You know that the term AI (Artficial Intelligence) is everywhere now. And, frankly, it seems to be one of the most overused and misunderstood terms amongst investors. So, let's try to cut through some the marketing fluff so you can gain a clearer take on what AI's use is in the context of investing.
We will distill AI's use in investing to two key components and currently available examples (we do not recommend either herewith):
1) Security Selection
Recently, we ran across Kensho Technologies, a firm with MIT and Harvard-educated folks that uses large amounts of data (read: Big Data) and some AI processing to create a stock index of next generation themes, like Clean Power.
The AI argument will be that ferreting through large amounts of data will identify those stocks that would likely deliver on the premise of Clean Power better than other general techniques for doing so. Note that there is no suggestion that alpha, excess return and/or lower risk, is part of the objective, which is alright, as this AI approach is simply about picking a security to meet the mandate and risk/return seems to be irrelevant. Rebalancing is on a 6 month basis so clearly this is a high beta investment process that requires simply a decision (by an investor) to invest in Clean Power and leave the rest to the machine.
What is the long-term value proposition of this form of AI investing? Very limited, as it is easy to replicate the holdings and the actual holdings are all big name companies that you likely are invested in elsewhere in broad baskets.
You can get the above at costs of 20 to 40 bps so it is pretty cheap and reflective of the value proposition.
2) Security Trading
AIEQ is a common reference ETF if you have read earlier blog posts of our's as it combines IBM Watson- supplied analysis of unstructured information on a basket of SP500-llke stocks, which is then further translated into a daily rebalancing decision with the goal of demonstrating alpha.
So, all the AI revolves around making a trading decision on the individual stocks by first using unstructured information and then combined with some form of quantitative analysis (they don't disclose how any of this actually works) on top of the unstructured info. The value proposition is far more distinct to potentially gather alpha. At 77bps for AIEQ, its cost is reflective of the issuer's belief that there is real alpha value embedded in this trading process.
Note that the Kensho-based ETFs have gathered a relatively paltry amount relative to the $100M+ in the single AIEQ even in light of cost difference to investors. So, clearly investors place a greater value on potential alpha generation regardless of cost.
Summing up - next time you see the words AI in the context of investing, ask whether the AI is using for Security Selection or Security Trading. That way, you can better assess whether the potential benefit you may garner from the use of AI is worth whatever cost is being charged for it.
There's more to this but hopefully this clarifies AI a bit for you.