Every great measure of progress in computing – from servers to personal computers, the Internet and smartphones – has offered opportunities for more people to innovate on the digital frontier. However, there is growing concern that this trend is being reversed with the new main edge of technology: artificial intelligence.
Computer experts say AI research is becoming increasingly expensive, requiring complex calculations by giant data centers, so fewer people have access to the computational power needed to develop the technology behind futuristic products such as autonomous vehicles or digital assistants that can see, talk and reason.
The danger, they point out, is that the pioneering investigations of artificial intelligence will be a field of privileged and dispossessed people. The privileged will be mainly a few large technology companies such as Google, Microsoft, Amazon, and Facebook, which will invest billions a year building their data centers.
In the field of the dispossessed, they warn, there will be university laboratories, which have traditionally been a source of innovations that end up driving new products and services. “The enormous computing resources these companies have to pose a threat because universities cannot compete,” said Craig Knoblock, executive director of the Institute of Computer Science, a research lab at the University of Southern California.
Warnings from research scientists come amid growing concerns about the power of big technology companies. Most of the approach has been based on the generation of today’s technology: search engines, online advertising, social networking, and e-commerce. However, scientists are concerned that there is a barrier to exploring the technological future when an enormous amount of computer information is required.
The modern data centers of large technology companies are extensive and airtight. Buildings are the size of soccer fields, or larger, and there is shelf after shelf with hundreds of thousands of computers. The doors are bulletproof. The walls are fireproof. Rarely is anyone from outside allowed to enter them.
These are the machine rooms of cloud computing. They offer a cornucopia of entertainment and information available on cell phones and laptops and allow millions of developers to program cloud-based software applications.
However, artificial intelligence researchers, outside the big technology companies, see a worrying trend in their field. A recent report from the Allen Institute of Artificial Intelligence observed that the volume of calculations needed to be a leader in AI tasks, such as language understanding, video games, and common sense reasoning, has increased nearly 300,000 times in the last six years.
All that computer fuel is needed to fuel the deep learning software models, whose performance improves with more calculations and more data. Deep learning has been the main driver of AI innovations in recent years.
“When it succeeds, there are enormous benefits,” said Oren Etzioni, executive director of the Allen Institute, founded in 2014 by Paul Allen, the billionaire co-founder of Microsoft. “However, the cost of conducting research is rising exponentially. As a society and economy, we suffer if there are only a handful of places where you can be at the forefront.
The evolution of OpenAI, an artificial intelligence lab, shows the changing economy as well as the promise of deep learning AI technology.
Founded in 2015, with the support of Elon Musk, OpenAI began as a non-profit research laboratory. Its ambition was to develop technology on the frontier of artificial intelligence and share the benefits with everyone. It was a vision that hinted at the computer tradition of an inspired programmer working alone on a laptop and coming up with a great idea.
This spring, OpenAI used its technology to beat the world champion team of human players who beat a complex video game called Dota 2. Their software learned the game through a constant process of trial and error over months, the equivalent of more than 45,000 years of playing it.