Whereas some AI researchers (and Metaculus AI questions) have substantial credence in high-level machine intelligence (HLMI) being developed within only a few decades, others have argued that the stock market provides evidence against this view (see, for example, this blog post).
The claim is that if HLMI was close, the market capitalisations of tech companies should be much higher than what we're seeing, given the presumed huge economic advantage of the company developing it.
Yet things might not be so simple. If we start with a tech giant's owning a large fraction of world output, and multiply through by the uncertainty of their actually achieving world dominance, stocks being potentially worthless due to bad corporate governance and eventual institutional collapse following HLMI, and so forth; we might get to a point where current prices do make sense.
Moreover, Robin Hanson published a paper 20 years ago arguing that the stock markets at that time did support an upcoming period of a fast exponential "growth mode", following three previous such shifts in human history: "hunting", "farming", and "industry".
To get a better sense of this disagreement, we ask:
Conditional on human-level machine intelligence in less than 20 years from 2019, is Alphabet's share price correct to within $300B?
The focus on Alphabet can be justified by noting their dominance in terms of for-profit labs dedicated to AGI (consider for example that DeepMind's close to 1000 employees constitutes by far the largest effort explicitly dedicated to pursuing AGI, and that the number of papers accepted papers by DeepMind and Google Brain to NeurIPS 2017 was almost twice that of the second ranked institution).
We shall use AI Impacts definition and say that “High-level machine intelligence” (HLMI) is achieved when unaided machines can accomplish every task better and more cheaply than human workers.
Alphabet's share price refers to the price of Class A shares (non-insider shares with voting power), with ticker NASDAQ: GOOGL.
We shall say that a share price is correct if it is equal to whatever the time-discounted sum of returns (in the form of dividends, buybacks, acquisitions and other premiums such as voting power) turns out to be.
There are some resolution subtleties which make this question more experimental than most other questions.
First, in the world where the conditional of this question has been verifiably satisfied, the likelihood of this site existing in a similar format, and an admin taking the time to resolve this question, is fairly low.
Second, the correct prediction changes as Alphabet's share price does.
Hence, the purpose of this question is to experiment with a way of eliciting reasonable, quantified discussion and aggregation of thoughts on the stock market's predicting HLMI; rather than winning Metaculus points.