ARK Invest is an investment adviser that was founded in 2014 and focuses solely on researching and investing in disruptive innovation. One of its investment focuses is on the next generation internet, which includes artificial intelligence and deep learning.
Other areas ARK focuses on include: genomic sequencing; robotics and automation, including 3D printing; energy storage, for example battery technologies in the shift to electrical transportation systems; and blockchain technology, involving crypto assets.
Catherine Wood, founder and CEO, describes ARK as “a sharing economy company in the asset management space”, which refers to the way research is managed.
ARK has set up a kind of open source research ecosystem in the financial markets, which relies on social media: networking through LinkedIn, branding with Facebook, sharing information and engaging communities on Twitter, writing thought leadership on Medium, and crowdsourcing views from platforms like Seeking Alpha.
There’s also an intranet in which ARK and its “theme developers” — professors, venture capitalists and private equity investors — exchange knowledge.
MarketBrains talked to Catherine Wood and ARK’s Next Generation Internet analyst, James Wang, about what it means that AI is coming into the ETF space.
MarketBrains: AI in ETFs can mean different things, how do you see this space developing?
Catherine Wood: From our perspective, every company is going to have to become an AI company of sorts because it’s going to impact every line item of the income statement.
We’re looking at the enablers of this kind of activity, and also the beneficiaries, but I think eventually we’re not going to be able to separate any business from artificial intelligence.
MB: And what about AI being used to make decisions on, for example, populating an index, or pick ETFs?
CW: We’re focused on it with interest but we’re not anywhere near there in terms of signals versus noise. For very short-term trading, sure you can impose it on that. But we are very long-term in our thinking and AI is not ready for what we are doing.
James Wang: We have seen some funds try to take deep learning, for example, to do trading.
These are very short-term trading techniques and when you look at what the underlying is, it could be anything: gold or stock or commodities — all could be done in theory that way.
We think the problem with that approach is you’re not in a single entity environment, you are in a multi-agent environment so you are competing with the quality of other people’s algorithms.
If everyone uses this technique it tends to cancel out and that’s our concern with that style.
MB: You are basically active fundamental managers, right?
JW: Our firm is focusing on the impact of artificial intelligence on society, and the economy as a whole to try and identify the best names in terms of that line of questioning and giving investors exposure before those names appreciate to reflect their value.
If you compare what ARK does to, for example, robo, or any of these other AI or automation-based ETFs, the real value add is in active management.
We have analysts who basically study this field fundamentally day in day out, and we manage it more like a mutual fund. You get a real exposure to AI, not just this smattering average exposure to AI.
There is no sense of conviction or what is actually important in their portfolio is my impression when I look at them.
“…we are seeing companies across the board trying to figure out how they can capitalize on artificial intelligence…”
CW: Focusing on Nvidia specifically, which we believe is the AI chip company, when we started focusing on it, NVIDIA was being tapped out of portfolios because it was deemed as nothing more than a proxy for the PC space and PCs were dropping at a double-digit rate.
During brainstorming sessions, as a portfolio manager I could see that other portfolio managers did not understand the impact Nvidia was going to have on artificial intelligence or autonomous taxi networks.
No one knew that the brains of those networks were going to be GPUs and that Nvidia is the leading manufacturer. So we were picking up on Nvidia’s impact on artificial intelligence and putting it in our portfolios when most didn’t even understand that Nvidia had a play there.
Estimates of revenues from the AI opportunity go anywhere from $10 billion to $60 billion, and because most analysts are so short-term focused, they can’t even incorporate that kind of number into their models. They are just going to wait to see it, and then extrapolate.
We are all about anticipating this sort of thing.
MB: How are investors responding to AI in ETFs? Has there been a change over time? Our sense is that investors are still quite wary.
CW: Elon Musk says beware of the risks of artificial intelligence, you could have nefarious actors involved, just like every new technology does. Half of the solution is understanding the problem.
Given what we do, we are seeing companies across the board trying to figure out how they can capitalize on artificial intelligence because it does change the competitive dynamic so much.
Those companies that don’t embrace it are going to lose competitively, so I think because of the way we are coming at it and James’ expertise and his background, we’re convinced that it is going to become very powerful and potent.
Salesforce.com for example is saying it is making a billion predictions per day, investors are seeing that something is stirring here, and companies who do not embrace it and investors who do not embrace it are going to miss a lot.
It also will be very interesting to track those funds that are being managed with artificial intelligence and those that actually invest in companies exposed to artificial intelligence or embracing it faster than others.
Today, the human mind is so much more pliable than machine learning, there are so many nuances in the portfolio decisions that we make that sometimes they seem contradictory.
Maybe machine learning will get to be that good but in my experience, I’ve seen it blow up a lot of portfolios because it assumed that the world would continue to work the way it has always worked, and it doesn’t.
MB: Still, some of the work in unsupervised learning counters that, but does require quite a bit of data and processing power for sure.
CW: I am intrigued to see how those respond and I know that all the games, like AlphaGo, that deep learning has conquered are complicated games. Investing is even more complicated I think.
ETFs by definition are backward-looking, they are based on equal weight or cap weighted stocks that are in a universe. In other words, how has the world worked? OK, those are the stocks we’ll play.
Well, the way the world is going to work could involve completely different stocks.
JW: How stock selection is done is a key question for a fund like this. When I read the materials for robos or their peers, my impression is they draw broad buckets and make some subjective calls. They make some scores and then a rules-based system.
They call it a machine learning-based approach, but that first step is very much a human-based approach.
For example, these funds don’t hold Amazon as a stock. Anyone who has actually listened to an earnings call will realize that Amazon is a machine learning and robotics company, it has the most sophisticated automated warehouses of anyone in the world. Alexa’s voice system is the most popular in the US and Apple is similarly exposed.
Alexa: I don’t know that.
JW: There you go, I just triggered my Alexa.
I think while people claim they are using AI for these things, if you actually look at a stock selection and how the process works day to day, it’s still very much contingent on the fund that is human-driven.
If a company CEO makes a certain comment about an AI initiative, that news flow can’t just automatically go into a factor that then reflects the weight. For us, we can act in real time.
This interview is edited and condensed