Defend Truth


ChatGPT presents an opportunity for investment managers to embrace, not scorn


Khadeeja Bassier is Chief Operating Officer of Ninety One, an authorised financial services provider.

Earlier this month, I attended the Emerging Technology Conference in London’s Guildhall. The venue was both ironic and fitting. The current building was constructed in the 15th century. Hardly an avatar, then, for bold futurism. But this grand hall stands in the city’s financial district, surrounded by firms which must relentlessly find ways to harness digital advances.

In that vaulted room, investment management participants had gathered to wrestle with the question of charting a way forward in what many feel to be unprecedented times. While in our industry change has always been the only constant, the pace of technological development has posed an existential question to us all: how do we harness technology to amplify rather than displace our value proposition? 

There is a degree of hubris required for the survival of humanity. Ideally, this should be coupled with self-awareness. Over the past five years, the sense of hubris when it comes to the power of technology has been staggering. 

We have all heard the refrains, “AI will have limited impact on me because my job is creative/uses judgement/relies on empathy”. And yet, this has been refuted time and time again. 

Sometimes, what we ascribe to judgement is simply a decision tree which has been pruned over time. Intuition and gut are more often than not learned experiences from millions of micro interactions. Essentially, neural networks enabling us to make sense of vast volumes of data, and this manifesting in heuristics for us to navigate the world. 

Moravec’s paradox states that “it is comparatively easy to make computers exhibit adult-level performance on intelligence tests or playing checkers, and difficult or impossible to give them the skills of a one-year-old when it comes to perception and mobility”. This paradox was formulated in 1988 and, ironically, contends that, in fact, it is the more abstract skill set that is easier to computationally replicate. 

Visit Daily Maverick’s home page for more news, analysis and investigations

Since the 1980s, computational power has accelerated at a pace that has narrowed that gap. 

ChatGPT-4 has just demonstrated the way natural language models will soon be able to mimic sight, i.e. to “see” images and interpret them. Hitherto, this was virgin AI territory for popular usage and now is a potential game changer in the path to robotic sentience.

I use the notion of sentience with tongue quite firmly in cheek. While there have been remarkable advancements, we (thankfully) have a long way to go to a robotic dystopia. However, my firm belief is that as an industry we need to move past the smugness of human ingenuity when relating to an AI model that has been “live” for four months. 

Four months of iterative learning is not a definitive yardstick for success. But embracing this development in our daily lives presents a way to power up our toolkits in dealing with the unpredictability of an increasingly multi-polar world. 

AI presents a compelling opportunity for asset managers to leverage a new baseline. Natural language models will enable us to distil the corpus of knowledge at a rate and in a way that will enhance how we think about the world. 

Alpha, or excess return, is generated through value-add above consensus. True alpha is a consequence of being ahead of conventional wisdom. The typical fundamental analyst will scour reams of reports, the laborious pursuit of which hopefully harvests fresh ideas and enables capital. What the latest language models offer is a way to condense information at scale that would open new areas in a way that was previously inaccessible in a time-constrained (even a 16-hour workday) context. 

AI doesn’t replace the diversity of thought and value of a good analyst. It enriches it. 

If one remains a sceptic on the broader direction of travel and direct work applications, the very notion that ChatGPT can devise weekly meal plans based on dietary requirements and preferences, and then populate a shopping list based on portions required and order that list by shopping aisle, should convince you this technology will save time. How this time saved translates to more meaningful investment decisions is yet to be seen.

Coming back to the conference in London and the general mood in the industry, I was struck by this juncture being our Road to Damascus moment. Immersed in that physical space where Late Middle Ages architecture formed the backdrop to a conversation around pushing the frontiers of innovation, the analogy for where we find ourselves as an industry today could hardly have been more apt. BM/DM


Comments - Please in order to comment.

  • Peter Atkins says:

    I love the AI term for making up or inventing a missing fact – “hallucination”.
    I’m looking forward to AI tantrums as AI models mature into the “terrible twos” and then rebellious teenager phases. However, I’d prefer to be driving my own car during an AI tantrum.

Please peer review 3 community comments before your comment can be posted