Because the research and development capabilities of artificial intelligence (AI) technology are concentrated in firms, there are concerns that the event of AI technology for the general public interest will step by step shrink.
The Financial Times (FT) identified on the twenty third (local time) that while generative AI technologies and application cases resembling ChatGPT are pouring out, power is being targeting AI firms. Researchers and policy makers are concerned concerning the possibility that AI technologies will only find yourself being developed commercially, as fewer individuals or firms control resources and knowledge.
Experts call the concentration of AI research and development capabilities in a small number of enormous firms “industrial capture.” The concept originated from a paper titled “Increasing Impact of Industry in AI Research” published within the journal Science by researchers on the Massachusetts Institute of Technology (MIT) on the 2nd.
The paper showed data showing that industry dominates three key elements of contemporary AI research: computing power, large datasets, and highly expert researchers. The proportion of Ph.D. graduates who studied AI got jobs in firms rose from 21% in 2004 to about 70% in 2020. The variety of AI professors employed by firms has increased eight-fold since 2006.
“Most of the researchers we talked to gave up on their research because they felt they couldn’t compete with the industry,” said Nur Ahmed, the paper’s writer. He concluded that large-scale language models, resembling GPT-4 specifically, can’t be built by scholars.
It’s because the technology requires huge amounts of knowledge and computing power that only large tech firms like Google or Microsoft (MS) have access to. Ahmed found that the enterprise share of the most important AI models increased from 11% in 2010 to 96% in 2021.
The paper also showed that there are significant differences between private and non-private investment in AI technologies. In 2021, US government agencies, excluding the defense sector, have allocated $1.5 billion (about 1.95 trillion won) for AI, and the EU Commission planned to spend 1 billion euros (about 1.4 trillion won). However, the private sector has invested greater than $340 billion (about 441 trillion won) in AI in 2021.
As AI R&D capabilities are focused on for-profit firms, AI technologies and applications are increasingly more likely to be developed in a industrial direction. Alex Hanna, director of the Distributed AI Lab, noted that “not many firms want to develop AI in fundamental fields resembling biodiversity assurance, climate science, or agriculture.”
Ahmed suggested an alternate: the federal government should step in and establish data centers exclusively for academics where researchers could conduct experiments. Nonetheless, the FT explained that this plan will eventually only increase the dominance of personal cloud firms.
Nonetheless, researchers and policy makers alike agree that they can not close their eyes to the issue.
Jeong Byeong-il, member jbi@aitimes.com