by The enormous potential of AI to reshape the future has led to massive investments from industry in recent years. But the growing influence of private companies in the basic research driving this emerging technology could have serious implications for its development, researchers say.
The question of whether machines could mimic the intelligence of animals and humans is almost as old as computer science itself. Industry commitment to this research direction has fluctuated over the decades, lThis led to a series of AI winters where investment has flowed in and then back again as technology has done could not meet expectations.
However, the advent of deep learning at the turn of the last decade has resulted in one of the most sustained streams of interest and investment from private companies. This is now starting to spawn some truly groundbreaking AI products, but a new analysis in Science shows that industry is also benefiting from thiswrinkleGdominant position in AI research.
It’s a double-edged sword, say the authors. Industry brings with it money, computing resources, and vast amounts of data that have accelerated progress, but it also refocuses the entire field on areas of interest to private enterprise rather than those with the greatest potential or benefit to humanity .
“The commercial motives of the industry drive them to focus on profit-making issues. Often such incentives lead to outcomes that are in the public interest, but not always,” the authors write. “While this industry investment will benefit consumers, the accompanying research dominance should worry policymakers around the world, as it means public interest alternatives for key AI tools may become increasingly scarce.”
The authors show that the industry footprint in AI research has increased dramatically in recent years. In 2000, only 22 percent of presentations at leading AI conferences had one or more co-authors from private companies, but by 2020 the figure was 38 percent. But the effects are felt most clearly at the front of the pack.
Advances in deep learning have been driven in large part by the development of larger and larger models. In 2010, the industry made up just 11 percent of the largest AI models, but by 2021 it was 96 percent. This has coincided with a growing dominance of key benchmarks in areas such as image recognition and speech modeling, where industry participation in the leading model has increased from 62 percent in 2017 to 91 percent in 2020.
A key driver of this shift is the much larger investments that the private sector can make compared to public entities. Excluding defense spending, the US government earmarked $1.5 billion for spending on AI in 2021, compared to the $340 billion the industry spent globally that year.
This additional funding means far better resources — both in terms of computing power and data access — and the ability to attract the best talent. The size of AI models is highly correlated with the amount of data and computational resources available, and by 2021, industrial models were on average 29 times larger than academic models.
And while in 2004 only 21 percent of computer science doctorates specializing in AI went into industry, by 2020 it was almost 70 percent. The rate at which non-university AI experts are hired by private companies has also increased eightfold since 2006.
The authors point to OpenAI as a marker of increasing difficultyj Doing cutting-edge AI research without the financial resources of the private sector. In 2019, the organization transitioned from a non-profit to a “capped for-profit organization” to “quickly increase our investments in computers and talent,” the company said at the time.
This additional investment had its benefits, the authors note. It has helped bring AI technology out of the lab and into everyday products that can improve people’s lives. It has also led to the development of a variety of valuable tools used by industry and academia alike, such as software packages like TensorFlow and PyTorch, and increasingly powerful computer chips tailored for AI workloads.
But it’s also pushing AI research to focus on areas of potential commercial benefit for its sponsors, and just as important are data-hungry and compute-intensive AI approaches that fit well with the things big tech companies are already good at. As industry increasingly dictates the direction of AI research, this could result in competing approaches to AI and other societal beneficial applications being neglected without a clear profit motive.
“Given how broadly AI tools could be applied in society, such a situation would give a small number of technology companies tremendous power over the direction of society,” the authors note.
The authors say that there are models for closing the gap between the private sector and the public sector. The US has proposed creating a national AI research resource consisting of a public research cloud and public datasets. China recently approved a “national computing power network system”. And Canada’s Advanced Research Computing platform has been running for nearly a decade.
But without intervention from policymakers, the authors say, academics will likely be unable to properly interpret and critique industry models or offer public interest alternatives. Ensuring they have the skills to continue to help shape the frontiers of AI research should be a key priority for governments around the world.
Photo credit: DeepMind/Unsplash
