The Wittgenstein Test
There’s a debate about whether LLMs such as GPT-4 have some sort of consciousness. That is, are they able to think, analyze, understand, etc., in a way that is somewhat analogous to how the human mind works? Back in 2022, Gwern argued that LLMs have agency:
Powerful generative models like GPT-3 learn to imitate agents and thus become agents when prompted appropriately. This is an inevitable consequence of training on huge amounts of human-generated data. This can be a problem.
Is human data (or moral equivalents like DRL agents) necessary, and other kinds of data, such as physics data, free of this problem? (And so a safety strategy of filtering data could reduce or eliminate hidden agency.)
I argue no: agency is not discrete or immaterial, but an ordinary continuum of capability, useful to a generative model in many contexts beyond those narrowly defined as ‘agents’, such as in the “intentional stance” or variational approaches to solving physics problems.
Thus, a very wide range of problems, at scale, may surprisingly induce emergent agency.
At one point he referred a famous anecdote attributed to Ludwig Wittgenstein:
He once greeted me with the question: ‘Why do people say that it was natural to think that the sun went round the earth rather than that the earth turned on its axis? I replied: ’I suppose, because it looked as if the sun went round the earth.’ ‘Well,’ he asked, ‘what would it have looked like if it had looked as if the earth turned on its axis?’
He is asking those who are skeptical of LLMs having agency to consider what their behavior would look like if they did have agency.
When I first discovered the Wittgenstein quotation, I immediately recognized that it almost perfectly illustrated the point I was trying to make about the crisis of 2008 being produced by tight money. Indeed I used the quote as an epigraph for my book entitled The Money Illusion.
It occurs to me that Wittgenstein’s insight has many applications in economics. Consider the following examples:
1. Economists often argue that the US current account deficit is caused by low rates of domestic saving. In contrast, for many people it looks like the deficit is caused by US firms being unable to compete with cheaper foreign rivals. One response to those with this common sense view might be: What would it look like if it looked as if a lack of domestic saving caused the exchange rate for the dollar to appreciate strongly enough to where many US firms were no longer competitive with foreign rivals?
2. Economists often argue that the US benefits from importing foreign goods. In contrast, to many people it looks like imports hurt the US economy by negatively impacting domestic firms that compete against imports. One response to those with this common sense view might be: What would it look like if it had looked as if international trade caused the US economy to reorient production toward goods in which we have a comparative advantage, boosting total output?
3. Economists often argue that inflation is caused by the effect of monetary policy on aggregate demand. In contrast, to many people it looks like inflation is caused by greedy firms raising prices and boosting profits. One response to those with this common sense view might be: What would profits look like if it had looked as if monetary policy caused aggregate demand to rise sharply at a time when nominal hourly wages were sticky?
The Money Illusion is basically a long examination of why I believe the Great Recession was caused by tight money, rather than housing and banking problems. Step by step, I show that the facts fit the tight money explanation far better than the conventional view. I show that all of the stylized facts that are typically viewed as causal factors are actually better viewed as the consequence of tight money. I also showed that the conventional view is inconsistent with most stylized facts, such as the fact that the sharp economic slump began several months before Lehman failed and the fact that the Great Recession was worse in Europe, despite the subprime crisis being in America. (Europe had tighter money.) And before you try to come up with an ad hoc explanation for those anomalies, recall that (in 2008) conventional economists almost universally expected the Great Recession to be worse in America, because they were working with the wrong model of the economy.
A geocentric model of the solar system with epicycles may work for a while, but with a close enough examination of the data you’ll eventually find anomalies. And I found dozens of inconvenient anomalies. Read the whole thing: