March 2, 2021

AI researchers rank the top AI labs worldwide


Google Deepmind head Demis Hassabis speaks during a press conference ahead of the Google DeepMind Challenge Match in Seoul on March 8, 2016.

Jung Yeon-Je | AFP |Getty Images | Getty Images

LONDON — Artificial intelligence researchers don’t like it when you ask them to name the top AI labs in the world, possibly because it’s so hard to answer.

There are some obvious contenders. U.S. Big Tech — Google, Facebook, Amazon, Apple and Microsoft — have all set up dedicated AI labs over the last decade. There’s also DeepMind, which is owned by Google parent company Alphabet, and OpenAI, which counts Elon Musk as a founding investor.

“Wow, I hate this question,” Mark Riedl, associate professor at the Georgia Tech School of Interactive Computing, told CNBC when asked to pick his standouts.

“Reputationally, there is a good argument to say DeepMind, OpenAI, and FAIR (Facebook AI research]) are the top three,” Riedl said.

AI investor Nathan Benaich, a partner at Air Street Capital, agreed. Google Brain and Microsoft could potentially be included in the top ranks, Benaich said, before adding that he believed Amazon and IBM weren’t quite in the same league when it comes to AI research output and impact.

Another AI expert, who asked to remain anonymous because he didn’t have approval from his company to speak publicly, told CNBC that DeepMind, OpenAI and FAIR were probably the top three pure AI research labs in terms of known funding, while IBM pushes out more patents. “The unknown question is the Baidus and Tencents of the world,” he said in reference to the Chinese tech giants.

Alphabet gives DeepMind hundreds of millions of dollars a year to carry out its work, while Microsoft invested $1 billion in OpenAI on top of the $1 billion that the founding investors contributed. FAIR’s funding is less clear because Facebook doesn’t break it down.

A.I.’s potential

One way to measure the impact of an AI lab is to look at how many academic papers it publishes at the two biggest AI conferences: NeurIPS and ICML.

In 2020, Google had 178 papers accepted and published at NeurIPS, while Microsft had 95, DeepMind had 59, Facebook had 58 and IBM had 38. Amazon had less than 30.

For the same year at ICML, Google had 114 papers accepted and published, while DeepMind had 51, Microsoft had 49, Facebook had 34, IBM had 19, and Amazon had 18.

PR vs reality

AI has been hailed as a technology that has the potential to bring about a new industrial revolution and dramatically change the world we live in. But, for now at least, it remains relatively nascent and “narrow” in its abilities — an AI that can play chess to a superhuman level doesn’t know how to make an omelet, for example.

DeepMind, OpenAI, and FAIR are widely perceived as the top three labs partly due to “strong PR games,” Riedl said. 

He believes that Microsoft, which carries out much of its AI work through Microsoft Research, could legitimately be included in the top ranks. “For whatever reason they fly below the radar sometimes,” Riedl said. “Salesforce, Amazon, IBM all have some really strong pockets of research but, again, fail to make big splashes.”

Riedl said he’s “not sure that you couldn’t swap any group of researchers from any of these companies with any other and make any difference.”

Neil Lawrence, the former director of machine learning at Amazon Cambridge, told CNBC that Amazon doesn’t have a large, centralized AI research lab because it’s more focused on bringing technology to customers.

“I would argue they’ve done that very successfully,” said Lawrence, who is now a professor of machine learning at the University of Cambridge. “But if you look at (academic) publications as a measure then they don’t rank.”

Lawrence said that Microsoft Research is personally the research lab that he admires the most but “Amazon really ranks up there in deploying AI … despite not having a (big) research lab.”

He added: “DeepMind, OpenAI and FAIR have definitely dominated the headlines. But it’s interesting how much of the research they are publishing might traditionally have been done in universities.”   



Source link