OK with AI?
The Lens Blog was on the road last week speaking to investment communities in London and Dublin about how they are using artificial intelligence (AI) to assist investment strategies.
We heard that some asset managers are becoming better informed about the potential benefits that AI may offer to their business, and more adventurous in putting these techniques to use.
For example, one quant investor spoke about the use of textual recognition, which enables fund managers to – in effect – read a mountain of analyst reports, far more and far quicker than a person possibly could, and to use the findings to gauge sentiment towards companies and markets.
This illustrates the main function of AI, which is giving investment firms advanced abilities to work efficiently with large datasets (‘big data’) and to identify patterns and insights on a scale that is rarely possible for a human research team. As one fund manager told us, AI means their funds are not constrained by the slowest or weakest human analyst.
Significantly, as an AI model is ‘trained’, the quality of the system’s output improves over time.
Yet the need for such training suggests scope for AI systems to make mistakes – just like a new human trainee would. It is understandable, then, if some firms would rather not take the risk of early implementation.
An AI consultant who the Lens shared a panel with last week argued it could be a mistake for firms to sit on the side-lines too long. Some of the successful pioneers of AI are already raising the drawbridge between themselves and those who choose to wait-and-see.
The avalanche of regulation aimed at fund firms over the past decade could be putting some firms off from early adoption. So much of that regulation is to do with data, don’t forget, and so full transparency of the underlying data, processing methods and decision-making are essential when using new analytical technologies.
Fortunately, AI machines are themselves demanding in terms of data quality and data expertise right from the start. The old term ‘rubbish in, rubbish out’, should become a thing of the past when referring to statistical-led investing which AI facilitates.