AI and Investing: The New Potential

  • |
  • 03 mins 33 secs
PIMCO's Mihir Worah discusses how new troves of financial data and faster computers have given PIMCO the means to apply artificial intelligence across a range of investment processes, from optimizing Treasury bidding to predicting mortgage prepayments.

Channel

PIMCO Canada


PIMCO is one of the world’s premier fixed income investment managers. With our launch in 1971 in Newport Beach, California, PIMCO introduced investors to a total return approach to fixed income investing. In the nearly 50 years since, we have worked relentlessly to help millions of investors pursue their objectives – regardless of shifting market conditions. As active investors, our goal is not just to find opportunities, but to create them. To this end, we remain firmly committed to the pursuit of our mission: delivering superior investment returns, solutions and service to our clients.

Visit PIMCO.ca

CMR2018-1002-357076

Visit PIMCO.ca

CMR2018-1012-358879

Speaker 0:

One of the big reasons sprinkles investing in a i artificial intelligence right now is simply the better availability of data in finance and the growth of computing power.


Speaker 0:

Essentially, artificial intelligence needs a lot of data, and it needs a lot of computing power. Let me give you an example.


Speaker 0:

It's been investing in mortgage backed securities for a number of years. We have a database of over two billion loans and data points on individual homeowners. But with the old techniques, we could analyze, at best, two percent off these two billion data points that we have.


Speaker 0:

So we've partnered with an outside artificial intelligence expert,


Speaker 0:

and even we've been working with them to analyze the data. And now we can analyze over ten to twenty percent of the data. And within a year, we already getting results that are better than our best models could deliver that we've built over decades.


Speaker 0:

In many ways, the way that farms like netflix and google is a is similar to the web. Pimco would use it, it's, taking large amount of data, analyzing them to try and predict what the future does. In some ways, what netflix and google have have to do is easier because the rules of the game a prescribed, given a certain amount of data given certain past behavior based on the data, the prediction of the future is somewhat simpler and more rule based financial markets. As you know, our self adaptive


Speaker 0:

once you do something market's changing, adapt. Also, data isn't as freely available and as clean and consistent in the financial markets. In the financial markets, you've gotta pick the right problem to analyze and then learn how to analyze it in a way that he dealing with the self adaptive nature of markets. So in some ways very similar, but some ways different, more difficult problems to tackle in finance.


Speaker 0:

An area of artificial intelligences car is an lp, a natural language processing. What many of the tech companies are interested is looking at. Users reviews on yelp


Speaker 0:

and using national natural language processing to come up with predictions on what they buy and what they wouldn't buy


Speaker 0:

were less interested in users reviews on yelp and, in fact, want to analyze companies quarterly and annual releases. Ten case and thank you's. And, for example, some of the researchers that we spoke to weren't even aware that there was this rich area of research when you could analyze ten queues and ten case and releases by companies, see what the market reaction is and try to come up with rules based on natural language processing off these documents much faster than what humans can do and try and try to understand them. Packs on markets. These are problems that interesting in finance that we're sharing with researchers and academics and hope to get results that we can use for a client's benefits.