 Hey everyone, thanks for tuning in and today it is March 30th, 2022. We're here in the Actinflab quarterly roundtable number one for 2022. Welcome to the Actinflab. We're a participatory online lab that is communicating, learning and practicing applied active inference. You can find us at some of the links here on this slide. This is a recorded and archived live stream. So please provide us with feedback so that we can improve our work. All backgrounds and perspectives are welcome here and we'll be following video etiquette for live streams. Today in the Actinflab quarterly roundtable number one for 2022, we're going to have a first section after our intros on lab scale updates. And then we'll talk about project scale updates which are nested within the lab scale and just close with some preferences and expectations and through and through. We hope that you'll see many affordances for participation, many ostentive cues, signals in your epistemic niche that elicit action, that make you curious about something, that connect with your pre-existing preferences or update them so that you get involved with active inference or possibly even the lab. So whenever you do see this or listen to it or some other combination, feel free to go to ActiveInference.org or email us at ActiveInference at gmail.com and we'll be in touch from there. So we can begin with just some introductions and warm-ups and it feels so different to do this without a pipe that we've read before. So for these intros we can just say hi and anyone can reflect on just like, how was your generative model updated this quarter? What did you perceive or learn? How did you change? And what might be something that you're looking forward to as we carry on the latter 75% of 2022? So I'm Daniel and a researcher in California. I think my generative model updated and fragmented into a thousand little pieces with CODA and the way that we used formal documents this quarter as opposed to previous quarters, but it's cohering in a way that is actually helpful, I hope. So I'll pass to Alex. Yes, hi everyone. I'm Alex Vatkin. I'm one of the founders of the lab and also researcher at Systems Management School and from the latest time I feel like we're sharing our generative model more and more and hopefully it will bring us to new developments. And I pass it to Blue. Hi everyone. I'm Blue. I am an independent research consultant in Lexifo. I think this quarter was pretty special, like the CODA restructuring was pretty awesome and just contributing to building like lab infrastructure in CODA and in like YouTube with like videos and timestamps and podcasts and I don't know, it was fun. I think probably the biggest update was the quantum paper because that was pretty epic. The one that we just have discussed over the last couple of weeks. And in the next quarter I'm looking forward to having some more quiet time, less talking, more doing. That's what I'm looking forward to in Q2 of 2022. And I will pass it to Ivan. Thanks, Blue. Yeah, that is very nice. I'm Ivan and I'm based in many now. And we finally are running a system thinking study group this quarter and I hope lab participants enjoy it and I see it that they do. And I hope we run next quarter another study group and it will be on conference. So that it. And I pass it to Dave. Yeah, a lot going on. I'm becoming a little less guilty about skipping at least half of the activities, but I get to hear at least to read the papers and listen to the live streams. And there's lots of work going on with the transcript, which is very gratifying. Even if Daniel no longer believes himself about how valuable it is to do the transcripts and put them in service. I still am sticking with the 10,000 times more value from a transcribed talk than just a YouTube audio with pictures. So who haven't we heard from Stan? Hello. Yeah, we're still here. We're still standing sort of. So that's great. I'm here in Toronto. For me, I'm appreciative for certain of the capacity the labs offered the connections, the conversations, the ability to bounce the ideas and that's been really, really helpful. And particularly with like Mark's soams and a lot of these ideas with quantum, which at first I thought some might be quite abstract. We're starting to synthesize and find ways to to pull them together. And so I'm hoping that can be used now. I'm working with some embodied workshops and taking that into communities and letting this working form how I coach, how I facilitate, how I try and explain. That's a hard bit, but I'll pass it back. And it could but shouldn't go without saying just massive appreciation to everybody who does and continues to show up for active lab. You see the tip of the iceberg on the live streams, but there are many others. So as we continue to formalize and determine and reduce our uncertainty about what kind of thing we are here. We'll hopefully have clear and clear ways to source attribution and to show recognition for all the people who contribute in just the million ways that are possible from beginner to very familiar. So we will just provide a brief overview of active lab from a narrative perspective before we jump into our strategy discussion. So in 2020, towards the middle and in the end, some of us wrote a paper in September was when it was published on active inference and behavior engineering for teams, which was addressing a gap that we perceived in the literature, which was the relationship between active inference framework and remote teams and communication. At the end of that year, after thinking about what our next steps would be after that paper, we thought that something like an active lab would be cool. So we sent out an initial call for collaboration during all of 2021 was our first year of active lab project operation. And the projects were in the three organizational units of education, communication and tools. This year in 2022, which is what we're going to be talking about today. A lot has happened at both the lab scale and the project scale. So we're going to first talk about the lab scale and the two pieces under lab scale updates are going to be the initial release of the lab strategy document. And we'll also discuss the initial advisory board cohort. And then we'll go into the project scale and discuss some of the ongoing projects that have resulted in all kinds of deliverables and outputs. So first to the lab scale. Alex, would you like to talk a little bit about just from a high level strategy? Yeah, thanks. I think as everything in our lab we are trying to do things in some new ways that was not used in before. And also trying to put it on some state of the art theories and approaches. So as for strategy, we consider strategy not as a concrete document, but as an ongoing process, which will have quarterly reviews and updates. And on high level we are following open-endedness approach. So for who is interested, it now can be put, can be approached from developments of Kenneth Stanley. And we are trying to avoid using classic objective missions, goals and other corporate stuff. So we are trying to see, in the focus of the future, we are trying to see some stepping stones that we can make next steps. And also for sure, from the beginning, we use active inference not only as a search field, but also trying to implement and use concepts from there, such as preferences and expectations. And also we consider that all our activities are on a different scale, so we need to follow and understand how it's applied on different scales. Thank you. So we will now go through the current state of the strategy document and just anyone here, feel free to raise your hands and of course anyone watching along, feel free to write questions and we'll get to them. So pause the video or check it out to read the full text here, but just to summarize our strategic document preamble. In Act in Flab, we're directing our attention to different phenomena across multiple scales. And the scales that we're most interested in or the ones that we are heuristically using are the personal, the individual human scale, the lab scale, that's Act in Flab scale. And then the community or the niche or the environment scale, which is the epistemic surroundings of the lab. And what's special about the lab scale is that is where the lab can implement policy. We can emit actions that are perceived as sense states by individual people or by the external community, and that may or may not change their policy selection. But the lab scale is where strategy happens because that's where we can actually implement policy selection. And we're taking this multi scale nested systems perspective to five different aspects of Act in Flab. And those five aspects that we're going to talk about in the next few slides are education, research, outreach and engagement, methods and open-endedness. Any general comments before we continue. Okay, so Act in Flab is an educational nonprofit. This plays out at all scales, as all things do. But we're especially interested here in the personal scale, in educating individuals, in having individuals that learn and update their generative model. And also respecting just the tremendous diversity of active inference journeys that different individuals are coming in and through relationship with the active inference lab. And at the more community scale, our educational efforts are oriented towards reducing research debt and increasing the applicability and the comprehensibility and the rigor and accessibility of active inference and related ideas. Any comments on education? The second aspect of the lab is the research aspect. And here it also plays out at the personal and at the community scale. At the personal scale, we hope to provide a niche that scaffolds and catalyzes the research projects by participants, which is something we'll come back to later. But we both do research in the lab, as well as on the interface and just hope to motivate and create a space for people to share their individual research updates. And at the community scale, we're recognizing that research is carried out alongside ongoing discussions about responsibility and ownership and many, many other terms. So there we hope to learn practices and improve and give our own twist on what the future of science, open science, decentralized science, DSI, just science with no adjectives and education. What will research and education be in our future? Steven? Thanks Daniel. Yeah, this is really helpful to read. And also the fact that we've got lots of people from different backgrounds coming together, we're able to... The research that's happening is quite transdisciplinary and it's been informed by lots of different ideas. So there's kind of a mixture of those processes. So it's helpful to have the systems engineering work that we'll probably come to later. But that's helpful in letting us get a feel for how the research doesn't get siloed as purely a paper. But okay, how could that be bounced off to maybe make a difference in the world in ways that maybe people who don't even know active influence find that they're using something differently because active influence has helped the way that their practice is shaped. So I like that. Awesome. The third aspect is outreach and engagement. And the active inference lab is working on a global scale with a participatory model. So at the personal scale, we welcome all participation and engagement. And that means all time zones and spoken languages and preferences and what people want to do and where they're at with active inference participation in a sense begins with where the participant is. And there's of course more that I hope everybody can say and add and show not tell or show and tell if they've brought enough for the class in the coming years. And then at the community scale, we also seek to be part of a emerging landscape of high quality science communication and participation and inclusion in research and projects around topics like active inference, which is what we're most attuned to in our regime of attention. But that doesn't mean we only talk to active inference people. So we hope that our outreach and engagement is meaningful and sustained and impactful. Any comments on outreach and engagement? Okay. Methods. The active inference lab applies state of the art methods across domains. That refers to the methodology of active inference and the emerging tools and ways that we have of working with active inference, as well as other methodologies like for project management for education for communication, etc. So at the personal scale, we're educating participants so that they can grow their expertise in different disciplines and also develop as a person holistically like keeping it balanced. You're not going to stay up and finish reading every single active papers. So staying balanced and having methodologies that help us be performant, but also balanced are quite important. And then the lab scale, we hope to implement the cutting edge, the Pareto optimal, the Bayes optimal methodology, which includes our switching strategies and we fix typos on the fly. That's what editable documents allow. Our affordances as an organization are the methodologies that we scaffold in projects. So, especially at the lab scale, we hope that we use methodology that's powerful and facilitates collective intelligence and also clarifies the distinction or the delineation the partition perhaps even between those who are following along and listening to live streams reading the papers in the game. That's awesome. And then those who actually on board into the lab and eventually have a role in a project will be able to increasingly be clear about what that means through the uses of different systems, ideas like practices, cases, responsibilities and concerns. Yes, Yvonne. Yeah, thank you. I would like to add that if we compile methods and outreach so the lab provides a lot of opportunities for the understanding of active inference and we provide tools and for those who already participate in the lab and for those who just start to deep inside the active inference. Awesome. And the last but not really the last aspect is open-endedness as Alex mentioned earlier and has been bringing and reminding us all of since the beginning that active lab is on the path of open-endedness. It doesn't mean that we sit and wait where the road forks or when we have uncertainty about the future. Those are fundamentals and in fact open-endedness is action amidst uncertainty. So at the personal scale, we recognize and support the open-endedness of the personal journey of individuals who are learning and applying active inference. So even if you don't know exactly where you're headed or what your finish line or your milestones, your checkpoint looks like, if you don't even know what your preferences are or you're unsure about some other parameter in your generative model, that's all good. At the lab scale, our strategy is to say it is a work in progress would be to almost assume that there would be a final strategy. But rather our strategy is an ongoing process that will always be updating in response to changes in the niche and the effect that our actions have over multiple time scales. So we do recognize and utilize approaches to achieve short-term success and realizing products and deliverables over short terms. But also there's a time and a space and a place for thinking about active inference and the active inference lab in deeper time. And that is especially where diverse participation and our lab advisory board come into play, which we'll talk about in a second. And then just lastly at the community scale, it's true in principle that active lab can participate in various kinds of activities that might be beyond education, research and outreach and engagement. We will always be considering those opportunities in a case by case way and assessing whether it's something that fits best within the Markov blanket and the kind of thing that active lab is or maybe some thing else could exist that could play that role in the ecosystem. Yes, Dave. Okay, Stephen. I'm interested by that way that the ecosystem, this question of what's in the Markov blanket and the non-equilibrium steady state and what's in the niche. And we, those three things are kind of dynamic. I think over time, looking at the way the world is, and without getting too evangelical, there's a lots of big, you know, there's big claims made about active inference and its potential. And as we've been having conversations, we've seen a lot of things at scale come up, which is quite exciting conversations around that, but often maybe a little bit fuzzy, a little bit conceptual. So I think some of the stuff that's going to come out of there, the lab's going to be spinning off as much as it might be holding on to. Excellent. Thank you. So that concludes the discussion on the strategy update. The second lab scale update is to mention our advisory board. So this information can be found at our website, activeinference.org, and on the top right, head over to where it says advisory board. The advisory board is a insight and guidance board. It's not a managerial board, and it exists to help the lab make better strategic decisions. So we engage our advisory board participants in terms of their expertise, different recommendations and questions they have for the lab. And we recognize and hope to draw on and synergize with their experience in many, many different areas. We have an image here from Kirchhoff et al. 2018 about nested systems and about the way that as systems develop, for example, as potentially a overdevelopmental time, a zygote goes from being a single cell enclosed by a lipid bilayer towards increasingly ramified levels of embryonic complexity or similar transitions could be happening over evolutionary time. Something like that is occurring as we elaborate and build the active lab. We have an awesome initial cohort of the advisory board, and it includes Bradley Alicia, John Boyk, Matt Brown, John Clippinger, Scott David, Jeff Emmett, Chris Fields, Carl Friston, Raphael Kaufman, Anatoly Levenchuk, Rosalyn Moran, Alba Serrano, Cheryl van Hoof, Tim Verbellen, Swan Webb, and Michael Zargum. And we have already received really important and meaningful feedback from them and really appreciate it. And for those who are interested in the kinds of work that we're up to and might have experience in a certain domain that could be helpful, for example, related to educational nonprofits, then perhaps get in touch with us. But this is a major change to the lab scale to introduce an advisory board, and we're off to a fun start with them so far. So we look forward to just seeing how that continues. Any other comments on ACTIMF at the lab scale? Or we can head to the projects. Okay, so as we head through these projects, feel free to raise your hand at any point if you want to add a comment or a question. And as a participatory lab, those who are listening and you've made it this far, recognize that all of these reflect the contributions of participants and reflect affordances for you to contribute further. Even if only to reduce your uncertainty about the documentation or the way that something is occurring. So hopefully if you're listening to this string of symbols, know that you can participate and perhaps you even should. We're going to talk about a few different projects today. In fact, 11 difference of them. And you can see loosely on the left that these are in three organizational units, EDU, COMs and tools. But the project scale is really where the team and the roles and the practices occur. So we're highlighting the project scale. And the project is also the unit of granularity where you can really get involved and make a contribution. So showing up for our synchronous organizational unit meetings is an awesome way to kind of paint a first coat and to be exposed to many of the projects. But by and large, the contributions happen at the project scale and asynchronously. So all backgrounds, all time zones, all levels of commitment, everyone is welcome to get involved with these projects, which we're going to go through now. They're going to be the active inference ontology, the active inference course, internal educational initiatives, like the systems thinking, reading discussion group and ontology working group, active inference journal, research updates, live streams, podcasts, social media, active blockference, robotics and embodied and active lab meta modeling. On we go to the ontology. We've just recently pushed an update to the active inference ontology. And it's available at this Coda link here. The goal of the active inference ontology is to increase the accessibility and impact of active inference by increasing clarity and interoperability and expressivity and all these other good things about terms and ideas. And when you go to this site, you'll find that there's a explanatory preamble and what this ontology release contains primarily are two different kinds of things. Proposed definitions and translations. So first the proposed definitions. This is what the proposed definitions look like for the core terms. The terms are on the left column here. And then everywhere there's an at and a blue is like a clickable link to another term. So it's kind of like how dictionaries use other words in that language. And it will say a is kind of like B, except it's different than C. And so that's what the definitions are. They're natural language answers to the question, what is X? And so these proposed definitions are currently in natural language, but there are other ways to develop them more formally. The proposed definitions are there for those who just say, well, what does this term mean? It also helps with computationally aided sense making. And then the second piece that is in this current release of the ontology is the translations. And so here's the current view of the top of the translations table. So we have the term in the English on the left side. And then we have so far complete translations of the core ontology terms into Spanish, Portuguese, French, Italian, Czech, Russian and Hebrew. And some German and some others that are sort of beginning. So especially if you have fluency or friends with fluency in some of these languages, but especially new and different languages. So this is a really awesome opportunity to translate some terms, which are mostly normal natural language terms into another human natural language. And that will facilitate the ability to, for example, deploy educational information or transcripts in their translations. Do it in a way that's rigorous and accessible and meets people where they are in terms of their language fluency. Yes, Dave. Okay. So that's the active ontology project where we're developing participatory ontology for active inference terms. Stephen. I was just asked, not putting you on the spot, but if I was to ask you, like, what areas have been the most challenging or the most kind of like most most difficult to overcome? And what are the sort of questions that are still being grappled with on the ontology, you know, and which ones have we overcome already and which ones are still sitting there in the background? I was curious if you've got any thoughts on that. Yeah, good question. I'll give it thought and then anyone else is welcome to if they want. One interesting thing is these definitions, although many papers do sometimes define their terms, either in the text or in a box or in a glossary. That doesn't mean that the way that the terms are defined are coherent or compatible or consistent. Sometimes it's like a yes and someone defines it one way and someone else just kind of takes a different spin on it. But other times we have come across several situations all together, all of us at least here where there were incompatible seeming uses of terms. And perhaps that happens with all kinds of linguistic elements. So that was one interesting thing was that it's not just like we're looking through papers and copying sentences. There's actually a really powerful process of the discussion that we had focusing week by week on just a few terms or one term where we were seeing all the different ways that different terms are being used inside and outside of active inference and just trying to work towards some kind of grip on that blue. So one of the biggest challenges that I think we face is dirty and incomplete data, which is like pervasive across fields of disciplines everywhere. So I mean trying to be the bridge to fill the gaps to make the data to complete the data. And I mean there's always opportunities that people want to get involved in filling out the data, helping us to have a complete picture of which to metamodel. Awesome. And there are other columns and features of the ontology that we're working on in the lab for anybody who wants to get involved. Like correct examples, just simple declarative, concise, positive examples of terms being used correctly. That's awesome. Incorrect examples. People who are using sentences, it doesn't have to be caught in the wild, but it could be. But just generating sentences that are like counter examples can be very helpful for both humans and computers in their learning. So we have examples, counter examples, formalisms, connecting to core terms and other terms to equations and papers and connecting it with the literature corpus. So there's so many angles. Those with curiosity and expertise in active inference and or ontology should absolutely get involved with this, which will be a long term project of the lab. And we hope to find new ways of deploying it. And just to give one little contextualizing slide here. This entire continuum can be considered to be different types or grades or like a taxonomy of ontologies. And so ontologies in their most informal can be as simple as a list of terms or glossaries, the sources or what we have right now with the current release, which is like a principled informal hierarchy and some relationships amongst terms. And that's what we have right now with the proposed definitions and translations. We're also working with the ontology working group, which we'll talk about in a few minutes on formalizing that ontology. For example, using sumo and suo kiff and thinking about how do we integrate the ontology with course development with knowledge engineering and literature integration. So some of the many things that people can look forward to. Okay, on to the second project, which is the active course. So in the active course project, we are seeking to create a learning platform that embodies the principles of active inference and facilitates the learning of its principles and concepts to learners of many different backgrounds. And this is something that we are not too far down the road on, but are eagerly looking forward to, especially in light of some recent epistemic changes in our niche, which we'll talk about soon. But there's a massive potential for active inference education. And we hope to have course offerings and educational experiences that meet that opportunity. Alex. Yeah, just a quick note to avoid possible misunderstanding that notion platform here is not like for some IT platform, but more about set of educational services, which will be supported to improve learning affordances. Yes, good call. Thanks for clarifying that. So if you're interested or curious about course design, education, education in a participatory way, nonprofits, those are all excellent cues for people who would like to come on board and help in a serious way for this course as we start to think about what it can be. Okay, third project, we have several internal educational initiatives in active lab. And so on the left, we see the systems thinking, reading discussion group, and on the right ontology working group, the OWG. Alex, or anyone would you like to describe or give a note on either of these sides? I just want to say that from very beginning, we consider that we should not only learn and educate active inference, but also to learn, as you mentioned before, state of the art practices. And system thinking was one of the main discipline from the beginning, what we think is very useful for collective work, especially in remote mode. So it's great that now more and more people start learning and it definitely will improve our communication and speed up our activities of the projects. Thank you. Yes. Yvonne has been facilitating the systems thinking, reading discussion group, STRDG, and we have an intrepid group of participants who are working their way through this online course and having regular discussions and asynchronous conversations as well around it. And as Alex said, this is really important for the ways that we work together and communicate. And on the right side, we have the ontology working group. So previously we were discussing the active ontology as a project where we're talking about active inference terms and how they're related in the ontology working group. We are increasing our competency in computational ontology. And what that has looked like primarily has been working with the sumo suggested upper merged ontology system in Sue Kiff and working through the textbook of Adam piece. And this has been an awesome experience. Dave has made a massive amount of contributions and ontological limericks of different formats. And that has also been a great learning experience that hopefully will fold back in towards improving the active inference ontology, but it's also just a learning initiative. And that brings us to a ongoing learning initiative, which is the release, literally in the last several days of the textbook by Thomas Parr Giovanni Pazzullo and Carl Friston. It's the active inference textbook. And we asked, well, what can the active lab do for slash with around the textbook? Do you want to join a textbook group, not just a reading group, just a textbook group, because we are going to be organizing cohorts of textbook groups. And you can email us for the form if you don't already have access to it. But you will fill out a form that asks whether you'd like to join an upcoming textbook group cohort, and it will be done in groups and cohorts. Yes, you're ready or no, you'd like to perhaps hear about that affordance in a few months. We ask for name, email address and country, whether you have a textbook or not, and then just for you to provide several sentences or more on your level of familiarity and your interest in active inference. And all backgrounds and familiarities, as with all activities are welcome. This is going to be a really awesome experience, I think to work through this textbook and see it as just one pebble on the ant hill the tip of the iceberg or any other metaphor that people use for the way in which this textbook will both be a visible and centralizing artifact, as well as something that is really only a few hundred pages long and so relatively small physical book. So it's just the beginning. It's not the end of active inference and those in the textbook groups will realize that and enact it. Any comments on internal educational initiatives? All right. Number four, the active inference journal. So in this project, we are publishing knowledge artifacts so that they are searchable, indexable and sightable. Most recently in the active inference journal, we uploaded a second improved version of the transcript of the Carl Friston symposium applied active inference symposium that was in June 2021. And from that, we are improving our methodology that we learned and learned by doing in that first symposium transcript process and applying that to other event recordings. For example, guest stream 16 with Mark Solms and live stream number 40 with a quantum free energy principle. So the journal is going to be a great place where we continue to build a participatory pipeline for improving spoken and written dissemination. Anything that anyone wants to add on this? Yes, Dave. Yeah, I sent the original paper and the three video segments to Sean Carroll along with a summary of the three big points in the summary. He says, this is cool. I will check it out. Which I'm sure incentivizes the gentleman who volunteered to go through our transcript and make it more presentable to get through that. And then we can send that to the very busy Dr. Carroll who just has a new appointment. He is something like the the chair of the philosophy of science and Johns Hopkins. So all his pain of being denied tenure is paying off. And he may pay off. He already I think gets along very well with Professor Kristen. He he says I have never had to work so hard in an interview with somebody who is so knowledgeable and well spoken and anxious to communicate. Awesome. Yeah. And it's a great example of a project that increases accessibility, for example, by taking two or six hours of YouTube discussions and making them searchable and translatable. And then also reduces researchers research debt because we can have somebody spend three or seven hours improving and curating a manuscript and finding. Oh, when somebody mentioned this paper, here's that specific citation that multiplied by thousands of people searching for that paper. Maybe they have access. Maybe they don't. Maybe they misheard the name. There's so much that we can be proactive about in terms of making our knowledge corpus really accessible and meaningful and woven according to modern practices. Okay. Awesome work though, Dave and others on the journal Maria and others as well. Okay. On five, we have research updates. So this is not a project that we actively steward per se in the lab, but this is just capturing research that's in progress and finished by lab participants. So just to highlight a few papers and products that have been released by active inference lab participants, but not necessarily as an active lab project in the last few months. So by Avell, Gwen and Carlou, a.k.a. Serval, a.k.a. Icarus, the paper cognitive agency in sociocultural evolution. And then a awesome collaborative paper that we wrote with myself, Sean, Arhan, RJ, Shadi, Avell, Lou, Yvonne, Sid, Amit, Jacob, Caleb and Alex, where we talked about decentralized science, DSI, we integrated some of the more technological perspectives on DSI, for example, related to Web 3 and blockchain, and juxtaposed that with some of the bottom up the view from the bottom participatory sense making and epistemic commons and produced an active entity ontology for science, and I think that's the chaos that we hope will allow us to have a way to model some of these complex epistemic ecosystems in the coming months. So that's just some of the research and there's other awesome research that's in progress. Okay. On to, yes, Stephen, go for it. I think this work with situated sense making, the decentralized science is really, really interesting. And I think there'll be ways that the other areas of sense making, which this can be helpful to bridge that active influence generative modeling approach with other ways. I'm really glad that you did this work, actually, as a group. And I think it's going to help tell another story, which might have been a bit too hard to tell from the ground up, but now you've got a bridge. So thanks. Cool. Okay, onto six. So we had awesome live streams as usual in this quarter at the Coda.io slash active inference lab active inf lab, you can find the past and upcoming live streams. We had several guest streams and other kinds of streams, but most of our streams in the last few months were the paper discussions. So in live stream number 35 dot zero one and two. So as with all of these, there was a dot zero background and contextualizing video that was prepared in the weeks leading up to the dot one dot two discussions which are participatory group discussions. So if you ever see a paper where you want to be in the dot one or the dot two and be part of a group and you don't need to say a ton, no one's going to call on you. You can ask anything you want, provide any perspectives, always welcome to join that way. And then something that's incredibly helpful for the lab is to commit assuredly ahead of time and contribute to the dot zero making the slides and helping contextualize the paper ahead of the dot one and the dot two. So we in the last few months have talked about in number 35. We talked about a tale of two architectures free energy, its models and modularity. Check out the live streams to learn more. I won't even attempt to summarize the papers other than just to state their titles in 36 we discussed modeling ourselves with the free energy principle reveals about our implicit notions of representation. We then dove a bit into the technical in 37 with free energy, a user's guide. In 38 discussed the evolution of brain architectures for predictive coding and active inference. In 39 discussed morphogenesis as Bayesian inference, a variational approach to pattern formation and control in complex biological systems. And then in 40, we discussed a free energy principle for generic quantum systems, which was as anybody who was there before or during or after it happened can attest was quite a special and interesting experience. And just to show what that code page looks like if you head over there, you'll see the upcoming streams. Here we are in the end of March. It's not in the live stream section. It's in the round table section. Here we are right now. But in live streams, you'll see the papers that we're going to be discussing in the coming months. So if you're interested in any of these upcoming papers, please get involved in the dot zero or the dot one and two. If you're interested in inviting anybody or yourself being a guest on a guest stream, that's always an affordance. These are like one off or small series that can be on active inference research or related topics. And we've had some great discussions in the last few weeks on guest streams. We have model streams and math streams that are for presenting some of the formal and model based components of active inference and org streams where we talk about organizational principles and organizational design. So those are just some of the live streams that we've had and it's been fun. So especially if you want to give a guest stream or if you can connect to somebody who does want to or encourage or recommend somebody or you want to be a facilitator for someone else to present or any other combination of affordances. Let's make it happen because we can get a lot of awesome modifications in our niche. I think what they used to call making content back in the web two days. So that's all for live streams podcast. Blue, do you want to give an update? Sure. So we are moving through the podcast really revisits a lot of the early live streams. So like if you're playing catch up, like if you caught a live stream, you're like, oh, I should go through the back catalog. Like maybe now you're on live stream 20 something like the podcast starts with the early episodes and I think we're through episode nine live stream number nine now. So yeah, check it out. It's fun. Also a great chance to get involved if you have like AV expertise would love to get some pointers also like I know some stuff. I've been clipping the podcast for a little bit. So if you'd like to learn like the basic things that I know and want to get involved in making the podcast reach out. I could always use help if you want to listen to more podcasts. Come help me make them Dave. Yeah, if you have listed the podcast after nine that you're planning to do. I'd like to get that because I want to get started with just doing basic text conversion on this. If that assumes that this these are the prior. Is that correct? These are the ones that you want to get in front of me. I don't know what the priority is there Dave but I will say like there's definitely value to having a transcript. I actually looked at pod bean who is our podcast platform the other day, and they will transcribe our podcast for us actually at 20 cents a minute. So if you really want to know what the value of transcription is just just the value to just put it into text now I don't know if that comes with with periods or punctuation or I don't know what that output looks like. But there's definitely value to having a transcript there. I'm not sure that these are necessarily the priority but it's just going through the back catalog it's helping me catch up. So I just kind of started where I wanted to start at the beginning when it started the beginning right. I don't know what what I love about these podcast is they're short. They're like from four to maybe 15 or 10 minutes long. And the first time that a speaker speaks, they introduce themselves. And there have been some probable clippings and rearrangements which are very subtle that help that listenable piece be like coherent like it's like an extended thought even one that's distributed across multiple speakers so like you hit play listen to it. And then you can think and reflect because sometimes when it's two hours on a paper, it feels like it's like a rail car coming through with it's like every single idea and thing and people say that they pause the live streams and there's so much and then also as the priorities. That's something we can figure out and discuss, but won't it be amazing when the live stream is, you know, the front car of that train piercing into the future. Hashtag doing it live. And then there's transcripts for all the live streams, and that can be on one hand connected to the ontology of transcript of the live stream itself could exist. There could be some clipping and some transcripts for the podcast so as we develop our informational ecology. All of these kinds of connections and tasks are in play it's just a question of how late can you stay up, which I've wondered about for a long time. David's no sleep. You know, the other cool thing about the podcast is that it's slide free. So where when you're on the YouTube watching the live stream, often it's associated with some visual image and sometimes there will be like a description of the image in the podcast but it's nothing that you need to like pull over and look at, which is sometimes if you're trying to like, watch the YouTube videos on the long drive, it's like, oh, that might not be safe, but the podcast are definitely, you know, commute friendly, because you don't have to see anything you can just listen. So that's another it just offloads that the visual input as well. Good for the treadmill stuff like that. Great point. Okay, on to eight. So some updates on our. I'll just call it social media. We have our Twitter with 1350 nine followers. We have about, I think we crossed 300 members in the discord. Over 770 subscribers on YouTube and some several hundreds of people reached in Facebook and anyone who is interested in automating social media that would be an excellent affordance blue. Yeah, definitely I was just you took the words right out of my mouth and Facebook is very new so we have like just 40 likes on our Facebook page so if you're listening to this and you use Facebook regularly for whatever I mean it will ping you to remind you about podcasts and sometimes live streams sometimes I say because like I said we need an automator so that's your expertise and you want to always be reminded about the live streams get in touch. Yes, just to be able to preload a buffer. Here in the front car of the train. It's sometimes hard to clean up the pieces. And so if you're feeling like that's something that's interesting to you then there's many opportunities to have engagements on different platforms and just feel free to just engage in the comments and in the threads however make sense. Great enough with social media. Okay, on to nine. So now we're heading more towards the tools part of this discussion. Primarily the work of Yacob Smithle has been just excellent on active block fronts, which is a package that we're developing in order to facilitate cognitive agent modeling. And if you look up active block fronts on GitHub you'll find the open source repository. What we've done with active block fronts is implement active inference entities, the kind that you will see graphically described in papers or the kind that have been implemented in Python via pyMDP. And we built upon pyMDP as well as a complex systems modeling framework called CAD CAD. And by integrating the active kernel of the entity action perception loop with the complex systems modeling framework. We hope to have engineering grade approaches towards modeling complex cognitive ecosystems. So if you are interested in Python or Julia or other aspects that are computational related to active inference and you like complexity and complex adaptive systems and everything else that you might see on this slide. Active block fronts is an open source project so go ahead and fork it, make your contributions and come get in touch with us if you'd like to be a little bit more in loop and really looking forward to seeing how this project develops in the coming years. In.tools also we've been having some awesome discussions around robotics and embodied cognitive entities. So we have been characterizing kind of like a literature review on different cognitive agent modeling frameworks. So active inference is often proposed as a unified cognitive entity model. And so we've been just curating many other cognitive agent modeling frameworks from the past and present. And then assessing them in terms of their language, what cognitive features they implement, whether they deal with embodiment, what is their mathematical underpinnings, and so on. And meanwhile characterizing some of those cognitive features and elements like anticipation, memory, counterfactuals, abductive reasoning, different kinds of cognitive features or motifs and then trying to understand how do these different cognitive model frameworks line up on the rubric of wanting to have expressivity across all of these different cognitive features and to integrate them. And that's a little bit more on the symbolic cognitive. We've also been thinking a little bit about the embodied, whether the embodied and spatial and bodily or the embodied and robotics. So here is a screenshot from the work of JF Cloutier, who's been an awesome participant in this past quarter, and some active robots that are Carl and Andy. I think people can probably guess what their last names are. And there isn't too much more to add here other than check out this woefully under acknowledged channel and work. And if you're interested in any of these areas like related to cognitive modeling and robotics and embodied cognitive systems, those are the kinds of things that will be awesome to scaffold in the lab. Okay. And to our final of the 11 projects, Acton Flab Metamodeling. So Acton Flab Metamodeling, aka Alma. So there's an A and an L and there's an M and there's an A. And I think that was Jessica's suggestion because Alma has nice connotations as well. This project is using metamodeling tools like COTA to provide the capacity for lab scale generative modeling updates and reconfiguring. So the ability for the lab to reflect and to modify itself and its niche in order to design the kind of participatory niche that facilitates successful policy selection by lab members. Like, what does the lab need to provide the individual as the action state of the lab sense state of the individual in terms of the communication such that that participant is chill, but also making the right decisions that are productive. And we're using primarily COTA. That's why the font is bigger here. But we're also implementing the active entity ontology for science that we discussed on the research slide and active blockference just two slides ago as a more formal way to get at some of these metamodeling. And this is a figure from our 2020 paper that first early paper on systems engineering and active inference and remote teams where we also looked at the informational niche, the shared informational niche of the team. This is their shared folder, their shared COTA document, their formal documents, their drive, the chat threads. That's the shared epistemic niche where actions to modify the niche by one participant can then be like Stigmergy modifying the environment so that themselves later or other participants then or in the future take different action selection as mediated by the way that that sensory state updates their generative model. And then we have narrative. And so we've been very interested for a long time in rhetorical and narrative approaches to cognition. And so now we're just beginning to see how some of these approaches can come into play in the context of active inference and in the lab. So that's like the meta on active lab. And this is also an awesome project set of affordances. If you like COTA, if you liked Douglas Hofstetter's Gertle Escherbach, if you like act in first serve, all of these would be great diagnostic cues that perhaps you would like active lab metamodeling. Yes, Stephen. Yeah, thanks for sharing. I was wondering how this also might even be able to tie in the future to micro phenomenological kind of modeling that kind of approach. So we've got the meta, the narrative story, and hopefully maybe some sort of micro phenomenological approaches as well. But I think we've got to walk, but we'll get there. Thanks. When you said micro phenomenological, I thought, oh, like the experiences of children. Yeah. Say hello. Cool. So that is it for the project updates. They're all ongoing. Let's just kind of go over some next steps at the lab scale. Several areas that were especially looking for assistance or contributions on would be on the legal side. For example, potentially connecting with a firm that would be interested in providing pro bono assistance who has experience with education, nonprofits, intellectual property, tech, etc. And also helping our lab navigate a course towards financial sustainability, whether through grants or other mechanisms, but just finding a way that we can reward and stabilize the contributions of participants. And then at the project scale, any and all participants who want to be in the strange loop in the strange attractor want to be on the path towards contributing to active inference lab as they also grow and develop as an active inferencer and as a person. So if any of these projects or some other opportunity that you see and want to take the initiative on any of these reflect affordances for you. So hopefully if you're listening this far in a minute and six hours in or wherever it's clipped out in a podcast, then just recognize that these are all reflecting the past contributions of participants. Like yourself. And this is just an opening. It's only the first quarter of 2022. We're just beginning. And so there's so much in the field and in the lab to be done that there will be something to do that's meaningful and chill and aligned with your preferences and expectations and availability. Just believe it. Okay. So that takes us close to the end. And a little modification of our common ending slide. Where do we go from here? What are people still curious about learning and applying? And what are we going to do differently maybe in the next quarter? Yes, blue. So I think I already put it out there at the beginning. But I'm really like interested in not talking about things anymore, spending less time talking about things and more time actually doing things. So that that's what I'm going to do differently in the next quarter. And that's me as like on a personal scale has nothing to do really with the lab, but it's I just want to just work now. I just want to put my head in the books and get to it. Awesome. Yeah, sometimes like it helps to get caught up through conversation and just being in the synchronous meetings. But one hour per week is not how we build a course or not how we go through a textbook. And so that is a very important balance to keep in mind. Steven or Dave, any other thoughts? Yes. Or no. There's a lot of good work going on. I'd say just keep doing it. And talk about, you know, talk about what's working well, what needs to be improved, what bottlenecks are. And mention some bottlenecks and sometimes maybe some of the fun projects that could be pushed back in favor of some of the more difficult things that actually will end up being leveraged heavily. I think it'd be ontology. It may not be fun to slog through the stuff that it's going to really pay off when it gets some transitive closure aspects. Awesome. For those who are studying or curious about active inference but feel kind of like on their own, whether no one else in their local niche or network or lab or university or organization is working on active inference. That's like one of the most powerful things is just knowing like, if you have an idea that you're not even going to pursue, but just somebody should do this or it would be cool if this existed. Having a space where you can add that pebble and maybe down the road or maybe sooner than you expect somebody else seeing that pebble is stimulated to act. And so that's what can happen when we have shared ways of working. And that's just what we're beginning to see. Steven. Yeah, sort of bouncing off that as well. I think I keep thinking about stages of readiness, which I am. I'm normally working with community participants who are marginalized, but I kind of, I've had that myself as like feeling a bit marginalized feeling and my. Like ready and it was like a little jump, but it can be a chasm, you know, trying to get to that stage and I think that can be for a lot of people. You don't quite know on different things, whether you're near that threshold or not. And you're often a lot closer to being ready and you maybe just need someone to give you a kick up the bum and say, let's go for it. You know, but having said that, that's when you appreciate the quality of getting getting some things under your belt on the way to the line. And I think, like Blue said, like it's time to act and for me personally, yes, like the stage of readiness I'm at is very different to where I was a year ago, for sure. Both in my knowledge when I look back to where I was, but also in my heart. And that's a that's not an insignificant piece of the equation. So thanks. Awesome. I'll just wait one more minute if anyone who's watching live has any questions that they would like to see addressed, but totally agree. It's, um, it's a multi year journey. I mean, we're all on a lifelong journey of some sort, but hopefully even our active inference journey is multi year. And so being able to connect our policy selection, our behavioral decisions in the next micro saccade of the eyes, the next place your eye is going to look the next action you're going to take in terms of opening up a hyperlink or reaching for a book or a pencil, but being able to connect those very rapid actions with a bigger picture and with a longer temporal timescale and acting amidst uncertainty. That's what active inference is about one aspect of it. And so if anything that you've heard resonates, just hopefully see this round table presentation as like an invitation less than a stand and deliver. Because there's so much openness in the area and in the lab and it maybe can feel like one is starting late, but that's just human knowledge development. It's a treadmill that's been going. And through coming together in the lab, we can actually leverage each other's contributions and catch up on fields that have thousands of hours of reading and hopefully get somewhere cool together. So, um, Dave, blue, Stephen and Yvonne and Alex from earlier. And all the other lab participants, big thanks for this amazing quarter. And it's been a fast one. And onto the next one. Okay. Bye. Thanks all have fun.