What is this?

Bonus: Jim on The Stoa, COVID & Game B 181

• was wondering if game b is just kind of a natural evolution that will be occurring from game a so instead of
• seeing game b is a reaction to game a it's more like an evolution and as a people that are involved in game being
• absolutely appropriate to think of game b as a natural evolution from game a unfortunately it's not the only one i

EP118 Matt Ridley on How Innovation Works 180

• rational optimist and the evolution of everything he's also chairman of the international
• and of course with the science education i'd always been aware of evolution but it had just been one of those things in
• you know that was partly why i wrote the book the evolution of everything was to try and point out that once you
• in the universe and you talk about evolution as one form and human invention which is another can
• but energy to create order is what uh biological evolution does it harnesses energy to
• i recall from my readings of the evolution of science of thermodynamics a whole lot of it was
• being this phase change in human evolution in some sense is true however you know there was an awful lot

EP51 Richard Bartlett on Self-Organizing Collaboration 179

• part-time work as a next natural evolution for some of the people then becoming full-time and actually living

Currents 016: Robin Hanson on Are We Living In A Simulation? 179

• linear evolution equation in that very high dimensional vector space which is

EP63 Michel Bauwens on P2P & Commons 178

• realm every cell has a membrane but the membrane has been designed by evolution very carefully for what it both what's

Currents 038: Connor Leahy on Artificial Intelligence 176

• for instance i'd love to point out that human evolution is now in the west at least substantially under the control

EP38 Tristan Harris on Humane Tech 173

• tree mod nature's seldom profligate in her gifts from evolution and one can go into the cognitive psychology of

EP137 Ken Stanley on Neuroevolution 171

• know that this is one of my obsessions neuro evolution when i retired from business back in
• tool it's kind of you know neural evolutions i like to say it's the weakest of all methods but can have
• neural networks through neural evolution and ken was one of four authors the others
• evolved over the course of natural evolution but now try to translate that into an
• evolve things inside the computer and so what neural evolution is is basically trying to evolve these neural
• interesting that the field particularly of neural evolution but even neural computing has always been
• working in this field try not to use the word evolution any theory on why it's kind of been a
• prove things and one of the things about evolution when you think about evolution is that it's sort of like
• to evolving stuff and so when you combine evolution with neural networks it's a little bit i
• become extremely popular um but evolution remains kind of on the outside and so
• roll the die type of random stuff but the real issue is that neural evolution to me is not really about optimization
• to me is not really about optimization no evolution is about how did we get something astronomically complex out of
• profound changes that's the kind of thing i think no evolution is about and ultimately understanding how
• came from pond scum right and it got there via evolution at least if you're not a believer that somewhere along the
• you talk about billions of years of evolution it's in some ways even more amazing because the neuron only arose
• it's like what is it that we don't know here like evolution is a theory that um that's you know kind of
• that ultimately help us on the road to ai as well since after all the evolution led to intelligence
• nature in evolution and you can think of it a little bit like an automatic breeding
• very well done so now we get the neural evolution instead of using backdrop let's take the core case we're using
• let's take the core case we're using evolution to discover weights and topologies
• performance and then running evolution back on the original strings is essentially the
• simple-minded core version of neural evolution is that fair enough that is fair so there's a lot of
• actually to recognize for people who think evolution is random it's definitely not a random process
• here because evolution isn't out there to be like the best
• think it's more interesting that evolution is more of a creative algorithm than an optimization algorithm
• and so what what you can do with evolution is that you can it is more of a kind of a creative
• which they did of course in real evolution that's not addressed by gradient descent
• but it would be also something extending it and evolution is a natural thing to do
• in deep learning between deep learning and their evolution where people are getting interested in
• that that then move it closer to what evolution does and the distinction is blurring in
• weights back in the history of neural evolution there was quite a bit of exploration of
• it's easy to evolve topologies in neural evolution because you did the problem in gradient-based method
• and and there were huge proliferation of neural evolution algorithms that would evolve topologies
• fits in this history of neural evolution sure yeah so around the year 2000 yeah i was in
• at the time like crossover was a big problem in their evolution because if you have like two different topology
• diversity like now like what evolution does it's like when you get a really cool new idea like you
• the most popular tooling in neural evolution that i could find at least yeah yeah it really became popular
• the early history the basics of neural evolution neat which was for quite a while
• know 2000 vintage other forms of neural evolution ain't going to cut it
• this world and where does neural evolution fit in in the giga size and eventually bigger
• thinking about what what do we need evolution for and it basically means that we need to be a
• we need something different maybe evolution is it now things look a bit different now that we can actually back
• discovered yet and it is a possibility that evolution is one way that we could get to those
• from humans um but it's not out of the question that evolution is on the table for that
• thinking um where evolution is basically a divergent system which is about how to
• published in the last two or three years where neural evolution seems to be at least as good and in some cases better
• of weights like it's an uphill battle for evolution because granny descent does have uh the
• like there's these evolution strategies that actually can optimize
• way that we use gpus and so forth evolution is sort of a different paradigm in a way so it's not as clear
• we've computed through something like an evolution strategy and so yes i think while it's getting
• uh it's not completely off the table um the evolution has some theoretical opportunity here to
• just wait space yeah it is interesting the co-evolution hardware software
• gpuis gradient descent and the rest was this very interesting co-evolution of hardware tools and approach and
• reason then perhaps general purpose neural evolution for all aspects
• yeah it's i mean you're right there's been a co-evolution it's quite interesting
• and um the the tightness of that co-evolution is like the algorithms improve and then the hardware improves
• quality results they got there could have been done through evolution they were later so we get you can get
• you know the whole world would have been circular circling around the evolution strategies like i'm not sure but
• just you know one of the worst things that can happen in evolution is if is if the diversity drains out of your
• go read his book let's get back now to neural evolution one area that even in my own little work
• using evolution to evolve evolution so evolving the
• deep learning might a hyper parameter b and how you might use evolution to search the space of hyper parameters
• learning rate or something like that in evolution it's things like the mutation strength
• a meta learning level or a meta evolution level um and it's expensive uh to do things at a
• do and so evolution is you know one of the options for that i mean because
• of the hyper parameters and so the evolution doesn't have to worry about that it can just evolve the
• another potential opportunity for evolution and by the way it also points to the fact that you know
• when we talk about applications of evolution and um things like novel research it's also not
• important eventually like neuro evolution has to be evolving and directing coatings if we want truly
• like the the which basically you would do is like the combination of evolution and development together and how they
• the evolution back in cppn space exactly yeah so you might actually be
• yeah so you might actually be effectively doing evolution in a low dimensional space
• dimension sequence affordances and the evolution of complexity which is a book i'd recommend you read by the way
• complex patterning and how you throw all that together in evolution and you can end up with some stuff that
• favorite topics which is baldwinian learning or the evolution of learning well i want to talk briefly about that
• well i want to talk briefly about that one the evolution of learning itself and learning mechanism
• basically like once you a lot of the time in their evolution if you evolve a neural network like you find a
• mechanism of learning itself was evolved and in that sense evolution is a meta-learning algorithm not just a
• and so if there's a holy grail for neural evolution it can't leave that out i mean that that's like a fundamental
• but you can do that you know why not like evolution can encode rules we call them
• like which are actually inspired by the neural evolution work so i think it's an example of the um
• like two competing paradigms like nerve evolution versus deep learning but they're very related to each other like
• on wildly recurrent networks and yet evolution doesn't care right you can put neural evolution to
• evolution doesn't care right you can put neural evolution to work on the most tangled crazed
• obsession with gradient descent that neural evolution might be able to open up much much wider
• recurrent networks to get at and that maybe evolution is the obvious tool for exploring that kind
• kind of a stereotype structure um and that neural evolution can really think out of the box in terms of like

EP93 Brent Cooper on Critique, Consensus & Politics 167

• about grand narratives etc and for fun i looked into the evolution of critical legal studies from started
• provide raw material for experiment and social evolution using components borrowed from various of these

EP116 Doug Erwin on the Cambrian Explosion 165

• domains the evolution of animal regulatory genomes and the origin and early
• genomes and the origin and early evolution of animals and the end permian mass
• is a radically rapid evolution and diversification of life again
• a bit about the difference up front between micro and macro evolution maybe tell us a little bit about
• 1800s even before darwin was born is whether or not evolution occurs on two levels
• adaptive changes that happen with the evolution of different kinds of flies
• it occurs within populations and is often responsible for the evolution of new species
• twice so the evolution of nervous systems is actually a really
• when those who appeared in the early evolution of animals and i've been to several meetings in london
• actually at the royal society just about the evolution of nervous systems and this is a
• when did these mutual evolutions these separate evolutions happen
• mutual evolutions these separate evolutions happen so the critical question in all of this
• i mean that's what technological evolution is about creating environments for various kinds
• to understand in the early evolution of animals is
• than people often tend to think that evolution may not have been solely a function of natural selection as darwin
• the short answer is that i think the evolution of life is actually pretty easy
• you start getting cumulative evolution as you did in humans 50 000 years ago or whatever does the rest of