

Didn’t read the paper, but the description suggests that the models which were tested were specifically created with a very limited set of training data. I can understand the argument that this might allow researchers to evaluate one property in a simplified environment, namely logic generalization. But it could also be argued that general logic is an emergent property, and limiting the LLM in this way prevents it from reaching that complexity threshold.
I’m not sure if the intent of including the two transformations in the training data was an attempt to provide the LLM with at least one opportunity to generalize between known phenomena before asking it to further generalize.
On the one hand, absolutely yes. Trump has solid control of one of the political parties, but in general, is a very unpopular politician. Yet our political systems have become so dysfunctional that we run a serious risk of him destroying the rule of law and creating an authoritarian presidency. There’s no authoritarian-proof political system, but we can do a lot better than this.
On the other hand, I think any reform that sufficiently addressed our current problems would be the end of the Democratic party as well. Getting a political party to sacrifice itself for the greater good is a tough sell. Not impossible, but I think it would look more like a popular takeover of the Democratic party to be used as a disposable vehicle. The Democratic leadership and their donor class will absolutely be opposed.
The tea party and Trump have shown that seizing a party from its current leadership is not impossible.