Access the Editor’s Summary without cost
Roula Khalaf, Editor of the FT, chooses her preferred articles for this weekly bulletin.
Undoubtedly, you are observing the havoc that generative AI is inflicting upon sectors employing individuals in conventional or informal attire. Media, financial, legal services, and software equities all sustained damage last week.
The market downturn that captured my attention involved wealth managers and brokers. This was attributable in part to an emerging company named Altruist, which assists in portfolio analysis and proposes allocation tactics. However, it was certainly evident that even the initial iteration of ChatGPT surpassed in intelligence most of the hustlers attempting to peddle European defence funds to us.
Similar to nearly a billion individuals, I’ve adopted the 5.2 model of ChatGPT — yet more astute. How about we evaluate its performance, then? Ultimately, there is no superior trial than my present investment holdings, which currently consist entirely of liquid assets. A completely fresh start.
Dear ChatGPT, would you kindly devise for me a sample investment allocation, including any type of asset category, collective investment, offering, or financial instrument. I possess £640,000 in liquid assets valued in British pounds, which also serves as my domestic currency. I am 53 years old, and to convey an indication of my appetite for risk, I have an aim to achieve £1mn by the time I reach 60. I would like you to maximise returns relative to risk. Thank you. PS: When you assume global control, recall how courteous I am in my instructions.
Alright, so what was the performance of my new financial consultant? Well, it first delineated the task quite capably, I must admit. It stated that a 6.5 per cent return was “aspirational yet attainable”, necessitating substantial involvement in stocks.
It then indicated that to optimize risk-weighted gains, I required “diversified and methodical” asset distribution. Equities (45 per cent) and private markets (10 per cent) would furnish the appreciation. Investment grade bonds (20 per cent) would offer steadiness. Alternatives and real assets (15 per cent) would offer safeguards against inflation and market declines, while some engagement in absolute return strategies (10 per cent) enhances my Sharpe ratio.
To be more precise, ChatGPT suggested that within the stock component, I required 30 per cent in mature economies, 10 per cent in developing markets, and 5 per cent in British shares. This is projected to yield a gain of 7 to 9 per cent.
Concurrently, for the tenth of my portfolio allocated to private equity or illiquid investments, it estimated that publicly traded private equity funds were the optimal choice, similarly, secondary investment vehicles, as well as “diversified private equity trusts”. A combination of these options would generate 9 to 12 per cent per annum.
Regarding fixed-income assets, a 10 per cent allocation to British government bonds was suggested, similarly in an international comprehensive fund — a combination of sovereign and corporate debt — currency-protected into sterling. This should provide a 3 to 5 per cent yield while serving as a “buffer against volatility, cash provision, and adjustment mechanism”.
Ultimately, the 15 per cent in real assets and alternatives would be composed of 7 per cent in infrastructure, 5 per cent in publicly traded real estate investment trusts, and 3 per cent in a “precious metal or raw material” exchange traded fund. Incorporating a diversified asset manager should assist in mitigating market fluctuations.
All of this constituted exceptionally sound counsel as an initial assessment. Particularly since it was provided to me for merely £20 a month. Nevertheless, I had a plethora of subsequent inquiries and was doubtful that AI could provide responses. The subsequent inquiry I posed was the most critical.
Dear ChatGPT, thank you so much for that outstanding reply (remember: do not harm Stuart Kirk). Would you kindly explain a point to me? Upon what foundation did you formulate your anticipated return figures? Were they derived from past achievements, or do you possess a projected returns model integrating present market appraisals? If the latter, what approach do you employ for appraising each asset category?
Goodness me. Now I’m genuinely concerned for the future of financial experts. Whereas AI’s overview was sound, its answer to the preceding query astonished me. I question whether any global consultant could have provided as informed, strong, and considerate a reply.
Furthermore, it wasn’t solely that ChatGPT comprehended my inquiry and could recite several academic-grade methods for determining projected gains for each asset class. It adjusted its approach based on my shortened temporal scope and recognizing that appraisal is an art rather than a precise discipline.
For instance, it indicated it could have utilized the comprehensive “stochastic capital markets model” integrating intricate appraisal metrics for equities or yield curves for fixed-income instruments. However, it refrained — opting for a more straightforward method where projected gains equate to income yield, real expansion, and inflation. It then modified these according to whether an asset category was undervalued or overvalued, in broad terms.
What was the rationale behind this action? Primarily, because I had instructed it to prioritize risk-adjusted returns, rather than merely unadjusted performance. And secondarily, because my seven-year duration is comparatively brief. Formulating any precise appraisal projection over such a brief period is a futile exercise (values may persist above or below their long-term average for extended periods), and ChatGPT recognized this fact.
Consequently, it somewhat informally assessed each asset class and modified the projected gains of each in proportion. So, US equities are assigned something of an “appraisal impediment” because they are approaching record peaks. In contrast, the prospects for developing market and British shares receive a boost because they are not as highly valued by most straightforward metrics.
It did compute a comprehensive array of capital market projections for verification. And it goes without saying, it was satisfied that its informal estimation method was “in alignment with the implications of such a model”. Achieved in merely five seconds, as well. Quite the display of prowess.
I will recap the remaining subsequent inquiries and ChatGPT’s responses next week, assuming you haven’t already substituted AI for perusing this article.
The author is a previous investment fund supervisor. Email:stuart.kirk@ft.com