work costs are high, and tedious labor such as data cleansing and content auditing arises. The push to create larger models has led to a surge in the energy consumption of artificial intelligence. Deep learning has also created a culture in which our data is often crawled without consent to train products such as facial recognition systems. The recommendation algorithm exacerbated the political polarization, and the large-scale language model failed to eliminate misinformation.

It is this situation that Gebru and more and more like-minded scholars want to change. In the past five years, they have tried to shift their priorities in this field from simply enriching technology companies by expanding the number of people involved in the development of technology. Their goal is not only to reduce the harm caused by the existing system, but also to create a new, fairer and democratic artificial intelligence.

“Hello from Timnet”

In December 2015, Gebru sat down and wrote an open letter. During her PhD at Stanford University, she participated in the Neural Information Processing Systems Conference, the largest annual artificial intelligence research gathering. Of the more than 3,700 researchers there, Gebru counted only five blacks.

NeurIPS (now well-known) was once a small conference on niche academic topics, and soon became the biggest AI job bonanza of the year. The richest companies in the world come to show off presentations, host lavish parties, and write huge checks for the rarest people in Silicon Valley: skilled artificial intelligence researchers.

That year, Elon Musk came to announce this non-profit enterprise Open artificial intelligenceHe, Sam Altman, then president of Y Combinator, and Peter Thiel, co-founder of PayPal, invested $1 billion to solve what they believed to be a problem: the prospect of superintelligence that could one day take over the world. Their solution: build a better superintelligence. Of the 14 consultants or technical team members he appointed, 11 were white.

When Musk was worshipped, Gebru was dealing with humiliation and harassment. At a meeting, a group of drunk men in Google Research T-shirts circled her, asked her to hug her reluctantly, kiss her on the cheek, and took a photo.

Gebru gave a severe criticism of what she observed: spectacles, cult-like worship of artificial intelligence celebrities, and most importantly, overwhelming homogeneity. She wrote that the boy’s club culture has driven talented women out of the field. It also leads the entire community to form a dangerously narrow conception of artificial intelligence and its impact on the world.

She pointed out that Google has deployed a computer vision algorithm to classify blacks as gorillas. The increasing sophistication of unmanned drones has put the U.S. military on the path to lethal autonomous weapons. However, in Musk’s grand plan, these issues are not mentioned, that is, preventing artificial intelligence from taking over the world in certain theoretical future scenarios. “We don’t have to predict the future to see the potential adverse effects of artificial intelligence,” Gebru wrote. “It has already happened.”

Gebru has never published her reflections. But she realized that something needed to be changed. On January 28, 2016, she sent an email with the subject “Hello from Timnit” to five other black artificial intelligence researchers. “I have always felt sorry for the lack of color in artificial intelligence,” she wrote. “But now I have seen 5 of you 🙂 and thought it would be cool if we start a black person in the AI ​​group or at least know each other.”

This email sparked a discussion. What makes black people provide information for their research? For Gebru, her work is largely a product of her identity; for others, this is not the case. But after the meeting they agreed: if artificial intelligence is to play a greater role in society, they need more black researchers. Otherwise, the field will produce weaker science-its adverse consequences may become worse.

Profit-oriented agenda

As Black in AI Just beginning to converge, artificial intelligence is moving towards its commercial pace. According to the McKinsey Global Institute, that year, in 2016, technology giants spent approximately US$2-30 billion on the development of this technology.

Driven by corporate investment, this field is distorted. Thousands of researchers have begun to study artificial intelligence, but they mainly want to study deep learning algorithms, such as the algorithms behind large language models. “As a young PhD student who wants to find a job in a technology company, you will realize that technology companies are all about deep learning,” said Suresh Venkatasubramanian, a professor of computer science in the White House Office of Science and Technology Policy. “So you transfer all your research to deep learning. Then the next PhD student who comes in looked around and said,’Everyone is doing deep learning. I should do the same.'”

But deep learning is not the only technology in this field. Before its prosperity, there was a different artificial intelligence method called symbolic reasoning. Deep learning uses large amounts of data to teach algorithms about meaningful relationships in information, while symbolic reasoning focuses on explicitly coding knowledge and logic based on human expertise.

Some researchers now believe that these technologies should be combined. The hybrid approach will enable artificial intelligence to use data and energy more effectively, and give it expert knowledge and reasoning capabilities, as well as the ability to update itself with new information. However, when the most reliable way to maximize profits is to build larger models, companies have little incentive to explore alternative methods.