Sharing: The Growing Influence of Industry in AI Research

More & More People Going From Academia to Industry

Sik-Ho Tsang
4 min readMar 8



The growing influence of industry in AI research,
2023 Science, 59.921 Impact Factor (Sik-Ho Tsang @ Medium)
Artificial Intelligence, Deep Learning, AI, DL

  • For decades, artificial intelligence (AI) research has coexisted in academia and industry, but the balance is tilting toward industry as deep learning, a data-and-compute-driven subfield of AI, has become the leading technology in the field. E.g.: Industry’s AI successes are easy to see on the news.
  • This article is quite interesting about how AI impacts both the academia and industry, so I just made some review or summary on it.


  1. Industry’s Input Dominance
  2. The Increasing Dominance of Industry in AI Research
  3. Drawbacks

1. Industry’s Input Dominance

Percentage of US artificial intelligence (AI) PhDs hired by industry.
  • Demand for AI talent has grown much more quickly than supply over the past decade.

In 2004, only 21% of AI PhDs went to industry, but by 2020, almost 70% were.

Growth of US university AI research faculty hired by industry, with a reference line for the total size of computer science research faculty. * Data normalized to 2006.
  • Academic institutions are struggling to keep talent. This concern is not limited to US universities.

Computer science research faculty who specialize in AI have also been hired away from universities to work in industry.

The total number of model parameters (a rough proxy for compute) for image recognition on ImageNet

In 2021, industry models were 29 times bigger, on average, than academic models, highlighting the vast difference in computing power available to the two groups.

  • In 2021, nondefense US government agencies allocated US$1.5 billion on AI. In that same year, the European Commission planned to spend €1 billion (US$1.2 billion).
  • By contrast, globally, industry spent more than US$340 billion on AI in 2021, vastly outpacing public investment.
  • As one example, in 2019 Google’s parent company Alphabet spent US$1.5 billion on its subsidiary DeepMind, which is just one piece of its AI investment.
  • In Europe, the disparity is smaller but is still present; AI Watch estimates that “the private and public sector account for 67% and 33% of the EU AI investments respectively”.

2. The Increasing Dominance of Industry in AI Research

The proportion of papers at leading AI conferences that have at least one industry co-author.

Research papers with one or more industry co-authors grew from 22% of the presentations at leading AI conferences in 2000 to 38% in 2020.

The fraction of the largest AI models that are from industry (3-year rolling average).

Industry’s share of the biggest AI models has gone from 11% in 2010 to 96% in 2021.

Periods when the state-of-the-art model for leading AI benchmarks were from academia, industry, or collaborations
  • When looking across the six benchmarks in image recognition, sentiment analysis, language modeling, semantic segmentation, object detection, and machine translation — as well as 14 more that cover areas such as robotics and common sense reasoningindustry alone or in collaboration with universities had the leading model 62% of the time before 2017. Since 2020, that share has risen to 91% of the time.

3. Drawbacks

  • Industry’s commercial motives push them to focus on topics that are profit oriented. Often such incentives yield outcomes in line with the public interest, but not always.
  • Some researchers are concerned that we maybe on a socially suboptimal trajectory that focuses more on substituting human labor rather than augmenting human capabilities.

Industry domination of applied work also gives it power to shape the direction of basic research.

  • Even absent public alternatives to industry research, one might imagine that regulation, through auditing or external monitoring of industry AI, could be the solution
  • As academics would be unable to build the large models as they can access to dataset and with computing power, some important researches e.g.: for example, toxicity in AI generated language, and stereotyping, cannot be easily conducted.

It is also important to provide the resources to keep top AI researchers in academia.

  • For example, the Canada Research Chairs Program (CRCP), which provides salaries and research funds, has proven to be a successful means of attracting and retaining top talent in Canada.
  • Otherwise, important public interest AI work will be left behind.



Sik-Ho Tsang

PhD, Researcher. I share what I learn. :) Linktree: for Twitter, LinkedIn, etc.