Come that brave new world, then not only will the streets be free of motor accidents - well, except for the few caused by cussed old fools who still insist on driving - but the financial markets will be free of investors; or, at least, those in the shape of homo sapiens.
Professional fund managers would go first. Why hire an expensive MBA to select stocks when a piece of software could do the job twice as well without demanding a fatter bonus when it got lucky? A few private investors - a bit like those old drivers - would linger on because investing is their hobby. Their numbers will diminish by the year, which will threaten the Investors Chronicle. No worries - the magazine will transform itself into an expensive bit of software whose tailor-made programmes are updated in a real-time download to subscribers who are content to let the package trade for them. That won't leave much room for Simon's small-cap picks, or Algy's stock screens, let alone Bearbull. No matter, their spirit will live on in algorithms written by a team of software engineers based in Lagos and supervised from Long Beach, California, by a maths PhD from Massachusetts Institute of Technology who specialised in quantum-field theory.
Okay, I caricature, but the question is: by how much? We all know that the so-called third industrial revolution is on its way. The first one did for agricultural jobs and the second for the blue-collar type. The third will make deep inroads into white-collar work; not humdrum clerical jobs - they have already been digitised - but the added-value stuff, which was supposed to be immune to technological advance - the law, medicine, accountancy.
Investment analysis and portfolio management will be victims, too. One can hardly imagine a skill set better suited to automation by smart software than an investment analyst's. It is happening already. The proliferation of quantitative funds tells us this. It ensures that a background in maths and computing is at least as important as economics or finance. However, the defining moment will be when sufficient quant funds are driven not just by software - that happens already - but by machine-learning algorithms, programmes that adapt themselves without being specifically programmed.
The process must be on its way. There is too much machine-learning software around for it to be otherwise. And some are free; in particular, the 'scikit-learn' library, developed by a programme sponsored by Google. In January, two academics from Johns Hopkins Carey School of Business in Baltimore produced a paper which showed that they used scikit to predict the price direction of 10 exchange traded funds (ETFs) that tracked broad market indices, such as the S&P 500 index of US shares, US Treasury bonds and the MSCI Emerging Markets index.
The academics used three types of machine-learning programmes - 'deep neural networks', which use multiple layers of programming through which data is passed; 'random forest' where lots of binary decision-making trees combine to produce suggested outputs for given inputs; and 'support vector machines', which classify new information into existing categories.
Using daily prices and trading volumes in the 10 ETFs for the five years January 2011 to 2016, the academics wanted to see if machine-learning programmes could predict changes in the direction of prices. Over a short period - up to 20 days - the programmes had no predictive power, supporting the notion that markets go on a random walk. Beyond that, predictive power materialised, peaking for the price movements over one to three months. As to the most useful input, that was the volume of stock traded.
Sure, as they acknowledge, the academics were scratching the surface. Big-time machine-learning programmes would require oodles of data and programmes more sophisticated than scikit can offer. Even this may be happening, but it's so secret that only rumour seeps out; the rumour, for example, that the Medallion fund, run by Renaissance Technologies, a New York investment manager, is powered by artificial intelligence.
Medallion is hugely successful. Bloomberg, the financial information provider, claimed that its returns averaged 72 per cent a year for the 20 years to 2014. Actually, few know the precise figure because the fund is wholly owned by Renaissance employees so does not publish returns.
Still, imagine that funds run by artificial intelligence - maybe like Medallion - are the way to go. Does that mean they will take over the investment world because they will obviously be so much better than plodding humans? Presumably not. The machines - just like humans now - would compete for 'alpha', excess returns available to the best. Returns to alpha, unlike the market's, are a zero-sum game - one investor has to snatch them from another.
This means that the machines will be fallible; some will be better than others. And in that smart-but-fallible world there must be gaps where humans can scrape their own bits of alpha. So maybe Investors Chronicle, the magazine, will survive. But there is little question that, if the investment present is tough for folk, then machines will make the investment future tougher.
