We are forever being told that artificial intelligence (AI) is coming to take our jobs. Is your financial advisor one of those who is going to be forced into early retirement? Or, to put it another way, would you trust your finances to be managed by a robot?
Our money is already being managed by machines in the form of algorithms — straightforward computer procedures which follow a programmed set of rules. They are part of everyday life, whether it’s Amazon trying to give you product recommendations or Facebook identifying which of your friends you find least tedious. Indeed, you may only be reading this article because an algorithm thought you might be interested in it. Algorithms are also used in the management of index-tracker funds — buying and selling stocks automatically so that the composition of the fund more or less mirrors its benchmark index.
Algorithms have a long and not always illustrious history in finance, however. Buying shares just after they have increased in value and selling them as they start to fall doesn’t always make much sense. The 2010 Flash Crash, where high-frequency trading algorithms were at least partly responsible for a 9 per cent drop and bounceback in the Dow Jones within just 36 minutes, is a notable case study in machines making perverse investment decisions.
Many providers including Invesco PowerShares, Vanguard and Deutsche Bank offer ‘smart’ trackers, which go beyond simply mirroring an index. They implement popular investment strategies — e.g., buying companies which have a strong track record in paying dividends — and apply them without human intervention. These are a bit more expensive than index trackers — typically 0.3 per cent of your investment in the fund, compared to around 0.1 per cent — but still cheaper than most actively managed funds. There is a wide variety in the strategies employed, and the smart tracker’s performance is entirely dependent on how good the underlying strategy is.
Taking more direct aim at the personal side of the wealth-management industry are so-called ‘robo-advisor’ funds which seek to replace the traditional role of the financial advisor in making product recommendations tailored to the individual. These include Nutmeg, Moneyfarm and Wealthify, among others. ‘-Technology has made the personal investment process much faster, with low-cost financial advice now accessible to all,’ says Scott Gallacher of Moneyfarm. ‘Where it can take up to a week to get investment advice from a traditional wealth manager, Moneyfarm’s own algorithms reduce this to seconds.’
Robo-advisors offer lower fees than traditional money managers, in the region of 0.45 to 0.7 per cent depending on the provider and degree of intervention. Fees charged by traditional wealth management can be more than 3.5 per cent, so there is a substantial possible saving. The downside is that, being relatively new to the market, their track record is quite short. The traditional money managers are unlikely to sit on their hands either so there’s a chance that fees will start to fall across the sector.
But none of this is AI in the strictest sense — it’s just machines implementing strategies devised by humans. The next step is to let the machines devise the strategy by identifying patterns from analysis of market data.
Machine-learning, as it is known, has been used by hedge funds for years, but remarkably it is only just beginning to be adopted in the management of the big public funds to which you and I have access. Blackrock announced plans last year to retire seven of its 53 human fund managers, replacing their traditional stock-picking nous with techniques more reliant on machine-learning and big data analysis. Their new China A-Share Opportunities Fund will rely on the wit of machine rather than man in a way previously open only to institutional investors. In the month since its launch, it is 3 per cent up on its benchmark index, but obviously a much longer period is required to assess whether this is a gamechanger.
If the machines do perform at least as well as traditional fund managers — not a difficult task — then there will be a bloody cull in the industry. But there are reasons to be cautious. For example, the complexity of the financial system could prove a challenge for the robots. The strength of machine-learning is that it can recognise deep underlying patterns which are invisible to humans. But the flip side of this is that these patterns cannot be easily justified or explained. And trained on past events, they will be just as vulnerable as humans when it comes to coping with previously unseen events.
Google’s AlphaGo — an application of AI in computer games — mastered the ‘Go’ through repeated simulation and analysis of millions of games which enabled it to differentiate between strong moves and weak moves in any given position. However, this approach works best when the situation under consideration can be precisely reproduced many thousands of times. In real world situations, where the actions of other players are unpredictable, individual companies are insufficiently similar, and the given market conditions at any one point are unique, this replication and subsequent precise analysis of a situation isn’t possible.
Another factor to consider is that it is relatively easy to outperform the market when moving small amounts of money. The more money that comes under the management of AI, the harder it will become for AI to outdo the rest of the market. It is the same as with human-investing: as more people copy successful methods, those methods produce ever more average results.
There is also a problem of trust. A machine-learning process encodes the patterns it finds very deeply within the machine’s ‘brain’, so that it cannot be unpicked and understood by humans. Financial regulators may be wary of software which cannot be robustly audited.
Moreover, will investors be happy to stand back and watch as decisions are made which seemingly don’t make sense? If what appeared to be a foolish investment does go badly wrong in the short-term, would you shrug it off and assume the computer can see something you can’t? Lacking human emotions, robots can avoid the mistake human fund-managers often make: losing faith in their strategy just at the wrong moment and changing to a less successful one. The machine can plug on and keep going — but would you, its customer, allow it to?
Unlike traditional fund managers, a robot will feel no compulsion to send out an apologetic missive outlining its much rosier expectations for the year ahead. Given that you are human, it may make you feel better just to know that there is a human fund manager to blame for losses, and that you could, if you wanted, reach over the desk, grip the bastard’s tie and ask firmly where your money went. You can’t do that with a robot — so perhaps there’s a future for fund managers, after all.