Michael Kearns: Algorithmic Trading and the Role of AI in Investment at Different Time Scales
00:00:00.000 | You've worn many hats, one of which, the one that first caused me to become a big fan
00:00:05.840 | of your work many years ago is algorithmic trading.
00:00:09.680 | So I have to just ask a question about this because you have so much fascinating work
00:00:13.960 | there.
00:00:14.960 | In the 21st century, what role do you think algorithms have in the space of trading, investment,
00:00:22.080 | in the financial sector?
00:00:23.880 | Yeah, it's a good question.
00:00:26.480 | I mean, in the time I've spent on Wall Street and in finance, I've seen a clear progression
00:00:34.040 | and I think it's a progression that kind of models the use of algorithms and automation
00:00:38.560 | more generally in society, which is the things that kind of get taken over by the algos first
00:00:46.320 | are sort of the things that computers are obviously better at than people, right?
00:00:52.280 | So first of all, there needed to be this era of automation, right?
00:00:57.080 | Where just financial exchanges became largely electronic, which then enabled the possibility
00:01:03.400 | of trading becoming more algorithmic because once the exchanges are electronic, an algorithm
00:01:10.560 | can submit an order through an API just as well as a human can do at a monitor.
00:01:14.440 | It can do it really quickly, it can read all the data.
00:01:16.960 | And so, I think the places where algorithmic trading have had the greatest inroads and
00:01:23.980 | had the first inroads were in kind of execution problems, kind of optimized execution problems.
00:01:29.860 | So what I mean by that is at a large brokerage firm, for example, one of the lines of business
00:01:35.600 | might be on behalf of large institutional clients taking what we might consider difficult
00:01:42.440 | trades.
00:01:43.440 | It's not a mom and pop investor saying, "I want to buy 100 shares of Microsoft."
00:01:47.320 | It's a large hedge fund saying, "I want to buy a very, very large stake in Apple, and
00:01:53.640 | I want to do it over the span of a day."
00:01:56.680 | And it's such a large volume that if you're not clever about how you break that trade
00:02:00.400 | up, not just over time, but over perhaps multiple different electronic exchanges that all let
00:02:05.560 | you trade Apple on their platform, you'll push prices around in a way that hurts your
00:02:12.960 | execution.
00:02:14.400 | So this is an optimization problem.
00:02:16.800 | This is a control problem.
00:02:19.560 | And so, machines are better.
00:02:22.840 | We know how to design algorithms that are better at that kind of thing than a person
00:02:27.440 | is going to be able to do because we can take volumes of historical and real-time data to
00:02:33.040 | kind of optimize the schedule with which we trade.
00:02:35.720 | And similarly, high-frequency trading, which is closely related but not the same as optimized
00:02:42.240 | execution, where you're just trying to spot very, very temporary mispricings between exchanges
00:02:50.600 | or within an asset itself, or just predict directional movement of a stock because of
00:02:56.800 | the kind of very, very low-level granular buying and selling data in the exchange.
00:03:03.560 | Machines are good at this kind of stuff.
00:03:04.800 | It's kind of like the mechanics of trading.
00:03:07.600 | What about the... can machines do long-term sort of prediction?
00:03:12.920 | Yeah.
00:03:13.920 | So I think we are in an era where clearly there have been some very successful quant
00:03:19.960 | hedge funds that are in what we would traditionally call still in the stat arb regime.
00:03:27.440 | So-
00:03:28.440 | What's that?
00:03:29.440 | Stat arb referring to statistical arbitrage.
00:03:31.760 | But for the purposes of this conversation, what it really means is making directional
00:03:35.540 | predictions in asset price movement or returns.
00:03:40.480 | Your prediction about that directional movement is good for... you have a view that it's valid
00:03:47.300 | for some period of time between a few seconds and a few days.
00:03:52.700 | And that's the amount of time that you're going to kind of get into the position, hold
00:03:55.800 | it, and then hopefully be right about the directional movement and buy low and sell
00:04:00.220 | high as the cliche goes.
00:04:02.560 | So that is a kind of a sweet spot, I think, for quant trading and investing right now
00:04:09.700 | and has been for some time.
00:04:12.140 | When you really get to kind of more Warren Buffett style time scales, right?
00:04:18.260 | My cartoon of Warren Buffett is that Warren Buffett sits and thinks what the long-term
00:04:23.020 | value of Apple really should be.
00:04:25.860 | And he doesn't even look at what Apple is doing today.
00:04:28.740 | He just decides, I think that this is what its long-term value is and it's far from that
00:04:34.540 | right now.
00:04:35.540 | And so I'm going to buy some Apple or short some Apple and I'm going to sit on that for
00:04:40.980 | 10 or 20 years.
00:04:43.100 | So when you're at that kind of time scale or even more than just a few days, all kinds
00:04:50.040 | of other sources of risk and information...
00:04:54.580 | So now you're talking about holding things through recessions and economic cycles.
00:04:59.620 | Wars can break out.
00:05:01.300 | So there you have to understand human nature at a level that...
00:05:04.300 | Yeah, and you need to just be able to ingest many, many more sources of data that are on
00:05:09.020 | wildly different time scales, right?
00:05:11.580 | So if I'm an HFT, I'm a high-frequency trader, like I don't...
00:05:17.700 | My main source of data is just the data from the exchanges themselves about the activity
00:05:22.160 | in the exchanges, right?
00:05:24.220 | And maybe I need to pay...
00:05:25.620 | I need to keep an eye on the news, right?
00:05:27.860 | Because that can cause sudden...
00:05:30.100 | The CEO gets caught in a scandal or gets run over by a bus or something that can cause
00:05:36.060 | very sudden changes.
00:05:38.020 | But I don't need to understand economic cycles.
00:05:41.500 | I don't need to understand recessions.
00:05:42.940 | I don't need to worry about the political situation or war breaking out in this part
00:05:47.700 | of the world because all I need to know is as long as that's not going to happen in the
00:05:52.780 | next 500 milliseconds, then my model is good.
00:05:57.620 | When you get to these longer time scales, you really have to worry about that kind of
00:06:00.940 | stuff.
00:06:01.940 | And people in the machine learning community are starting to think about this.
00:06:04.700 | We held a...
00:06:06.600 | We jointly sponsored a workshop at Penn with the Federal Reserve Bank of Philadelphia a
00:06:12.020 | little more than a year ago on...
00:06:14.220 | I think the title was something like, "Machine Learning for Macroeconomic Prediction."
00:06:19.740 | Macroeconomic referring specifically to these longer time scales.
00:06:23.580 | And it was an interesting conference, but it left me with greater confidence that we
00:06:32.980 | have a long way to go to...
00:06:34.980 | And so I think that people that...
00:06:37.100 | In the grand scheme of things, if somebody asked me like, "Well, whose job on Wall Street
00:06:42.220 | is safe from the bots?"
00:06:44.060 | I think people that are at that longer time scale and have that appetite for all the risks
00:06:49.020 | involved in long-term investing and that really need kind of not just algorithms that can
00:06:54.820 | optimize from data, but they need views on stuff.
00:06:57.420 | They need views on the political landscape, economic cycles and the like.
00:07:03.900 | And I think they're pretty safe for a while, as far as I can tell.
00:07:07.860 | So Warren Buffett's job is safe for a little while.
00:07:09.500 | Yeah, I'm not seeing a robo Warren Buffett anytime soon.
00:07:13.500 | Should give him comfort.
00:07:14.540 | [End of Audio]
00:07:15.540 | Duration: 2 minutes
00:07:16.540 | 1
00:07:16.540 | 2
00:07:17.540 | 3
00:07:18.540 | 4
00:07:19.540 | 5
00:07:20.540 | 6
00:07:21.540 | 7
00:07:22.540 | 8
00:07:23.540 | 9
00:07:24.540 | 10
00:07:25.540 | 11
00:07:26.540 | 12
00:07:27.540 | 13
00:07:28.540 | 14
00:07:29.540 | 15
00:07:30.540 | 16
00:07:31.540 | 17
00:07:32.540 | 18
00:07:33.540 | 19