The world is changing fast and to keep up you need local knowledge with global context.
Artificial intelligence is fast replacing humans at their jobs. Thirty years from now will there still be a role for humans in Investment Management?
This discussion is sponsored by Absa, a member of Barclays. David Williams is with Vivienne Ming, who is a theoretical neuroscientist and founder of Socos. Vivienne, you’re a person who looks broadly at improvement of human achievement and a neuroscientist, which sounds interesting. How does all this apply to investment in particular?
It’s fascinating from the perspective because of both the incredible growth in automated systems and artificial intelligence investing and what’s coming next and the implications? In the financial world we’ve seen the growth of companies, like Sig Fig, and Wealthfront in using automated investment advice for customers but in fact, what’s much more fascinating and relates to work I’ve done recently, is the growth of truly powerful artificial intelligence systems. For example, the one that recently won the Go Competition against the world’s Go champion called Alpha Go, developed at Google. What is most fascinating about these systems is they can learn on their own. That they need only to watch the stock market, watch the investment strategies of individuals or even companies. If you give them some criteria like maximise my return over the next five years, it will learn how to do that all by itself, better than a person can.
That has some fairly big implications, both theoretical implications for what is when markets are full of AIs competing with one another but also very human implications. For what it means in terms of the investment strategies of individuals but also the professional prospects for people who’ve made this their careers.
If the machines are competing with each other doesn’t it get to the point where they all know what to do and therefore, there’s no superior or inferior performance?
It turns out their part of these interesting theoretical implications of if they are coming up with near optimal strategies, all of them, then what, in a sense, is the point of the market? One of the interesting discoveries in this space, it’s been known for a while, is there is no stable strategy. That what happens when AI’s like this compete is that it becomes this complex series of booms and busts, where they reach a seaming stable strategy with one another and then a new one comes in with something completely new, and the whole system falls apart and has to start over again.
Initially enough, although it sounds a lot like humans, where we engage in the series of boom and bust strategies ourselves, so maybe in some sense they’re not fundamentally different except for the fact that very quickly their strategies will become nearly impenetrable to us. They will look like random moves that somehow magically produce superior outcomes. In fact, I’ve often said that the future of financial advising is more like historian. To analyse historical data and try and figure out what these machines were doing in the past because we won’t be able to keep up in real time.
One of the best examples of this would be the high frequency trading markets, and how the use of artificial intelligence there is meant to get milliseconds of advantage, over another high frequency trader. People argue about the additional liquidity in the market. I’ll leave that debate aside and say clearly; the competition between these vehicles, as opposed to the additional liquidity, they may collectively add into the market is a questionable margin of value. I think there’s a real point here. If we can develop optimal strategy systems that do well enough against a human investor that there’s no point in having those anymore, then the question becomes is there real value in the market or should we leverage these systems simply of, you know markets are about resource allocation. Could we simply transition to a system where AI’s are doing resource allocation directly, rather than markets?
Needless to say, this is not an overnight change in what has been a very successful capital system but again, it has these real implications. The title of my talk is the ‘Investment Singularity’ and there are at least three possible singularities. This is in reference to a moment in time, beyond which it is impossible to make prediction, and one of those is, is there a moment in time where we will cross over between the rationality of running markets or not. This isn’t a distant future possibility.
The other two singularities are will there be a moment in time where human work in markets is no longer valuable and that’s an implication for the broader league markets. Then the last is the classic one. It’s the one that Bill Gates and Elon Musk talk about, which is will there be a moment in time when no human job is a marginal value and what do we do with ourselves at that point? None of those are so distant to question that it’s not worth thinking about now.
Even cooking food and growing food and all that?
There are amazing technologies. Robots that visually distinguish between crops and weeds, they use millimetre precision fertilisation jets to kill the weeds or fertilise the crops. Effectively giving you conventional scale, organic farming faster than humans can do it. The only question about human labour is does it become so low value that it’s worth maintaining at that end of the scale but what we’ll end up with very quickly is two kinds of labour, one which is worth less than an AI, and one which is worth more.
I think quite apart from theoretical issues in finance and artificial intelligence there are very human discussions about what it means to be on one side of that split versus the other. Let’s be frank, there will always be a marginal value in subsistence farming. I just don’t think it’s the value proposition most people want for their kids.
Yes. What about trusting the machines (the AI), to evaluate the impact of a Trump being elected president?
It’s an interesting story. Right now, the beginnings when we talk about artificial intelligence, probably most people are thinking of something we built but behaves much like us. The reality is we’re talking about self-driving cars, image recognition technologies at Google and Facebook and the ones, which we’re most familiar with aren’t even really intelligent. In fact, most of the financial management systems run by Wealthfront and Sig Fig, for example, are just you answer a set of questions and an algorithm takes those and says ‘here’s how you should invest’. It’s called just a look up table.
We don’t need to debate what is true artificial intelligent and what isn’t to say that the given political models are still on that look up table end of things. Currently, they’re failing. They’re suggesting that there’s some real possibility, I mean how do you assess a Trump, going forward when your models are based on historical data? What’s fascinating about AI, the possibility of an Alpha Go level investment strategy, is if we can link Alpha Go with another famous AI, IBM Watson that can go through and read every newspaper, read every article, it can run surveys, assess peoples’ temperament and sense about Trump. Maybe even in further implicit sentiment, the things they’re not willing to say to another person, and then integrate that into Alpha Go as, an essentially betting strategy. The result is again, a fully deployed, a system which is surely much more accurate and what I think we’ll find is Trump probably is much more appealing than people are willing to admit, to someone running a survey.
From an investment point of view?
Now, from an investment point of view I think that the implications are a lot less certain because, I will say honestly, what Donald Trump plans to do as President, is probably not well indicated but what he said he’s going to do.
In addition, the American system will prevent him from doing a lot of what he wants to do.
It is a fundamentally, by design, checks and balance system, sort of dysfunctional by design but what I think he’s shown in the past is he’s a very burly, phony type innovator, let’s say in the political sphere and I think his whole goal will be to figure out how to get around that. I think the only question is will he get bored? If he gets bored then very little will happen, if he doesn’t, if he really wants to go after creating a difference as he sees it, which I find of great concern, then you have that opportunity for something truly different happening. Where he might be able to leverage me, push politicians around in the States and create big change. What that would mean for the financial markets? I actually think it should be a long-term negative because despite all the frustrations everyone on the ground the world has with the US Government; it is still a point of stability. If that changes, I think it will inject a level of uncertainty into markets, which will be unfriendly for most people that are not investing outside of China, and quite frankly, we’re the single biggest market for China as well, so I think, if nothing else, it injects huge amounts of uncertainty in the future.
Thank you Vivienne Ming, she is a Theoretical Neuroscientist.