There’s a lot of interest within the EA community in investing in certain securities to benefit when AGI is developed. Which stocks, ETFs, etc. would you invest in and why?
[Question] Which stocks or ETFs should you invest in to take advantage of a possible AGI explosion, and why?
It seems altruistically very bad to invest in companies because you expect them to profit if they perform an action with a significant chance of ending the world. I am uncertain why this is on the EA forum.
I believe this section of this post by Zvi addresses your concern about this:
Q: Should I invest in AI companies?
Short Answer: Not if the company could plausibly be funding constrained.
Long Answer: Investing in AI companies, by default, gives them more funding and more ambition, and thus accelerates AI. That’s bad, and a good reason not to invest in them. Any AI company that is a good investment and maximizing profits is not something to be encouraged. If you were purely profit maximizing and were dismissive of the risks from AI, that would be different, but these questions assume a different perspective. The exception is that the Big Tech companies (Google, Amazon, Apple, Microsoft, although importantly not Facebook, seriously f*** Facebook) have essentially unlimited cash, and their funding situation changes little (if at all) based on their stock price.
The post was published on 1 March 2023, but it seems like the entire tech sector has been more cash constrained recently, what with the industry-wide layoffs.
I think this question is worth discussing, but I downvoted your comment for suggesting that my post does not belong on the EA Forum.
I imagine that at any point in time either big tech or AI safety orgs/funders are cash-constrained. Or maybe that at any point in time we’ll have an estimate which party will be more cash-constrained during the crunch time.
When the estimate shows that safety efforts will be more cash-constrained, then it stands to reason that we should mission-hedge by investing (in some smart fashion) in big tech stock. If the estimate shows that big tech will be more cash-constrained (e.g., because the AI safety bottlenecks are elsewhere entirely), then it stands to reason that we should perhaps even divest from big tech stock, even at a loss.
But if we’re in a situation where it doesn’t seem sensible to divest, then investing is probably also not so bad at the current margin.
I’m leaning towards thinking that investing is not so bad at the current margin, but I was surprised by the magnitude of the effect of divesting according to Paul Christiano’s analysis, so I could easily be wrong about that.
I believe this paper gives a clear picture of why it would be advantageous for a charity to invest in the very companies they are trying to stop.
In short, there is not really any evidence that investing in a publicly traded stock materially effects the underlying company in any way. However, investing in the stock of a company you are opposed to provides a hedge against that company’s success.
Buying a publicly traded company’s shares marginally increases their price and the price of the company’s corporate bonds in expectation. In expectation, this allows the company to raise new capital more easily by issuing new shares and bonds at better prices, lower interest rates or while diluting existing shareholders else less.
Buying a publicly traded company’s shares marginally increases their price and the price of the company’s corporate bonds in expectation.
I don’t think this is true and I as I understand it this statement is controversial among economists.
Here’s how I think about asset pricing. For an individual stock, there are typically large pools of informed traders (e.g., fund managers) who control sizeable portions of all assets. They each have a “target price” for a stock—the price they think reflects the true value of a stock. They will buy the stock if it falls below their target and sell it if it rises above. Collectively these funds have trillions in capital which they can shift around to buy up stock they think is undervalued. Their demand is so enormous and elastic that you can not apply “supply and demand” thinking to stocks. The price of a stock is set by how large institutions value it, not by trading activity.
Some have argued that large scale shifts towards passive investing have reduced market elasticity. However, if this reduction of elasticity has occurred, it has been because of tens of trillions of dollars of AUM shifting towards index strategies—it is not the kind of thing that even the largest charity could impact.
For counterexamples at the extremes, we have retail-driven stocks like GameStop, AMC and BBBY, but these had small market caps. I’d also guess Tesla stock prices have been affected by retail investors.
Institutions’ price targets could be affected by retail investment, and not all strategies are based exclusively on price targets, e.g. momentum investing. Retail investment interest could be used a measure of sentiment and is evidence for future revenues, especially for companies whose revenues primarily come from consumer sales rather than business-to-business sales. For example, Tesla retail investment levels should be evidence for future Tesla vehicle sales.
Over 40% of Tesla shares are owned by retail investors compared to less than 25% of Microsoft shares and around 30% of NVIDIA and Google shares, so I doubt that the difference in percentages is mostly explainable by passive investing. I’d guess >10% of Tesla shares are owned non-passively by retail investors.
EDIT: The Tesla stock price also seemed to be affected by the first stock split, another reason to believe it is driven by retail investment or at least strategies not based exclusively on price targets. According to your model of markets, stock splits in general seem irrational for companies to do and shouldn’t really affect prices, but I think the point is to increase demand for the stock, by making it more affordable for retail investors (or giving that perception; maybe it’s mostly just buying positive sentiment). NVIDIA and Google have done stock splits in the past few years, too. EDIT2: It seems there are other possibly more important reasons, e.g. “The most common ones are to achieve an optimal price range for liquidity, to achieve an optimal tick size and to signal managements’ confidence in the future stock price.” http://erepository.uonbi.ac.ke/handle/11295/10052
Maybe all of this is less applicable to the companies contributing most to advancing AI, though. But either way, I’d expect our impact on company capital and work (e.g. advancing AI capabilities) to be pretty negligible in expectation anyway, so I don’t think EAs or EA orgs should hold back from investing because of such concerns. We’d still be a tiny share of non-passive retail investment.
Yeah I’ve asked the same question (why invest in AI companies when we think they’re harmful?) twice before but didn’t get any good answers.
Right? I might be misunderstanding something about investing, but my presumption was that if you invest in a company, you help it do more of what it does. Please do correct me if I’m wrong.
There’s no good evidence that investing in a publicly traded company helps it at all. You are buying shares from other shareholders, not from the company itself, and the company does not care who hold’s its shares. Investing in a venture capital deal is a different story. It would definitely not be advisable for a charity to provide VC funding for a company opposed to it’s charitable mission because they could plausibly expand the total available funding for the company.
In general, it’s very difficult for stock market investors to make bets related to large shifts in technology because the benefits of such growth are typically concentrated in firms that do not yet exist. Someone who saw the economic importance of operating systems in the 1960s couldn’t have capitalized on that foresight because Microsoft and Apple weren’t founded until the 1970s.
It’s also unclear whether the benefits of AGI would go to companies which create AGI itself, companies which control critical data needed to train AGI, or companies who are well positioned to utilize AGI. For example, you can imagine a world where Google and Microsoft both develop AGIs but make little profit from them because stiff competition keeps the price of using an AGI extremely low.
A good strategy may be to invest in all marketable securities (with an ETF like VT and an equivalent bond ETF). You’d be betting that any company might experience rapid growth due to AGI.
Another strategy might be to tilt your portfolio based on regions which seem to be hubs for AI innovation. For example, since most AI innovation seems to be concentrated in the US right now, you could choose a US total market ETF such as VTI. I believe such an investment decision would be a mistake, however. Although US firms are developing AI rapidly, international firms seem to be equally well positioned to utilize AGI.
In the short term, some companies may be well positioned to capitalize on the AI boom. Microsoft comes to mind, as do manufacturers of data center equipment (chips, cooling equipment, etc.). However, these short-term possibilities are already priced into those securities. In the long run, however, AGI could easily destroy the strong market positions of these companies—why pay for Microsoft Word when you can simply have an AGI develop a new word processor just for you? Why by a chip fab system from ASML when an AGI can design you a new, better semiconductor platform?
Edit: clarifying what I think is priced into the market right now.
This post argues that markets do not seem to be expecting an AGI explosion (which, of course, could be interpreted as evidence that such an explosion is unlikely to occur and the market is correct to not price the possibility in).
In that paragraph I meant to refer only to the “AI boom”—essentially all the recent LLM stuff. In general I don’t think it matters to the investor whether the possibility of AGI is accounted for in the markets because the benefits of such growth would likely be concentrated in firms that do not yet exist.
The post you linked to also discusses the possibility of trying to use incorrectly priced debt instruments to take capitalize on the potential development of AGI. However, such a strategy is not realizable in practice because you’d need to find a counter-party/lender. At best, you’d end up with a callable debt—not very useful over an extended time horizon.
If you haven’t seen, a long analysis of this:
AGI and the EMH: markets are not expecting aligned or unaligned AI in the next 30 years
TLDR: bet on real rates rising
Against using stock prices to forecast AI timelines
Some cautionary points ¯\_(ツ)_/¯
Tricky question. I generally agree with Erin’s reply.
A possibly clever but risky strategy might be to think about where an AI would get its data from, and invest in companies and protocols that are well-positioned as automated, high-speed feeds of reliable data. Crypto oracle protocols such as Chainlink are aiming to do that for smart contracts, but they could also do it for AIs. The difficult issue is how secure any such data feed services are to cyberattacks or game-theoretic/consensus protocol attacks on their data integrity.