Wyckoff-Tweedie Photography
PIMCO Manny Roman told the Morningstar conference that machine learning isn't always a solution.
- Manny Roman, CEO of $1.7 trillion PIMCO, doesn't expect every technology project to go well - in fact, some of them will "miserably fail," he said at a recent financial conference.
- PIMCO is experimenting with machine learning, among other technology initiatives. The firm's found some things that work - mortgage models based on machine learning, for example - and some that don't, such as predicting metrics like the Consumer Price Index that aren't money makers.
- "You've got to be able to do things differently," Roman said.
- Visit Business Insider's homepage for more stories.
Even $1.7 trillion PIMCO is finding the limits of technology, including buzzy artificial intelligence.
The asset manager's CEO, Manny Roman, said the firm undertakes research and development with the expectation that some projects just won't work.
"I've got to be totally honest - some of the things we'll try are going to miserably fail. And that's OK," Roman said at last week's Morningstar Investment Conference in Chicago.
See more: PIMCO's CEO highlights 4 tech areas he's boosting and what keeps him up at night
PIMCO has long focused on data science. Earlier this month, the firm's chief technology officer told Business Insider how he's overseeing the construction of a central data platform to enable every employee to become a "citizen data scientist."
PIMCO's also seeking new ways to add alpha by using big data techniques, finding success for certain uses, like sorting mortgages via machine learning, Roman said.
In other cases, the firm's trying both human and machine learning efforts to shape data. For example, when trying to predict how the Federal Reserve will change interest rates, PIMCO consults global advisory board member and former Fed chairman Ben Bernanke, and just last week brought in Janet Yellen.
"We discuss what the Fed may do, depending on China, depending on business activities in the next six months," Roman said. "We're trying to predict possible outcomes. And it's hard."
On the data side, the firm's using machine learning to derive patterns from all of the Fed's press releases and conferences.
"I don't know which one is better ... but you've got to be willing to do things differently," Roman said.
Even when PIMCO can better forecast certain metrics, they're not always useful. For example, Roman said that if PIMCO could predict the Consumer Price Index before its release, it's not clear how that number could be integrated into the investment process.
"There are certain things that you could figure out but they may not be money makers," he said. "Sometimes you don't know what's going to add value for a while."
- Read more:
- The tech head at $1.7 trillion investment firm Pimco told us how a new tool is turning every employee into a data scientist
- BlackRock is quietly building a team of 30 data scientists to create a next-generation stock-lending platform
- Artificial intelligence is transforming a $22.9 trillion investing strategy - but the cutting-edge technology comes with a new set of problems
- JPMorgan's head of asset management explains how he's been able to poach workers from Silicon Valley
- BlackRock now has a higher percentage of technologists than JPMorgan, and it says a lot about the future of the money-management industry
- Investing startup Pagaya just raised $100 million in a bet that technology can reshape the consumer credit markets