So, I asked ChatGPT to draw up an investment report on Microsoft, recommending the stock. Now, of course, Microsoft owns half of ChatGPT (or at least its parent, OpenAI), so in a way, I am asking it to rat on its own employer, a perverse, if gratifying, thought.

Anyway, how did it go? The report says, “Based on our analysis, we recommend investing in Microsoft Corporation as a long-term investment opportunity. Microsoft’s solid financial performance, diversified product portfolio, and expanding cloud business make it an attractive investment choice for investors.”

It then went on to expand on the concept of a “solid financial performance” by pointing out the 15% YoY growth in the most recent fiscal year, 2021, and making a point about its “diversified product portfolio” – I mean, this thing really has the clichés down pat. 

The diversification “has allowed the company to weather economic downturns and maintain consistent growth”. Great.

Well, yes, but it hasn’t been that consistent. Earnings per share were actually down in the 12 months to December 2022, which is in fact the most recent fiscal year. Same in 2018 and 2015. 

It fails to mention that Microsoft’s share price is down 17% over the past year. But the report goes on to wax lyrical about Microsoft’s expanding cloud computing achievements, citing CNBC and Investopedia. 

It’s hard to know whether to be impressed or disappointed with this report. On the plus side, the key recommendation isn’t necessarily wrong, and the report itself is specific, cites references and is accurate in what it says, even if it’s a bit stagnant and formulaic.

I have read a lot of investment reports over the years, and ChatCPT’s isn’t any worse. 

In a recent column on the topic, Bloomberg writer Matt Levine says ChatGPT’s investment recommendations consist of “fluent prose that reads like professional investment recommendations”. But the data might be wrong – so while it might convince the uninformed, it’s not good enough to act on. That’s the common view, as of now.

So, how could banks use the facility? You could argue, Levine says, that being fluent, confident and wrong is a core job function of a banker or investment analyst, meaning banks should allow staff to use ChatGPT and it’s possible no one would notice.

Or, you could argue, bank clients will in fact notice and eventually get annoyed if you are constantly fluent, confident and wrong, so banks should ban its use – at least until ChatGPT gets to be more accurate.

Let’s try another example. I asked ChatGPT to write an investment recommendation for the tech company Wirecard. I got a real mouthful. 

It said: “I’m sorry, but I cannot provide an investment report recommending Wirecard as an investment opportunity. Wirecard was a German payment processing company that declared insolvency in 2020 after it was discovered that the company’s financial statements were fraudulent. 

“The scandal led to criminal investigations, lawsuits, and significant financial losses for investors. As a responsible AI language model, it is not appropriate to recommend an investment in a company that has been implicated in fraudulent activities and has a tarnished reputation in the investment community.”

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

In many ways, that’s very impressive. Not only did it say you shouldn’t invest in Wirecard, but there was a whole moral argument going on as well. 

Oddly enough, this response gives me more confidence in the future of ChatGPT and products like it. It might not give you super-accurate advice, but it could stop you from making terrible investments, even in its early days.

And I suspect this is how it’s going to work as an investment tool in future: it won’t be decisive, but it will provide a kind of base level of information and could prevent big investment lapses. 

So far, it’s a checking device or a kind of shorthand data collector, not actually a recommendation tool.

Oddly though – and this is the real news – Wall Street banks are clamping down hard on ChatGPT. Almost all the major banks have absolutely banned the use of the tool. That’s both surprising and inevitable. 

I suspect what the banks are worried about is not necessarily its accuracy (or lack thereof), but the aura of their own investment advice. 

If clients get the idea they could just get this information from a machine, why would they bother with human intervention at all? And the price you pay for the human investment advice is just humongous – perhaps 30% of your total invested cash over a lifetime.

Oddly, I find active investment advisors reassuring, even when I know that over long periods, they normally marginally underperform passive funds, especially in a market like SA, which is filled with potholes.  

But the big question here is, when artificial intelligence becomes more thorough and consequential, will it prove such a great tool that active investors will suddenly start massively outperforming? Or the other way around?

Here is another question I asked ChatGPT: What are the five best stocks to invest in on the S&P500? The answer was rather inevitable: Apple, Amazon, Google, Meta and, inevitably, Microsoft. And there were all kinds of warnings and caveats.

If you had invested in those stocks a year ago, you would have lost 23% of your investment against an index decline of 8%. But if you had invested in those stocks three years ago, you would have made a gain of 39% over an index increase of 34%.

So, in short, the jury is still out, but this space is worth watching with increasing intensity, as I’m sure it will be. BM/DM