Antifragility in complex dynamical systems • Hacker News Board

Re: Antifragility in complex dynamical systems

#111

Post by soulofmischief » Wed, Aug 14, 2024, 8:35 AM UTC

taeric wrote: post #109

soulofmischief wrote: post #107

Earlier quoted context omitted.

I guess it's important that the negative signal be relatively small compared to the affected system, like antivenom doses, but also it needs to be predictable/differentiable to some degree. If two different doses of venom never had the same molecular patterns then it would be impossible to immunize against it. A one-time supernova event would destroy a solar system, but the universe itself is antifragile to supernova…

Right. If you have systems that can adapt to stimulus in ways to prepare for recurring stimulus, you can look like this antifragile system. If the answer is to just be bigger... It is not a compelling model. So building your immunity by using low doses doesn't mean you are antifragile. It means you are trainable. And even then, you are within tolerance levels. And cannot ever withstand some poisons. Indeed, training…

I don't think it's very common to experience antifragility outside of abstract complex systems such as economic markets. My intuition is that physical systems which exhibit antifragility tend to be metasystems, and the antifragility is described in the context of achieving specific states of subsystems.

For example, with the universe in relation to supernovas, we can say that a supernova wipes out a solar system. But new solar systems then take their place, with a different metal distribution that may be more conducive to advanced intelligence. If advanced intelligence is the goal, we might say that the universe is antifragile to supernovas, or more generally to the effects of gravity.

However, it's definitely a matter of perspective given that supernovas themselves are entirely inconsequential to the universe as a whole, and are only relevant to subsystems at specific emergent layers within the universe.

We could also form similar conclusions about natural evolution in general.

We might even be able to generalize further and posit that there is a natural tendency for order to arise out of chaos, and so order itself is antifragile to chaos, given that order thrives under the right chaotic conditions[0] and isn't simply immunized against it. So antifragility may exist as a fundamental property of the foundation of stuff, but it's not automatically an inheritable property of subsystems.

It all becomes very abstract, but at least we can define some constraints, namely that antifragility should be described in terms of optimization goals and measurement of specific states, and are thus subjective to the observer and not an innate property of the universe or any other system.

[0] https://en.wikipedia.org/wiki/Edge_of_chaos

Re: Antifragility in complex dynamical systems

#112

Post by RaftPeople » Wed, Aug 14, 2024, 2:52 PM UTC

bob1029 wrote: post #20

JackeJR wrote: post #12

Earlier quoted context omitted.

I am quite curious about your approach, Do you have an example code somewhere?

I'd start here: https://arxiv.org/abs/2406.12552 I will say that less is more and "The Bitter Lesson" applies here. Chasing biologically-inspired rabbits, such as STDP and LIF (see paper above/wikipedia), does seem to be a waste of time, especially when we have this thing entirely outside of biology that can arbitrarily replicate, serialize, mutate and simulate billions of instances of the same parent candidate in mi…

> Then clone that 1000 times and apply subtle mutations to it. Then, put these clones in an FFA arena and see who wins based upon a very well crafted fitness function. Then, do all of that over and over thousands of times per hour.

I spent many years on an ALife+ANN simulation that did this. For each new generation, I kept the top X% unchanged, the next Y% were changed a little, next Z% changed a lot, etc.

It was pretty fun and I wish I had the time+money to continue on a larger scale.

Re: Antifragility in complex dynamical systems

#113

Post by RaftPeople » Wed, Aug 14, 2024, 2:59 PM UTC

jcims wrote: post #84

bob1029 wrote: post #77

Earlier quoted context omitted.

> Unless physiological compatibility with a biological brain is considered an interesting endpoint? Excellent point. I am focused entirely on synthetic simulations without biological integration (for now). I think that could be an interesting next step - determining a way to modify a SNN that was trained in a synthetic time domain such that it can process real time signals. Training these things online is not really…

I had heard about spiking neural networks but I didn't really think about them until your post here. Lately I've been kind of deep diving into biological neurons (just a lay perspective) and what we know about their mechanism of action at the individual cell level. I'd also just watched the 8hr episode of Lex Fridman's podcast with the folks from Neuralink, so my brain was primed to make this connection and holy moly…

> Lately I've been kind of deep diving into biological neurons (just a lay perspective) and what we know about their mechanism of action at the individual cell level

I've been doing the same for a long time, it's a really fun and interesting topic. I really like reading about each new discovery about something that is computationally important, people don't realize how complex a neuron is and that the scientists don't have the full picture yet.

If you haven't looked into it, check out how astrocytes are involved in computation (e.g. visual processing), their internal calcium wave signaling and their control of neurons.

Re: Antifragility in complex dynamical systems

#114

Post by taeric » Wed, Aug 14, 2024, 3:16 PM UTC

guerrilla wrote: post #110

taeric wrote: post #97

Earlier quoted context omitted.

My argument here would be that this isn't that new, all told? And... I would be shocked to find a lot of disagreement there. Indeed, other threads are already pointing out that terms already existed that covered this general idea. My problem with antifragile, as often offered, is that it is positioned as something that gets stronger from being damaged, full stop. But... there is literally nothing on earth that would…

There are limits. It's not about arbitrary negative stmuli and the entire idea only applies to certain kinds of systems. These criteria are covered in Taleb's books. The limits are well known in biology.

I read his books. Will try them again sometime.

My gut is my main gripe is with many of the examples used. They often ignore that they are all on things that are in a healthy state of growth and that what is largely happening is that growth is refocused based on stimuli. More, there is no guarantee that you can refocus growth based on feedback in a way that will be ideal for future stimuli.

Which, I confess is silly for me to really get upset about. Luckily, I'm confident I'm sounding far more upset on this message board than I am on it. :D

Re: Antifragility in complex dynamical systems

#115

Post by taeric » Wed, Aug 14, 2024, 3:19 PM UTC

soulofmischief wrote: post #111

taeric wrote: post #109

Earlier quoted context omitted.

Right. If you have systems that can adapt to stimulus in ways to prepare for recurring stimulus, you can look like this antifragile system. If the answer is to just be bigger... It is not a compelling model. So building your immunity by using low doses doesn't mean you are antifragile. It means you are trainable. And even then, you are within tolerance levels. And cannot ever withstand some poisons. Indeed, training…

I don't think it's very common to experience antifragility outside of abstract complex systems such as economic markets. My intuition is that physical systems which exhibit antifragility tend to be metasystems, and the antifragility is described in the context of achieving specific states of subsystems. For example, with the universe in relation to supernovas, we can say that a supernova wipes out a solar system. But…

My intuition is that antifragile is often mistaken for controlled/directed growth. In that it works well in examples where things are able to die off and to be regrown. Muscles, in that regard, are able to withstand tears and are able to grow back. They will do so largely where they were torn.

That said, loving this exploration of it! Thanks!