December 30th, 2024
December 30th, 2024
Are there backward Arrows of Time out there?
Are there backward Arrows of Time out there?
At the moment, there are two places where I could imagine that (in principle) would have a backward arrow of time, i.e. situations where thinking (i.e. trying to compute probabilities) backwards would be easier than going forward
At the moment, there are two places where I could imagine that (in principle) would have a backward arrow of time, i.e. situations where thinking (i.e. trying to compute probabilities) backwards would be easier than going forward
Asset prices: the rationale here would be that 'forward sparsity' would imply something is easy to predict (as a time series), and intuitively, this would lead to more or less 'obvious' arbitrage opportunities being left open; there would in fact be a market pressure to remove these, and indeed to reduce the sparsity of the forward dynamics
•
Asset prices: the rationale here would be that 'forward sparsity' would imply something is easy to predict (as a time series), and intuitively, this would lead to more or less 'obvious' arbitrage opportunities being left open; there would in fact be a market pressure to remove these, and indeed to reduce the sparsity of the forward dynamics
These are environments where the (implicit) goal is definitely not to make anything easier in terms of understandability
These are environments where the (implicit) goal is definitely not to make anything easier in terms of understandability
Arrows of time in nature
Arrows of time in nature
The question of the presence or not of an arrow of time in nature is probably quite subtle
The question of the presence or not of an arrow of time in nature is probably quite subtle
My initial intuition is that besides the asymmetry related to the creation of entropy, there is not much of an arrow of time in nature, because there is no 'forward-sparsity' enforced by nature (unless there is a computing device)
My initial intuition is that besides the asymmetry related to the creation of entropy, there is not much of an arrow of time in nature, because there is no 'forward-sparsity' enforced by nature (unless there is a computing device)
Vassilis thinks it is not necessarily so simple, because the observables we care about (macroscopic ones) in various systems definitely introduce a description bias towards things that are simpler (and conceivable simpler forwards, rather than backwards)
Vassilis thinks it is not necessarily so simple, because the observables we care about (macroscopic ones) in various systems definitely introduce a description bias towards things that are simpler (and conceivable simpler forwards, rather than backwards)
And that's a plausible point... it is still not very clear
And that's a plausible point... it is still not very clear
For games like chess, it is not unconceivable that the description of a game by top players (at least if the state corresponds to the positions) would be easier backwards than forward: there is definitely something about a forward game that should not be predictable by a small model, by definition of what it means to have a very strong model, it sees things that weaker models don't see, while there is nothing so clear about the backwards dynamics
•
For games like chess, it is not unconceivable that the description of a game by top players (at least if the state corresponds to the positions) would be easier backwards than forward: there is definitely something about a forward game that should not be predictable by a small model, by definition of what it means to have a very strong model, it sees things that weaker models don't see, while there is nothing so clear about the backwards dynamics
Arrows of Time with Causality
Arrows of Time with Causality
January 28th, 2025
January 28th, 2025
One of the promising things with the LLM Arrow of Time is that it appears to bridge the arrows of time in a thermodynamics context with that appearing in the context of probabilistic causality
One of the promising things with the LLM Arrow of Time is that it appears to bridge the arrows of time in a thermodynamics context with that appearing in the context of probabilistic causality
Basically, the statement is that if we observe a system that computes, there is a causality asymmetry, but if we take a 'typical system' with some evolution, there is no such asymmetry
Basically, the statement is that if we observe a system that computes, there is a causality asymmetry, but if we take a 'typical system' with some evolution, there is no such asymmetry
Ultimately, all the 'causality' (in the probability sense) things are related to either a construction by humans or a human description of things; the 'traditional' notion of such causality would be really in terms of cause and effect, in terms of actions taken versus not taken
Ultimately, all the 'causality' (in the probability sense) things are related to either a construction by humans or a human description of things; the 'traditional' notion of such causality would be really in terms of cause and effect, in terms of actions taken versus not taken
The 'sparse' definition (which is considered weaker) is definitely more fundamental, as it is related to a descriptive view of the world
The 'sparse' definition (which is considered weaker) is definitely more fundamental, as it is related to a descriptive view of the world
Now, there is also the notion of causality which comes in physics with the ideas of special relativity, and also the notion of the arrow of time
Now, there is also the notion of causality which comes in physics with the ideas of special relativity, and also the notion of the arrow of time
The question is a little bit about the 'epistemic arrow of time', which says that there are no memories of the future
The question is a little bit about the 'epistemic arrow of time', which says that there are no memories of the future
.