Seeing Silicon | Concerns about society run high among AI researchers

Excitement, money, breakthroughs in AI and machine learning (ML) technology, whispers of new startups and job talks were palpable in the hallways of the Messe Wien Exhibition Congress Center, a glass-walled conference venue that was packed with thousands of students, researchers, scientists, professors and industry professionals. As were conversations about sightseeing in Vienna, attending jazz concerts and swimming in the Danube in the last week of July.

“We started as a tight-knit global community of about 500 researchers 20 years ago,” said Katherine Heller, a research scientist at Responsible AI at Google Research and the ICML 2024 program chair, who has been a regular at the AI ​​conference for two decades. She has spent the past few months reviewing submissions, creating schedules, inviting speakers, and deciding on workshops and tutorials for the weeklong event. “This year, we reviewed over 10,000 papers from 12,000 authors and selected about 2,600 to showcase,” she said, describing the growth of the ML community as “exponential.”

Before 2022, AI technology was largely confined to research projects in university labs and companies. When OpenAI released ChatGPT to the public in late 2022—more as a test than anything else—its rapid user adoption turned the mostly academic field into a gold mine. Overnight, ChatGPT became a household name, and companies working on generative AI and multilingual models sensed an opportunity—a general-purpose technology that could not only be used for other scientific research, but also execute code, search, and generate synthetic words, images, and videos for everyday use.

According to Bloomberg Intelligence, the market size for consumer generative AI is expected to grow from $40 billion in 2022 to $1.3 trillion by 2023. The market potential has since quickly poured talent, money, and more research into the field.

“AI research has the potential to have a huge impact on issues related to healthcare, education and climate, and its important research should not be left out in India,” said computer scientist Sandeep Juneja, who heads the Center for Data and Decision Science at Ashoka University and was at the conference to present a paper that aims to help understand new directions for research in this rapidly evolving, hot field.

The constant pursuit of profit and talent

The large corporate interest promotes and emphasizes a certain type of machine learning, while pushing other types of ML research into the background. The most common thematic keywords in the research papers were large language models (LLM), reinforcement learning, deep learning, and graph neural network – all studies focused on building generative AI and multimodal AI products for profit.

“The research community is at the forefront of what’s happening technically in the AI ​​universe,” said Sheila Gulati, managing director and founder of Seattle-based Tola Capital, an AI venture capital firm that attended the conference to find new directions for AI investments.

One of the 10 best-paper awards went to GENIE (Generative Interactive Environments), a program developed by researchers at Google DeepMind that demonstrated technology for creating synthetic virtual environments from text. Another award went to VideoPoet, an LLM that generated videos from scratch created by researchers at LumaAI, Google DeepMind, and Google Research.

“It takes money to do this research, scale it up and have an impact on more people,” said Heller, who was an academic before coming to Google to work on health systems. There’s a lot to be gained from having access to these kinds of resources in the industry, but you have to align your goals with what the for-profit company is most interested in at the time, she explained. “The industry is capitalistic, and researchers recognize that,” she said.

But there’s a flip side to this breathtaking pace. “Making sure that AI is delivering consistently drives research and stresses out researchers,” said Dr. Sunita Sarawagi, a professor and founder of the Center for Machine Intelligence and Data Science at the Indian Institute of Technology Bombay, who has been working in the field since 2003. “While it’s great that machine learning is having such a big impact, there are a lot of risks in using AI unwittingly, and we need to be careful,” she said, adding that because AI researchers are in such high demand in the industry, she finds it difficult to attract students to her center to do research.

AI/ML experts are in high demand not only by tech companies and startups but also by the financial industry, and academics like Sarawagi have to compete with everyone else. This was evident in the conference’s recruitment channel. Dozens of job offers popped up every day for PhD candidates, engineers, AI scientists and others, while recruiters actively interacted with students in the hallways. “I saw several interesting candidates with good research and programming skills,” said M Saquib Sarfraz, a technical manager at Germany-based Mercedes Benz Tech Innovation, whose one job posting for an engineer has already received more than 50 responses.

Matthew Leavitt, a California cofounder and chief scientist at Datalogy AI, sponsored the conference, hoping to brand his fledgling startup in the community. “The conference sponsorship and travel costs will be recouped if we hire even one candidate, because that’s less than what we pay a recruiter for an AI researcher,” he said, adding that there’s stiff competition (their booth is right across from a large Google installation) but that they’ve managed to generate interest because they’re working on data collection, a “frontier research problem” that’s an important research area for the data-hungry AI community.

Reflections on the Impact of Artificial Intelligence on Societies

As society understands the technology’s capabilities, owners of AI models are facing government regulation, copyright infringement lawsuits, and concerns about potential security threats. That has sparked a surge in interest in the technology’s ethical, social, copyright, and security implications. “There’s been a surge in workshops, researchers, research topics, groups, and labs that are thinking about the impact of AI on society, how to mitigate the negative impacts and amplify the positive impacts in nuanced ways,” says Weiwei Pan, associate director of Graduate Studies in Data Science at Harvard University, who attends a number of AI conferences each year.

As the data used to train all these models is increasingly contested, the next area of ​​research has become data—its sources, legality, preservation, and ethical issues surrounding its acquisition. One of the best-paper awards went to “Measure Dataset Diversity, Don’t Just Claim It,” a research paper that focused on data diversity and introduced social science principles to define dataset diversity.

“We need clearer definitions of terms like diversity, which can mean different things to different people,” said Dora Zhao, a Stanford University scholar and the paper’s lead author. Like Zhao, many researchers who are earning their Ph.D.s are increasingly focusing their work on the social implications of AI, which wasn’t the case a few years ago.

Another paper that also won the best paper prize featured two AI models debating each other to find the right answer for humans. “This makes us, the users, the judges, while two AI experts argue,” said Akbir Khan, the paper’s lead author, who is pursuing a PhD at University College London. Khan, who ran a startup before turning to research, believes that with the many AI models available today, it should be easier for humans to find the right answer among them all. His research is a first step in that direction.

Is artificial intelligence only for the rich?

Not all science is equal, and this is especially true for machine learning. Of the 10,000 attendees at this conference, half were from the United States and Western Europe, followed by China, with about 1,000. Apart from the conference costs, there was minimal representation from the Global South, with only 100 from India. This is due to the high costs of conducting AI research.

“The global South lacks the computing power, resources and capabilities to compete in the AI ​​space,” said Vukosi Marivate, a professor at the University of Pretoria in South Africa and co-founder of Africa’s largest research conference, Deep Learning Indaba, who gave an invited keynote at the conference. “There’s the United States, China, a bit of Europe, and then there’s India, which like everyone else, including Africa, has crumbs,” Marivate suggested that instead of pursuing profits, these countries should focus on the needs of their societies.

“I worry about companies overhyping AI,” Sarawagi said, adding that computer science researchers need to assess their community’s needs when it comes to the technology. “AI can help developed countries reduce their dependence on humans, but for us, it won’t solve social problems like living conditions, hunger or climate change,” she said. Instead of following the developed world, we should build our own systems, encourage local and relevant research and development, and change government policies.

Shweta Taneja is a Bay Area-based writer and journalist. Her biweekly column will reflect on how emerging technology and science are changing society in Silicon Valley and beyond. Find her online at @shwetawrites. The views expressed are personal.