1
00:00:00,330 --> 00:00:04,530
Thank you, and thank you for inviting us, and it's lovely to be part of this audience,

2
00:00:04,530 --> 00:00:10,770
and I'll tell you a little bit about what I do and some of my colleagues do around A.I. in finance.

3
00:00:10,770 --> 00:00:18,870
And when I do this with my MBA students, I always begin by asking them what is finance about in one word?

4
00:00:18,870 --> 00:00:28,410
What do you think finance is about? If you had to describe it in one word and everyone who has not worked in finance says it's about money.

5
00:00:28,410 --> 00:00:34,430
But everyone who is working or has worked in finance is something else.

6
00:00:34,430 --> 00:00:44,390
And guess what, how Brits risk finance is about risk, the everyday life of finance practitioners is really all about risk,

7
00:00:44,390 --> 00:00:50,330
is about measuring risk, about quantifying the risk and then about pricing the risk.

8
00:00:50,330 --> 00:00:54,500
For example, if I ask the bank for a loan,

9
00:00:54,500 --> 00:01:00,200
the bank then uses whatever information they have about me and they come up with

10
00:01:00,200 --> 00:01:06,200
a probability or a number that describes the likelihood that I would pay back.

11
00:01:06,200 --> 00:01:11,720
And then based on that number, they can attach some kind of a decision whether to give me a loan or not,

12
00:01:11,720 --> 00:01:19,280
or in some cases, also a rate that goes with that loan depending on regulation and time place.

13
00:01:19,280 --> 00:01:24,050
And they can do that because of course, I'm one person and I will pay yes or no.

14
00:01:24,050 --> 00:01:30,020
But if there's lots and lots of people like me, that individual risk translate to some kind of distribution and it's like a casino.

15
00:01:30,020 --> 00:01:36,110
They can come up with a pricing that would compensate them for the risk that they take.

16
00:01:36,110 --> 00:01:40,130
So this is really what what's happened in finance in most aspects of finance.

17
00:01:40,130 --> 00:01:48,080
Something like what I described. But this is really where technology and I can have a big impact because the

18
00:01:48,080 --> 00:01:55,130
bank in this case is trying to estimate the likelihood that I will pay back.

19
00:01:55,130 --> 00:02:05,240
So the bank doesn't know as much as I do. So if they could go inside my head, they will have a better idea of likelihood that they will pay back.

20
00:02:05,240 --> 00:02:09,800
But if they can go into this, they will have a much better idea.

21
00:02:09,800 --> 00:02:16,160
I would pay back because this knows better than me the likelihood that I would pay back any law.

22
00:02:16,160 --> 00:02:19,780
OK. And I know it sounds like you're excited by that.

23
00:02:19,780 --> 00:02:23,270
Too scared by this. But this is this is this is what we've seen.

24
00:02:23,270 --> 00:02:28,640
And if you remember the everybody remember the scandal, which Facebook and the election,

25
00:02:28,640 --> 00:02:34,220
but it came from a project that started by Facebook, although they don't like to talk about it, called my personality.

26
00:02:34,220 --> 00:02:38,390
It started as a Facebook project with the Cambridge people before they started off,

27
00:02:38,390 --> 00:02:45,890
where they looked at the correlation between Facebook likes, which at that time was a public of the publicly available information.

28
00:02:45,890 --> 00:02:52,280
And what we call personality is measured by the Big Five. And this is the standard way of measuring facade.

29
00:02:52,280 --> 00:02:55,490
And if you remember the result, they found that if you have enough likes,

30
00:02:55,490 --> 00:03:00,880
then the algorithm can predict your personality better than your best friends.

31
00:03:00,880 --> 00:03:05,440
In fact, they came up with this number, which I think was 200, but I'm not sure that with 200 likes,

32
00:03:05,440 --> 00:03:11,380
the algorithm can predict with more accuracy than your partner could think about.

33
00:03:11,380 --> 00:03:14,950
You know, the person who lives with you, they can predict better.

34
00:03:14,950 --> 00:03:21,730
So clearly, the information is phenomenal and and this is both revealing we can stop here because you can see both the

35
00:03:21,730 --> 00:03:31,420
the opportunity and the challenges around algorithm making decisions for us and using the technology.

36
00:03:31,420 --> 00:03:39,490
Now this using technology using information via the internet and other devices is not new.

37
00:03:39,490 --> 00:03:42,370
It's new in finance, but it's been used in other areas.

38
00:03:42,370 --> 00:03:51,820
So, for example, in the travel industry, you know, until eight, eight years ago, something like that,

39
00:03:51,820 --> 00:03:56,020
if you travel somewhere, you know, you could have a bad hotel, you could have a good hotel.

40
00:03:56,020 --> 00:03:59,890
There was lots of variability. There is still variability, but a lot less.

41
00:03:59,890 --> 00:04:03,880
Yes, it's very unlikely that you'll go to a place and it will be completely,

42
00:04:03,880 --> 00:04:10,060
completely horrible because of TripAdvisor, because people share the information and the places know that.

43
00:04:10,060 --> 00:04:13,630
So they know that nobody will come if they get really, really bad trips.

44
00:04:13,630 --> 00:04:20,830
So the reviews on TripAdvisor, so that that sort of is a way of aggregating information and reducing risk.

45
00:04:20,830 --> 00:04:27,580
And maybe even a better example of that is with the insurance industry. And so if you are young drivers or if you have children who are young drivers,

46
00:04:27,580 --> 00:04:32,410
you're all familiar with this thing called insurance and insured the box or the box, the different names for it.

47
00:04:32,410 --> 00:04:40,300
But it's essentially a device that you put in the car and it transmits large information about the driver all the time.

48
00:04:40,300 --> 00:04:46,090
So if you think of the example I started with with the loan, it's pretty clear that this is a way of reducing the asymmetry.

49
00:04:46,090 --> 00:04:53,590
So you're saying, I really don't know, I'm the insurance company. I think young people are, you know, statistically bad drivers.

50
00:04:53,590 --> 00:04:55,540
So, you know, I have to give you a really,

51
00:04:55,540 --> 00:05:01,720
really bad rate to compensate for the fact that I don't know anything about you and you are young, but you can say, Look,

52
00:05:01,720 --> 00:05:06,550
let me convince you that I'm a good driver and this is a credible way for me to

53
00:05:06,550 --> 00:05:10,720
use technology and transmit the information life all the time in a way to do that.

54
00:05:10,720 --> 00:05:17,320
And just if you know, if you're familiar with fintech, you know, there's lots of apps like that that that help people build credit history and so on.

55
00:05:17,320 --> 00:05:21,580
In much the same way as what I just described to, you know,

56
00:05:21,580 --> 00:05:26,590
so this is a lot of the new stuff is known as fintech or financial technology, and it's relatively new.

57
00:05:26,590 --> 00:05:31,240
The reason it's new is because until sort of 2011 12,

58
00:05:31,240 --> 00:05:39,220
the regulator stopped any kind of use of technology in finance for because the regulators view was we are here to protect

59
00:05:39,220 --> 00:05:45,790
the public and anything that involves people lending money to each other or doing all kind of risky stuff is bad.

60
00:05:45,790 --> 00:05:52,990
And that's what we would say. No. And that has changed after the financial crisis and after the government have instructed the FCA,

61
00:05:52,990 --> 00:06:01,330
the financial used to be the FSA, the regulators to work with the start-ups and help them come up with all these new things.

62
00:06:01,330 --> 00:06:08,590
And so we have this explosion of all this new digital banks and all these kind of applications that you may have

63
00:06:08,590 --> 00:06:15,450
heard of that are exactly using that using technology in finance and coming up with all kinds of new things.

64
00:06:15,450 --> 00:06:22,780
Let me give you a couple of example and some of the much loved company Wonga, if you heard of them.

65
00:06:22,780 --> 00:06:27,780
So one guy, they're no longer in business. But the idea was to give loans what they now call payday loans.

66
00:06:27,780 --> 00:06:37,000
So to give very short a small amount of money for short amount of time for people charging quite high rates for that.

67
00:06:37,000 --> 00:06:43,120
And the idea is these are people who wouldn't be don't have credit cards, don't have any other ways of loaning money.

68
00:06:43,120 --> 00:06:47,830
The banks wouldn't touch them. And so it is offering them something that they would use.

69
00:06:47,830 --> 00:06:52,390
But the key to that was to use this technology because they need the money typically now.

70
00:06:52,390 --> 00:06:58,690
And the decision has to be made very, very quickly. Normally on the smartphone, using the information they have.

71
00:06:58,690 --> 00:07:02,320
Another example, a bit more big scale, but similar is Funding Circle,

72
00:07:02,320 --> 00:07:09,070
which is a company that started by two of our undergraduates from the business school from Keeble College and is now a publicly traded company.

73
00:07:09,070 --> 00:07:17,320
So you may have seen their adverts on television. They're quite big, but Funding Circle is a peer to peer lending, so we take money from us,

74
00:07:17,320 --> 00:07:25,390
the public and we lend it to small and medium enterprises, businesses, small shops, small businesses that are starting.

75
00:07:25,390 --> 00:07:33,730
And again, you have this idea that there is a gap in the market because the banks after the financial crisis have stopped lending to these people.

76
00:07:33,730 --> 00:07:38,350
They need money. Some of them will, you know, not bad companies. Yes.

77
00:07:38,350 --> 00:07:43,130
And the public, on the other hand, is getting kind of close to zero interest rates on their account.

78
00:07:43,130 --> 00:07:49,030
So why wouldn't the public entities and thing kind of share some of the rewards from that?

79
00:07:49,030 --> 00:07:54,490
And so they use and they created a very, very successful base in business based on that.

80
00:07:54,490 --> 00:08:00,160
There's many more like them, but hopefully you've heard of either of them now.

81
00:08:00,160 --> 00:08:04,300
Both of them, as I said, are offering the service in a place where there is a gap.

82
00:08:04,300 --> 00:08:09,760
So they're offering something that didn't exist before. That's a good thing. Both sides are happy.

83
00:08:09,760 --> 00:08:14,890
Roughly speaking with that being, both sides rely heavily on algorithms to make decisions.

84
00:08:14,890 --> 00:08:19,600
So in Wonga, it's you can't have a person. It's not worth your while having the person.

85
00:08:19,600 --> 00:08:23,020
So making a decision of a loan of £10.

86
00:08:23,020 --> 00:08:29,500
But with funding secured, even though it's much bigger sums, it's still highly dependent on the algorithms and with funding circle.

87
00:08:29,500 --> 00:08:33,400
And again, as an example, there is algorithms and two sides of the market.

88
00:08:33,400 --> 00:08:39,400
So when the company, when a small, medium company and small is asking for a loan,

89
00:08:39,400 --> 00:08:44,270
it is the algorithm that essentially make the decision whether to give it the loan or not.

90
00:08:44,270 --> 00:08:47,740
I think there's nothing particularly sophisticated about this algorithm,

91
00:08:47,740 --> 00:08:51,130
which I haven't seen it, but it's using the same information as the bank would use.

92
00:08:51,130 --> 00:08:59,620
And so to make the decision, but also on the other side, on the public side, we can't invest the money directly in the businesses that we like.

93
00:08:59,620 --> 00:09:04,960
What happened is they take our money and they split it between a minimum of 100 companies.

94
00:09:04,960 --> 00:09:09,250
OK, so if you put a thousand pounds, it goes to a minimum of 100 different companies.

95
00:09:09,250 --> 00:09:18,310
And the idea is that you get diversification. And if one or two or even five systems fail, you will still get decent returns for for your buck.

96
00:09:18,310 --> 00:09:27,700
So it all makes sense, but you can immediately see that this is different kind of finance where all the decisions are made by the algorithms.

97
00:09:27,700 --> 00:09:36,760
So this algorithms in finances is very similar to what you got called gets in and and you know,

98
00:09:36,760 --> 00:09:43,240
it's got big advantages, obviously, because it allows us to do things we couldn't do before.

99
00:09:43,240 --> 00:09:50,290
And as I said, both sides are happy with it, but it also comes up like you've just described, and that's good with lots of challenges.

100
00:09:50,290 --> 00:09:56,470
And the challenges in this case is how are we making sure that the algorithms do what they're supposed to do?

101
00:09:56,470 --> 00:10:02,710
And it's not because the algorithms are bad, it's because we don't really have control over that black box.

102
00:10:02,710 --> 00:10:09,140
And so that create all kinds of issues for the regulators that they wouldn't be familiar with before and how to deal with that.

103
00:10:09,140 --> 00:10:12,820
So for example, and I'm not suggesting any of these companies are doing that,

104
00:10:12,820 --> 00:10:16,990
but you can see that there is pressure on them to match the demand and supply.

105
00:10:16,990 --> 00:10:24,160
So that is a two sided market. Yes. So you have to give loans and you have to have money invested in these loans from the public.

106
00:10:24,160 --> 00:10:33,470
And what happens if there is more people signing up to lend money than businesses asking for money?

107
00:10:33,470 --> 00:10:40,670
So, you know, does that create pressure on them to maybe slightly change the button of the algorithm and say yes to more loans?

108
00:10:40,670 --> 00:10:45,620
Yeah. This is the kind of things you do you really need to think about.

109
00:10:45,620 --> 00:10:49,430
And in the case of one guy, I can say, because that's public information.

110
00:10:49,430 --> 00:10:52,640
That's exactly what happened. And the regulator for that got very, very,

111
00:10:52,640 --> 00:11:01,610
very upset with them because they felt that the buttons were changed and they weren't, you know, fully informed.

112
00:11:01,610 --> 00:11:12,030
There's no suggestion of that with funding secured with any of the other companies, but that's it gives you a taste of what it's like.

113
00:11:12,030 --> 00:11:16,650
Now, there is one part of finance where algorithm trade three,

114
00:11:16,650 --> 00:11:21,750
there's one find in finance where A.I. and algorithms are not new and this is algorithmic trading.

115
00:11:21,750 --> 00:11:28,290
And this is algorithmic trading and this is what we do in the in the army and also some of the colleagues,

116
00:11:28,290 --> 00:11:36,330
my colleagues in the business school looked into that now algorithmic trading, the reason that the so it started in the 80s.

117
00:11:36,330 --> 00:11:39,990
So for example, my the is man.

118
00:11:39,990 --> 00:11:47,280
And actually the full name is my age. Gender and age is initials of three statisticians that came from Cambridge.

119
00:11:47,280 --> 00:11:56,670
Apparently, there's a university there and they and they came up with one of the first black boxes that were that were used in to trade.

120
00:11:56,670 --> 00:12:01,890
And the reason the regulator didn't really worry about hedge fund is because hedge

121
00:12:01,890 --> 00:12:06,090
funds takes money from very rich individuals and from very sophisticated investors,

122
00:12:06,090 --> 00:12:13,320
so they don't have to worry about it. And so we have a case study in A.I. in algorithms going from the 80s.

123
00:12:13,320 --> 00:12:20,430
And it's a really interesting one because with algorithmic trading, there's been a lot of start-ups, a lot of funds,

124
00:12:20,430 --> 00:12:28,950
some of them super successful, like ages like Millennium, like some of the Renaissance, some of the big ones, but some that have failed.

125
00:12:28,950 --> 00:12:39,620
And so we can learn from these things and see how and when algorithms are used properly and when don't use that inappropriately.

126
00:12:39,620 --> 00:12:50,820
And let me just two more minutes to finish on this. So. So in two minutes, I want to say that algorithmic trading, as you said, is from the 80s,

127
00:12:50,820 --> 00:12:56,760
but it's still growing and developing because there's new sources of data.

128
00:12:56,760 --> 00:13:03,750
For example, nowadays satellite, you can buy that data with satellite images and you can look at Compaq's of companies and

129
00:13:03,750 --> 00:13:10,950
estimate the demand from that or when the CEO of publicly traded companies speak to analysts.

130
00:13:10,950 --> 00:13:18,840
There is now lots of boats that translate that into numbers and deduce from that or kind of sentiment analysis and so on,

131
00:13:18,840 --> 00:13:23,520
and if there is optimism or pessimism and then trade based on that.

132
00:13:23,520 --> 00:13:31,320
So all these things are in exists and actually a couple of quick one, this one colleague of us in the business school is working specifically on that.

133
00:13:31,320 --> 00:13:35,190
So it is so, so it's very dynamic. There's a lot, a lot going on.

134
00:13:35,190 --> 00:13:39,780
And as I said, there is a lot of examples to learn from.

135
00:13:39,780 --> 00:13:49,290
And you know. Let me just give you maybe a couple of sentences how I think about it then is this

136
00:13:49,290 --> 00:13:53,140
it sounds a bit like the Matrix when there's good programmes and bad programmes.

137
00:13:53,140 --> 00:13:59,050
But I think there are good algorithms and bad algorithms and bad algorithms is like the guy who's now.

138
00:13:59,050 --> 00:14:02,110
I think he just the Americans decided not to put him in prison.

139
00:14:02,110 --> 00:14:09,670
This guy from Onslow who did this flash crash, the trader that uses his algorithm trading to trade really, really quickly.

140
00:14:09,670 --> 00:14:17,230
And he basically moved the whole the American stock market and he is now.

141
00:14:17,230 --> 00:14:22,570
It was in trial. I think they decided not to put him in prison for various reasons, but.

142
00:14:22,570 --> 00:14:28,150
But that's clearly an algorithm just used badly because it increases the risks.

143
00:14:28,150 --> 00:14:33,590
But good programmes, good algorithms is when you use the information to reduce risks.

144
00:14:33,590 --> 00:14:39,850
If, for example, you kind of trade lots and lots of different markets, which would be very difficult for a person to do.

145
00:14:39,850 --> 00:14:47,980
But in a way reduces the risk because you are placing bets in many, many more different baskets than you would otherwise.

146
00:14:47,980 --> 00:14:50,830
And as I said, this is a fascinating area,

147
00:14:50,830 --> 00:14:59,830
and I'm looking forward to hearing from you at the end and also collaborating with other people around university because I think it looks different,

148
00:14:59,830 --> 00:15:06,602
but it's very, very similar questions that we are facing. Thank you very much.