Chat session 1 (12:30-2:30)

workshop-oee4 chat record of 1st session 12:30-2:30

12:36:10 From Lisa Soros to Everyone : #workshop-oee4

12:39:15 From Alastair Channon to Everyone : https://openended.protolife.org/

12:39:23 From Lana S. to Everyone : :astonished:

12:41:13 From Olaf Witkowski to Everyone : @Lana what do you think vs. Mimosa?

12:44:23 From Lana S. to Everyone : I don’t see it as “versus” at all! There can never be too many places to bring small communities together. I do hope Mimosa can be used to advance collaborations and keep up with debates and consensus, and maybe have some synergies with this OEE Discourse forum

12:45:09 From Hiroki Sayama to Everyone : This is great, thanks Norman! How can we make contributions to this repository?

12:45:12 From Olaf Witkowski to Everyone : Yes, I didn’t mean versus as in opposed to, but as in idea transfers :slight_smile:

12:45:23 From Bert Chan 2 to Everyone : Very neat website!

12:45:26 From Olaf Witkowski to Everyone : Thanks Norman, that’s great!

12:45:40 From Odin to Everyone : Not a question, but a related idea on origin of life network: https://oolen.org

12:45:43 From Aymeric ViĂŠ to Everyone : Thanks a lot this is really nice

12:45:55 From Lana S. to Everyone : I guess journals used to be that “space” but they can never keep up with the pace of contemporary exchange

12:46:22 From Lana S. to Everyone : every research community should have such a forum

12:47:24 From Olaf Witkowski to Everyone : I think we should link (or even embed?) it in the current ISAL website, if you agree

12:47:24 From Lana S. to Everyone : :+1::+1::+1::+1:

12:47:53 From Lana S. to Everyone : I signed up :slight_smile:

12:49:31 From Norman Packard to Everyone : Of course links come for free… happy to discuss other forms of embedding/hosting.

12:50:06 From Lana S. to Everyone : Since @olaf nicely brought it up, shameless plug for Mimosa, the Open Collaboration Platform I’m building Loading Share Experience

12:50:13 From Lana S. to Everyone : let me know if you want to become a beta tester

12:50:42 From Olaf Witkowski to Everyone : @Norman I’ll ping Emily about it!

12:53:41 From Norman Packard to Everyone : “detection of quantitative change” => unsupervised learning in parallel w/ evolutionary process, observing it?

12:55:41 From Nicholas Guttenberg to Everyone : There’s a big conversation to be had there. Basically, different kinds of unsupervised learning might capture anything from quantitative to qualitative changes. If e.g. you try to predict/reconstruct the details, then variations on a theme still increase the entropy. But if you abstract and then try to reconstruct the abstractions…

12:56:13 From Nicholas Guttenberg to Everyone : The hard thing is putting tension on those abstractions so they don’t collapse to trivial things.

12:56:31 From Norman Packard to Everyone : Can unsupervised learning have the capability to “be surprised” as SS?

12:56:56 From Olaf Witkowski to Everyone : I know it doesn’t have to be, but when I hear qualitative change detection (from either within or outside the system), I can’t help but think of the subjectivity (which was partially addressed yesterday with Jessica’s coarse graining to escape individual subjectivities)

12:57:02 From Nicholas Guttenberg to Everyone : So for example, you could train a GAN to produce a distribution of data. Then given a new point of data, you could ask ‘does the latent exist which generates this new point?’

12:57:06 From Norman Packard to Everyone : as SS describes? If yes, maybe we don’t “always need post hoc measures”

12:59:12 From Steen Rasmussen to Everyone : We can measure whether a evolution/learning process is optimizing its fitness or whether it expanding its abilities.

13:00:17 From Nicholas Guttenberg to Everyone : (what SS is talking about right now can be mapped almost directly to meta-learning in neural nets, with nested ‘outer loops’ to train model-generators)

13:00:18 From Steen Rasmussen to Everyone : The hard problem is that the impacts depend on the context

13:00:29 From Olaf Witkowski to Everyone : @Steen I guess in that spirit in a discrete system like a CA you might simply count the number of implemented functions by each (emergent) entity

13:00:31 From Sharon Minsuk to Everyone : I would want to know how many levels of M0-Mn exist in nature. I wouldn’t want to hold artificial evolution to a higher standard than natural evolution! Evolution of multicellularity is an example that feels like M2 to me. What is higher? Gaia as M3? (If it exists)

13:00:35 From Norman Packard to Everyone : “expanding abilities” hard to measure without prior on space of possible “abilities”.

13:01:02 From Olaf Witkowski to Everyone : @Norman absolutely, I find it hard to define in general, out of the said CAs

13:01:59 From Odin to Everyone : I was about to ask the same that Sharon. If life as we know it is an open ended evolutionary system. What is the Bhanzaf model of our ever-branching species?

13:03:12 From Odin to Everyone : Sharon, I think it is possible to have a higher range in artificial (or other possible types of live) evolution than in the one we have on earth

13:03:59 From Sharon Minsuk to Everyone : Well, given that we are so far from achieving the OEE that nature exhibits, that might be a bit ambitious!

13:04:39 From Sharon Minsuk to Everyone : At least for now. :wink:

13:05:22 From Odin to Everyone : Experts may pop here, but cultural evolution and technological evolution has its own rules!

13:05:53 From Penny Faulkner Rainford to Everyone : I think the stages of models in nature and in artificial systems might be different enough that the number of levels here is not really a good comparison as a thing to “achieve” nature has a lot of pressures and limitations that artificial systems don’t

13:06:00 From Steen Rasmussen to Everyone : Question, agree or not: The problem is that it is not clear how to measure meaning/semantics as which properties is new depends on the context

13:06:07 From Hiroki Sayama to Everyone : There are many quantities in math and computational theory that are not definable or computable. Could it be that the open-endedness in this line of definition may also end up being one of such “immeasurable quantity”?

13:06:08 From Olaf Witkowski to Everyone : Absolutely, I think this is promising too

13:07:43 From Olaf Witkowski to Everyone : ^this, as in the measures mentioned in Jessica’s keynote talk

13:08:55 From Norman Packard to Everyone : Steen didn’t hear Nicks comment on abstraction issue.

13:09:58 From Norman Packard to Everyone : diagree.

13:10:05 From Olaf Witkowski to Everyone : What does it mean to “break the model”? Does it mean that we have a different (enough) mechanism, as captured under a mathematical standard measure?

13:10:10 From Nicholas Guttenberg to Everyone : this one is fun :slight_smile:

13:10:30 From Norman Packard to Everyone : obvious that no sign of biological evolution stopping its open endedness

13:10:44 From Nicholas Guttenberg to Everyone : Mixed agreement/disagreement there I think. I think historically bio evolution is sometimes not open-ended until something outside of biology kicks it, e.g. ‘boring billion’

13:11:31 From Lana S. to Everyone : I agree that we shouldn’t try to limit ourselves to “real” biology

13:11:33 From Olaf Witkowski to Everyone : Thanks Susan, obviously very thought/disagreements provoking :slight_smile:

13:11:48 From Stuart Bartlett to Everyone : yes, amazing talk, thank you Susan (y)

13:12:12 From Sharon Minsuk to Everyone : Penny, good point. But I think the “pressures and limitations” on artificial evolution may actually be the reason why natural OEE happens… pressure and limitation (i.e. “necessity”) is the mother of invention!

13:12:14 From Lana S. to Everyone : Maybe Takashi’s motto sums it well, “artificial life larger than biological life” ?

13:12:20 From Olaf Witkowski to Everyone : (About real biology: isn’t it going to be about a model of said biology in any case?)

13:12:34 From Sharon Minsuk to Everyone : *Sorry, I meant on natural evolution.

13:12:39 From Hiroki Sayama to Everyone : +1 @lana

13:13:34 From Lisa Soros to Everyone : “pressure and limitation (i.e. “necessity”) is the mother of invention!” → though too much pressure/limitation/pressure to optimize leads to convergence and the opposite of OE

13:13:57 From Penny Faulkner Rainford to Everyone : I think the biological limitations might be different which means that artificial systems would not be without pressures of limitations but that we can change those, possibly making it possible to have higher levels of models where the inability to change physics for instance might limit biology

13:16:29 From Nicholas Guttenberg to Everyone : I suppose my pet theory about OEE (rather than just OE) is that subsequent stages expand in spatio-temporal scope in order to erase realities of the smaller scales, and that for it to go forever there always has to be something you can’t erase at a given scale. So e.g. if you’re chemicals in soup, you’re at the mercy of the external environment (dilution, toxins, etc). By building something bigger than a molecule, a molecule can ignore the environment but now you have to maintain the bigger thing. So the bigger thing organizes into a film so that individual cells don’t deal with everything, but now the film does. So …

13:17:21 From Nicholas Guttenberg to Everyone : So if you want to exceed the physical universe in the artificial one, you need to have things which are unavoidable at any scale in our universe, but which are avoidable at some scale in the artificial one (or you just need there to be more things in general I guess)

13:17:56 From Nicholas Guttenberg to Everyone : however, if that’s the case, then how do you fit that inside of the physical universe? :slight_smile:

13:18:18 From Norman Packard to Everyone : but… cells don’t really escape molecular detail, however much they try. single molecules can seriously screw them up (e.g. virus).

13:18:20 From Olaf Witkowski to Everyone : Your pet theory sounds related to multilevel selection theory (if it had a good mathematical formalism, which I guess it doesn’t other than the Price equation perhaps)

13:18:24 From Olaf Witkowski to Everyone : @Nicholas

13:19:05 From Nicholas Guttenberg to Everyone : Biology is interesting in that it leaves a line to the smaller scales rather than deleting them entirely. Maybe related to Kuni Kaneko’s ‘minority control’ theory?

13:19:54 From Nicholas Guttenberg to Everyone : Whereas most endogenous coarse-grainings (using Jessica Flack’s terminology) are more thorough about deleting that link

13:21:10 From Odin to Everyone : Olaf, the multilevel theory has been advance a bit from a group related with Martin Nowak, Arne Traulsen and collaborators. I am not updated as I would love

13:21:43 From Dragana Laketic to Everyone : @Nicholas G. : “erase realities at smaller scales” - shouldn’t it be rather something like “abstract the details about smaller scales”?

13:23:17 From Nicholas Guttenberg to Everyone : @Dragana that’s why I added the evolution bit back in. As a pet theory, I’m proposing a hypothesis about the driving force. That is, e.g. along the lines of Watson’s stuff about Natural Induction, there are tensions at one scale which the system can escape by organizing at a larger scale, and that drives the movement to organize at that larger scale.

13:23:45 From Nicholas Guttenberg to Everyone : This could absolutely be wrong :slight_smile:

13:23:52 From Olaf Witkowski to Everyone : @Odin It’s related to Nowak’s work indeed, with the two Wilsons, but it’s not very satisfying to characterize open-endedness I believe, as it starts high level

13:25:06 From Penny Faulkner Rainford to Everyone : I think that the scales and abstractions are part of it, but artificial systems don’t just have to abstract they can actually drop their smaller scales and grow new larger scales if they need to continue with in the limits of the physical universe.

13:25:06 From Stuart Bartlett to Everyone : @Nicholas a theory for the self-engineering of emergence and its driving forces?

13:25:26 From Odin to Everyone : @Olaf, right, it is still a bit of two scales with a gap in the middle (and sometimes not filled because theoretical biologist do not fully agree with multilevel selection)

13:25:32 From Nicholas Guttenberg to Everyone : yes

13:25:51 From Stuart Bartlett to Everyone : sorry, which Watson is this?

13:26:17 From Dragana Laketic to Everyone : Is it Richard Watson?

13:26:33 From Nicholas Guttenberg to Everyone : Yeah

13:26:45 From Norman Packard to Everyone : Search results for 'watson' - OpenEndedness

13:26:46 From Dragana Laketic to Everyone : He had some interesting work on modularity and the likes.

13:27:15 From Olaf Witkowski to Everyone : @Norman nice live demoing

13:27:24 From Norman Packard to Everyone : :slight_smile:

13:33:26 From Norman Packard to Everyone : Pretty sure there is more than one example in current models archive (Models - OpenEndedness)

13:34:13 From Alastair Channon to Everyone : Qualification: autonomous artificial systems

13:34:24 From Norman Packard to Everyone : Does “indefinite scalability” of diversity simply mean diversity is an extensive thermodynamic variable?

13:35:00 From Steen Rasmussen to Everyone : @Alistair: What do you do when you can’t identify whether an object is “new” or not e.g., if it leans via Lamarckian learning?

13:35:12 From Charles Ofria to Everyone : @Alastair: I love the shadow system model as a control, but it requires us to be able to clearly identify what an individual organism is in order to remove selection in the shadow. How do you build a shadow system for a model where evolution changes the organisms (and thus WHAT is being selected, not just how). It seems like the moment the shadow system falls out of sync, it will no longer be able to produce relevant results.

13:35:48 From Lana S. to Everyone : (I hope this chat gets archived at @norman ’s website)

13:35:53 From Steen Rasmussen to Everyone : @Charles: agree, that’s another version of my question

13:36:25 From Charles Ofria to Everyone : @Steen: Agreed! I saw yours appear as I was typing mine.

13:37:13 From Norman Packard to Everyone : @Lana: will definitely do!

13:37:28 From Tanner Lund to Everyone : @Lana I agree. It would be a shame to lose the discussion/questions. Not much going on in the slack channel for this workshop.

13:39:34 From Norman Packard to Everyone : not so hard…

13:39:36 From Olaf Witkowski to Everyone : @Lana I agree, I just wish I could think of a good way to have (part of) this chat live directly on Norman’s platform

13:40:06 From Norman Packard to Everyone : Lamarkian system not so hard to detect innovations…

13:40:30 From Steen Rasmussen to Everyone : @Norman, please tell us how

13:43:13 From Norman Packard to Everyone : Let’s discuss in context of a concrete example. (e.g. you could enter your model in Models - OpenEndedness so I could look at it concretely. :slight_smile:

13:46:39 From Olaf Witkowski to Everyone : I’m unsure, but for innovation detection I guess there are some algorithms for coarse graining that could be used, or some approximations of statistical complexity, and some others mentioned in Susan’s talk?

13:46:51 From Olaf Witkowski to Everyone : (And maybe something autoencoder-based)

13:48:45 From Norman Packard to Everyone : re. current talk (Teruto Endo): coevolution of environment gives interesting issue for activity measurement (cf Channon), also interesting echo for Guttenberg comments on multi-scale nature of OE systems and ‘where are the drivers?’

13:50:34 From Steen Rasmussen to Everyone : @Norman, ok we’ll connect regarding the details

13:51:22 From Michael C to Everyone : I like the idea of keeping a pool of congestion control methods that are diverse and suited to different network dynamics, and perhaps training something that can then select different control algorithms on the fly

13:53:00 From margaretasegerstahl to Everyone : In considering biological OEE. I’d say that energy, environment(al stochasticity), reproduction and survival perhaps the 4 main things to bear in mind when thinking of the drivers of complexity evolution and OEE behaviour in any system.

13:54:42 From Olaf Witkowski to Everyone : @Margareta that’s interesting, could you unpack this (here or perhaps on Slack to have more space)? I think the point of view of a biologist can help

13:57:30 From Claus Aranha to Everyone : I wonder if they tried to put the robot in front of a mirror. :slight_smile:

13:58:28 From Luc Caspar to Everyone : They did put it in front of another alter if I remember correctly. Does that count?

13:58:43 From Norman Packard to Everyone : @claus: ask this question! (if Takashi doesn’t cover it)

13:59:09 From Olaf Witkowski to Everyone : @claus +1

13:59:28 From Hiroki Sayama to Everyone : This looks like the behavior of many researchers when no one else is around

13:59:39 From Norman Packard to Everyone : Looks like his memory is introspection of his state (not looking in mirror)?

14:00:13 From margaretasegerstahl to Everyone : @Olav, I Could try - in slack maybe. You asking me to summarise some of core insight gained from past few yers of theoretical work that finally hope to start publishing on this year… :wink:

14:00:23 From margaretasegerstahl to Everyone : …Olaf…

14:00:29 From Imy Khan to Everyone : @Margareta I would also like to unpack this further in Slack if possible - in particular, what are we calling “energy” and “survival” here (or what is the difference in biological systems?)

14:01:05 From Norman Packard to Everyone : @Hiroki :slight_smile:

14:01:16 From Olaf Witkowski to Everyone : @Margareta only if you want :slight_smile:

14:02:26 From Hiroki Sayama to Everyone : “The Imitation Game”

14:02:36 From margaretasegerstahl to Everyone : “Olaf. Sure! I think I can do it in terms of some biobasics to give an idea.

14:02:43 From Norman Packard to Everyone : maybe more interesting if “behavioral noise” is increased.

14:03:34 From margaretasegerstahl to Everyone : Re Alter: Could not be more impressed about its lifelikeness!

14:04:00 From Hiroki Sayama to Everyone : We need a control experiment with two postdocs!

14:04:08 From Olaf Witkowski to Everyone : I’m curious to know what the channel bottleneck looks like, compared to the amount of computation/information throughput on both ends?

14:04:42 From Tim Taylor to Everyone : I love that the Alter looks like a Centaur - half human, half horse

14:05:27 From Olaf Witkowski to Everyone : From this framing, it looks like they are both ends of a same robot :slight_smile:

14:06:33 From Norman Packard to Everyone : answer to @claus… mirror not so interesting.

14:07:54 From Olaf Witkowski to Everyone : re. Mirror: as I understand, there is a self-memory loop, which is something like a mirror, so they both have their own mirror alredy

14:08:39 From Michael C to Everyone : Related to David Ha’s world models perhaps?

14:08:45 From Roberto Gallotta to Everyone : @Olaf I believe the perfect synchronicity of the mirror enhances this effect

14:09:14 From Michael C to Everyone : Memories could be used to predict future motion as well

14:09:25 From margaretasegerstahl to Everyone : I was also thinking of similarity to what David Ha presented earlier today.

14:09:56 From Hiroki Sayama to Everyone : I think humans have order of magnitude larger behavioral data in their memory (coming from 20+ years of life experience). What if you also upload tons of prior behavioral data to Alter, perhaps his experience from past exhibitions and art performances?

14:10:11 From Olaf Witkowski to Everyone : @Michael that sure would be nice

14:10:47 From Hiroki Sayama to Everyone : Because Alter himself already has a highly accomplished life experience.

14:11:12 From Norman Packard to Everyone : @Michael Presumably the GANs have some temporal memory…(?)

14:13:20 From Olaf Witkowski to Everyone : This picture looks familiar, perhaps an internal memory in this conference :slight_smile:

14:14:15 From Michael C to Everyone : GANs would be a cool tool there. As humans don’t exactly plan motion perfectly, or try to recall an old movement and recreate it, we have long since compressed our learned movements

14:15:19 From Olaf Witkowski to Everyone : Sure, thanks Takashi!

14:15:29 From Olaf Witkowski to Everyone : Also, about puppetry: https://youtu.be/0hoK3RFvxwM?t=161

14:17:30 From margaretasegerstahl to Everyone : @Hiroki about human learning for 20+ … But different from humans, e.g. a horse is born almost ready - runs, but also communicates and acts and interacts and examines its environment the way that horses do, starting from, say, day 2 or 3 after birth - thinking how on earth it is possible simply blows my mind!

14:18:33 From Susan Stepney to Everyone : don’t disagree that bio is OE – just that the jury’s still out :slight_smile:

14:18:41 From Norman Packard to Everyone : @Olaf on puppetry: reminds me of Tokyo Alife entertainment…

14:20:09 From Norman Packard to Everyone : @Susan: Is the jury destined to be always out? For both bio and non-bio evolving systems?

14:20:52 From Hiroki Sayama to Everyone : @margareta perhaps because we are born quite prematurely and need to learn how to behave over 20+ (or 30+, 40+) years?

14:21:01 From Olaf Witkowski to Everyone : @Norman I guess it literally is…

14:21:06 From Michael C to Everyone : Is that what makes us more intelligent? Or capable of intelligence?

14:21:17 From Michael C to Everyone : That our brains have less pre-wiring at birth?

14:22:14 From Hiroki Sayama to Everyone : @michael not sure if more intelligent (because animals are also very intelligent for their own purposes). Maybe we are more adaptive and versatile (at the cost of many years of learning before becoming independent)

14:22:23 From David King to Everyone : @ Hiroki or did we sacrifice baked in knowledge for larger brains that could learn knowledge over time - based on human preference / ability to raise children for longer time periods than other animals. An evolutionary choice.

14:23:33 From Michael C to Everyone : A trade-off of neural plasticity, more adaptable means capable of more open-ended learning perhaps? Whereas less plasticity means fewer opportunities for discovery

14:23:34 From Olaf Witkowski to Everyone : @Hiroki as a society I think so (with the augmentation of all items of education, culture, and technology), but as an individual, arguably much less…

14:23:36 From Hiroki Sayama to Everyone : @david likely

14:25:09 From Hiroki Sayama to Everyone : @olaf but even at an individual level, babies are the most wondrous learning machines

14:25:41 From Hiroki Sayama to Everyone : They just can’t not stop absorbing

14:25:53 From Olaf Witkowski to Everyone : @Hiroki I can’t deny that:)

14:26:20 From David King to Everyone : @ Hiroki, I also agree that animals are also quite intelligent. It is quite a mystery how their intelligence is passed from generation to generation where, as Margareta wrote, horses can intuitively communicate, run, etc., within moments of birth… while us mere humans can only do rudimentary things in the same time scale.

14:27:17 From Olaf Witkowski to Everyone : I guess there are properties of adaptive intelligence (fitness to a given environment), and evolvable ones and ability to learn to learn

14:27:31 From margaretasegerstahl to Everyone : Re babies: …Or it is the clumsiness of the motor side especially re talking… From what a deaf parent told: his children have been able to talk by sign language already at age 5-6 months. Like say precisely what they want or feel (tired, wet nappy, want milk etc.)

14:27:56 From Susan Stepney to Everyone : precocial / prey animals need to do this in order not to get eaten too soon

14:28:28 From David King to Everyone : Survival!

14:30:17 From Tim Taylor to Everyone : @James very interesting talk - thank you!!

14:30:19 From Olaf Witkowski to Everyone : @James is there someone who has offered measures of retention in cumulative culture?

14:30:34 From Imy Khan to Everyone : Very interesting James!

14:31:08 From margaretasegerstahl to Everyone : Thank you James, great presentation. I will certainly look into this more!

14:31:35 From Alastair Channon to Everyone : @Olaf, your hand went up then down. Would you like to ask a question / make a point?

14:31:47 From margaretasegerstahl to Everyone : Isomorphism - good that this was mentioned!

14:32:31 From Olaf Witkowski to Everyone : Sure, I just thought we might be running out of time :slight_smile:

14:32:38 From Olaf Witkowski to Everyone : It’s a sort of minor point too

14:32:42 From Olaf Witkowski to Everyone : I can ask James later

14:33:28 From Stefano Tiso (RUG) to Everyone : @James what do you think about the parallelism between horizontal gene transfer and cultural transmission?

14:34:01 From Stefano Tiso (RUG) to Everyone : It also makes genetic evolution much messier

14:34:08 From Tanner Lund to Everyone : You can get cycles in cultural evolution (ex. metagame strategy). I wonder how often that comes out or how long it is tends to be sustained when it is observed. That wouldn’t be cumulative, at least on the short scale, as it repeats.

14:34:32 From Tanner Lund to Everyone : but of course, games have much lower dimensionality than the rest of life :slight_smile:

14:35:20 From margaretasegerstahl to Everyone : So we come back to Susan, how to measure qualitative aspects of systems and their organization - related how to look for whether there are isomorphisms maybe?

14:35:38 From Hiroki Sayama to Everyone : Thank you all

14:35:39 From Olaf Witkowski to Everyone : Thanks James, could you share the resources later?

14:35:40 From Stavros Anagnou to Everyone : Thank yu!

14:35:42 From Stavros Anagnou to Everyone : yu*

14:35:43 From Olaf Witkowski to Everyone : I’ll ask on Slack

14:35:45 From Stavros Anagnou to Everyone : you*

14:35:49 From Michael C to Everyone : Thanks all