Which investments for aligned-AI outcomes?
16Tamsin Leake
2tailcalled
2Tamsin Leake
15gwern
4tailcalled
4gwern
3Timothy Underwood
2JBlack
1RogerDearnaley
New Answer
New Comment
2 Answers sorted by
16-3
I continue to think that, in worlds where we robustly survive, money is largely going to be obsolete. The thing that maximizes the terminal values of the kind of (handshake of) utility functions we can expect probably aren't maximized by maintaining current allocations of wealth and institutions-that-care-about-that-wealth. The use for money/investment/resources is making sure we get utopia in the first place, by slowing capabilities and solving alignment (and thus also plausibly purchasing shares of the LDT utility function handshake), not being rich in utopia. (maybe see also 1, 2)
12
I think it's exceedingly unlikely (<1%) that we robustly prevent anyone from {making an AI that kills everyone} without an aligned sovereign.
*
154Some people say to invest in land because land prices keep going up. But I can't help but wonder if there's a pyramid scheme element to this. In the end, we'd expect land to increase in value because people who need land get more productive and can therefore afford to bid it higher, but if the value of human labor drops, shouldn't we expect the value of land to drop too?
In the macroeconomic division of an economy into 'labor', 'capital', and 'land', 'land' is a rather misleading sort of term (similar to 'non-shared environment' in behavioral genetics). It's basically a residual definition, 'all fixed facts about the world which are not a human doing work ("labor") or an improvement/change ("capital")'. To help distinguish the two, I'll capitalize the factors. To quote WP:
In economics, Land comprises all naturally occurring resources as well as geographic land. Examples include particular geographical locations, mineral deposits, forests, fish stocks, atmospheric quality, geostationary orbits, and portions of the electromagnetic spectrum. Supply of these resources is fixed.[1]
So when someone says 'maybe you should invest in Land if AGI', what you're thinking is something vaguely like "buy land"? 'I should buy suburban houses or random patches of desert in Arizona because someone will buy them from me because... AI??? but why? can't people just build their houses somewhere else or put on another story if they need an additional bedroom?' That is indeed confusing.
But what you should actually be thinking is something like '"Land" is the price of a cubic meter of space in central Manhattan next to the stock exchange which minimizes latency for AI HFT traders due to the fixed speed of light', or 'Land right now is an acre of land in the Appalachians where the only ultra-pure quartz in the world can be mined and which is a critical irreplaceable rate-limiting input to the TSMC chip fabs churning out the most advanced GPUs which run the AIs', or possibly, 'Land has become an apartment building in Cerebral Valley where the handful of AI researchers still capable of meaningfully improving DL architectures will reside next to their office building and willingly pay rents of hundreds of thousands of dollars, because, with their annual incomes of hundreds of millions of dollars, they don't want to waste even an extra precious minute commuting to the office in the final days left (or aren't allowed to work remote)'.
Essentially, in the slow takeoff, as there is a tsunami of Labor and Capital pent up behind a few bottlenecks (almost by definition for a 'slow' takeoff - if there were no bottlenecks, it'd be 'fast'), whoever owns those bottlenecks may be able to extract enormous profits. By definition those bottlenecks are now 'Land' (because that's all that's left if you assume abundant Labor and Capital), and so something in it must become extremely valuable. Concretely, if the world goes into hypergrowth where all deployment and economic activity becomes gated by getting more GPUs, then if you control some unique resource for GPU manufacturing, you may be able to exact a toll of billions or trillions, like the quartz mines. You still produce the same quartz, but now it's worth vastly more because it limits growth, rather than Labor or Capital components. (Imagine in this scenario that each ton of quartz you sell is worth a thousand immortal von Neumanns running 24/7, and that is now the major contributor to how Capital is turned into more Labor - no quartz, no von Neumanns - you could probably charge a lot for it, right? Millions, at least.)
Hopefully put that way, it's more obvious both why Land might become super-valuable and also why you, as an individual, may have serious trouble investing in the right Land. (There is no 'Land ETF' on Vanguard, any more than there is a 'Capital ETF' or 'Labor ETF'.)
4
I don't think one should 'do' anything about it. It is simply an implication of one model of economic growth. On a personal level, equities are so much better in many ways that even if they would be outperformed by some hypothetical ideal 'land' portfolio, that you'll be just fine investing solely in equities. (What must be avoided is stuff like cash, which won't grow or benefit, or investing into fixed instruments like debt - if you expect any sort of takeoff soon with any probability, investing in a 30-year Treasury or something would be insane. You would want to be taking out 30-year loans instead!)
3
You might capture value out of that relative to broad equities if the world ends up both severely deflationary due to falling costs, and where current publicly traded companies are mostly unable to compete in the new context.
In the end, we'd expect land to increase in value because people who need land get more productive and can therefore afford to bid it higher, but if the value of human labor drops, shouldn't we expect the value of land to drop too?
Will the value of human labour drop? The immense scale of economic disruptions in a future where there is widespread AGI makes the question itself hard to interpret. There are many theories of value that are relatively similar in modern economies but come apart in very different ones such as these hypothetical futures.
Regardless of that, I think that returns on any relevant capital investments would skyrocket for a while, followed by pretty major rises in rates of return on investment in whatever the new limiting factors turn out to be - very likely some forms of economic rents.
In the end, you presumably don't much care about returns directly in monetary terms, but in what it gets you in the new civilization. Goods, services, influence, security - whatever. I think land as a general rule is pretty unlikely to get less of these than it does now, and in some cases likely gets very much more - though as with any investment there will be risks.
This is basically an all-in bet on land, which I guess could be a great bet if land increases in value, but extremely bad if land decreases in value.
The value of land (other than farmland) depends wildly on its location: cities are far more expensive. Cities are currently popular for two reasons:
- High population densities and short distances make delivering a wide range of goods and services more efficient by shortening the lengths of pipes, wires, roads, and transmission networks.
- Having a lot of people working in the same industry in the same place produces social network effects and makes it easy for people to change employers without having to move.
Of these effects, 2. would go away, but 1. would remain (since of a lot of it is basic physics). So while cities would still be favored, they might become less favored. Thus the value of land might change. So I'd recommend against investing in land in cities where effect 2. is currently strong (like the SF Bay Area, Seattle, Los Angeles, New York) and instead investing in land either outside a city (but perhaps near one) or where 2. doesn't currently apply (such a currently-blighted city in the rust belt, say Detroit). Also allow for climate change and sea-level rise.
Curated and popular this week