When MIT and IBM launched a joint research lab in 2017, the New York-headquartered company pledged $240 million over a decade to chip away at the fundamental obstacles keeping artificial intelligence from transforming industries like healthcare and cybersecurity.
Now, the Cambridge, MA-based MIT-IBM Watson AI Lab is growing in personnel, funding, partners, and square feet, says David Cox, a former Harvard professor who leads the program for IBM (NYSE: [[ticker:IBM]]).
This week, the lab welcomed its first batch of outside partners. They include South Korea conglomerate Samsung, medical device company Boston Scientific (NYSE: [[ticker:BSX]]), construction tech firm Nexplore, and financial data provider Refinitiv. Cox said there is a healthy pipeline of other companies that may tap into the lab’s expertise.
 |
Want more AI, machine learning, and drug discovery content? Network and engage with industry leaders online this June. Learn more. |
The investment that each partner brings to the lab could lead the joint research venture to end up with double the $240 million it expected from the outset. The funds will help researchers work on making artificial intelligence more autonomous and easier to apply to real-world problems.
“We could imagine as much as doubling that investment over time as the program grows,” Cox said, declining to share financial figures for how much the partners are chipping into the endeavor. “We can substantially increase the scale of the investment we make in MIT.”
Cox explained, though, that the final list of partners won’t be too extensive.
“We don’t foresee it being an extremely large program,” he said. “There are other membership models where you pay a very small amount of money, and everybody is part of it. We really want to have a relatively small number in deep, deep engagements.”
In terms of research staff, the numbers are also higher than initially planned. The lab’s charter envisioned the equivalent of 100 full-timers. The reality now is the lab has 70 projects underway, and MIT and IBM both provide at least one staffer per project, Cox said. (Back in March 2019, Cox told Xconomy the lab had 49 research projects up and running.)
To accommodate the bigger team, the lab is moving a few blocks south to the heart of Kendall Square. The new home for the lab will be 314 Main Street, a 440,000-square-foot MIT building under construction that is expected to also be the home of the MIT Museum, the MIT Press Bookstore, and Boeing’s Aurora Flight Sciences research unit. Cox said work on the building will be completed next year.
Cox said one of the best mile markers for the lab’s progress is publications, and so far researchers have got 110 papers in journals.
The lab is likely to keep its focus on healthcare and cybersecurity, both local concentrations for IBM’s Watson business, Cox said. Teams will also continue their research into how artificial intelligence systems function at their core.
An appealing angle on fundamental AI research for the MIT-IBM teams has been something called neural-symbolic AI, which combined the popular deep learning technologies with symbolic reasoning techniques needed to learn abstract concepts and solve problems, Cox said. Combining the two could help develop more flexible AI systems that call for less “hand-holding” from humans to clean data, tweak algorithms, and carefully set up a framework for the AI system to explore.
The vision for neural-symbolic AI systems would be one that could think a lot more like humans think, need small sets of data to understand abstract concepts and be more transparent in the decision-making process.
“What’s interesting about that is it’s a progression where less and less is predefined,” Cox says, “and the system is more and more genuinely autonomous.”
