This column could have been written by artificial intelligence (AI). Indeed, I asked ChatGPT to write it for me, specifying that it consider what SA investors should do about AI, though the results were a cringeworthy amalgamation of clichés and even-handedness.
The first line, dear reader: “In the tumultuous seas of technological evolution, one term has emerged as both a beacon of promise and a harbinger of uncertainty: Artificial Intelligence”. It didn’t get better. I fear the editor of this august title would have little patience with such meaningless verbiage.
AI has had a rapid rise and apparent plateau as a theme for investment markets. Stocks such as chipmaker Nvidia and ChatGPT investor Microsoft have rallied but trended sideways more recently. AI has been hyped to revolutionise everything from banking to stock trading strategies, though perhaps these revolutions will be perpetually around the corner.
After the shock of just how human-like ChatGPT and similar tools looked at first, we seem to have calmed down, particularly as the limitations of large language models (LLMs) have become more obvious. Put very crudely, LLMs work by training algorithms with masses of data to predict the most likely next word in a sentence based on the previous words. They no more understand them than a dishwasher understands dishes.
There are problems, perhaps most famously “hallucinations” where LLMs make up things that sound plausible but aren’t true, because it is guided by probabilities rather than concepts of truth or falsehood (a good way to see this is to ask ChatGPT or Anthropic’s Claude for a reference to read on something. You often get plausible sounding titles and authors that probably don’t exist).
These underlying weaknesses are being addressed essentially by layering various human-designed “fixes” on top of the basic model and they gradually will get better, but the rate of change will be slower than the initial explosion.
The next front of AI may be “artificial general intelligence” (AGI) which is more akin to real thinking that could potentially assess right and wrong and much else, though experts are sharply divided on when this might happen. The truth is we don’t have a full model for how the human brain works, let alone an understanding of what it will take to replicate it digitally.
The human-like output of the LLMs has excited the imaginations of many. Some credible experts are saying AGI is just a few years away (for example, Elon Musk says a year away, Meta’s chief AI scientist says decades away, and Open AI’s CEO refuses to put a precise timeline on it).
We may blunder into AGI almost by accident because it turns out that general intelligence emerges from a deep level of machine learning combined with several additional interventions. But in truth, we just don’t know. Given the history of false predictions (for example in 1970 computer scientist Marvin Minsky predicted it would arrive in three to eight years) I think it wise to be sceptical.
Financial markets are highly attuned to the possibilities. The Financial Times caused waves last week with a front page proclaiming both Meta and OpenAI were preparing to launch new models capable of “reasoning”. The hype fizzled quickly when the market realised there was no radical AGI breakthrough, but rather upgrades with some additional new fixes for the weaknesses of their existing LLMs.
What should investors on the southern tip of Africa make of it? To state the obvious, there is no direct exposure to AI through our capital markets. The scale of investment required to be at the leading edge of AI is too enormous for anywhere but the world’s biggest pools of capital.
OpenAI was last valued at R1.5-trillion, about five times the value of Standard Bank, though it’s not available on public markets. Shares with a big exposure to AI such as Meta (market capitalisation R25-trillion) or Microsoft (R58-trillion) are bigger than all the JSE’s listed companies, including the dual listeds, combined.
The price to play at the leading edge of AI research is far too high for anything outside the US. Instead, local stocks such as financial services might be affected by AI indirectly, though potentially in conflicting ways as they benefit from new AI-enabled tools, but risk displacement by AI. The net effect will take time to determine. You can be sure, though, that our resources stocks will remain a good source of uncorrelated performance, perhaps benefiting if the productivity impact of AI breakthroughs does lead to a global economic surge.
So the investor strategy is the same boring one as always: a diverse portfolio is likely to be your best bet. As valuations shoot up on the potential of AI for tech stocks, they are becoming a bigger part of the investor universe, so anyone investing in a broad global portfolio will be gaining exposure without trying. I’m not sure investors should be doing anything more than that.
• Theobald is chair of research-led consultancy Krutham.
Would you like to comment on this article?
Sign up (it's quick and free) or sign in now.
Please read our Comment Policy before commenting.