How AI will change investment and research

0 0

Unlock the Editor’s Digest for free

The writer is a former global head of research at Morgan Stanley and former group head of research, data and analytics at UBS

Sir Isaac Newton lost a fortune betting on the South Sea Company. Perhaps he did not take into account the “precautionary principle”. If you don’t understand something, even if others seem excited about it, it’s better to do something else or wait until you do get it.

This principle should perhaps be borne in mind by investors and investment banks as artificial intelligence is developed for market and financial research. It could create a lot of noise.

Much has been written about shrinking “sell-side” research by investment banks. Even more about the troubles of active portfolio management. Blame regulation or passive funds, but the struggle to add value increases when uncertainty rises — for example, as a result of geopolitics or products such as Ozempic or Tesla that radically change the competitive landscape. At such times, the market whipsaws, with swings driven by a diffusion of views. This could be made much worse or better by the democratisation of generative AI.

Any operator will be able to create plausible comments with credible data. This might be considered by some as an alternative to the diminishing grip on the “official view of the future” that traditional financial intermediaries have today. A subscription to OpenAI and an X account will be all it takes to pour fuel on the fire of uncertainty. The total cost could be as low as $30 a month for the soap box and the content generator. 

But overconfidence in a potent AI tool will replicate the error of Sir Isaac. Even powerful models are influenced by context like humans. When the St Louis Federal Reserve compared the performance of inflation estimates between a pool of professional forecasters and GPT3, the AI tool had less bias and lower errors but it was subject, like the professionals, to something akin to mood swings.

This is not just a potential problem for the inexperienced “day trader”. We only need to look at how increasingly complex versions of value-at-risk models gave the confidence to many financial institutions to overextend risk in their balance sheets. Sophisticated firms may fall under the spell of a unique model that was excellent at explaining the past but is not sensitive to new uncertainties. In a way, firms focused on investing through quantitative systems have expertise in dealing with this problem, as market regime changes render previous models useless. But the opaque characteristics of a GPT will complicate the forensic analysis. And while it is tempting to lean on the rich information that AI tools may add, eventually usage hits the law of diminishing returns.

However, generative AI will bring benefits. It is a wonderful tool to test many alternative hypotheses at a very low cost. There will still be flaws that come from a backward-looking perspective but if we look at a problem from many angles, the combined prediction error of the ensemble will be radically reduced. It will be possible to use an AI lens to analyse Disney’s strategy, for example, looking at past reactions to factors such as price changes in streaming, success of blockbusters or how weather affects its amusement parks. All of this can be done by large research outlets or multi-manager platforms. But it requires a major investment in infrastructure, diverse subject-matter experts and lots of communication within the firm. This is difficult to sustain.

The next ecosystem could be very different. On one hand, imaginative professionals will be able to prompt multiple inquiries at a fraction of the cost. But alternatively, large firms will be the ones able to afford highly specialised, built-for-purpose models to chase multiple sources of returns. This could be done by an alliance with a tech company. Or we may see firms uploading data to modify the generic AI models. These options appear likely as firms like Open AI may create a platform for other businesses. It is hard to say which is the winning hand. 

Like top chess players, investment professionals may need to add new skills. With the arrival of spreadsheets, we all thought that there was a lot of value in building a model that represented the future under certain assumptions. Instead, the value may lie in the proper interrogation of the AI tools. We can see that large firms may need to hire more engineers to fine-tune the process of interrogating the models. A discipline in its own right.

The Lumiere brothers commercialised the Cinematographe in 1895. By 1905 they believed they had exhausted all the creative ways of the medium. That was wrong then. We should use our creativity to expand what Generative AI has to offer.

 

Read the full article here

Leave A Reply

Your email address will not be published.

This website uses cookies to improve your experience. We'll assume you're ok with this, but you can opt-out if you wish. Accept Read More

Privacy & Cookies Policy