 Hello and welcome, it's February 28th, 2023. It's active guest stream number 37.1 and we're here with Boba Brinkman. This should be a great discussion and emergence. Boba, thank you so much for joining. Let's just jump right into it. What brings you to this day? How and who are you? Thanks, thanks Daniel so much for having me here. It's a real honor to join you and to dip into this community which I've been aware of for a while and like to try to follow the work as much as I can but I'm not a scientist myself. I'm a rapper. I have been exploring academic topics in hip hop for quite a while. I did my master's thesis on Chaucer's Canterbury Tales. I was a Renaissance and medieval scholar and I rewrote the Canterbury Tales into rap and I used to tour around the UK rapping my wife of Bath's Tale and Miller's Tale renditions which was a lot of fun and then at one point it was around 2007 or 2008 an evolutionary biologist, a geneticist saw my Canterbury Tales show and he reached out to me and said, I love what you're doing for medieval literature, popularizing it with rap but there's other topics that need rap popularization. I think his exact words were would you be willing to do for Darwin what you did for Chaucer and this was in the run up to the Darwin Bicentennial in 2009 it was the 200th anniversary of the birth of Charles Darwin also the 150th anniversary of the publication of the Origin of Species. I was of course a big fan of Darwin and of evolutionary theory but not very expert. So I said to this scientist, yes I'll write some raps about evolution but I could use some guidance. He sent me a whole bunch of suggested reading sources, books, peer reviewed papers, everything so I took a sort of six month deep dive on the state of evolutionary theory absorbed a whole bunch of books and wrote these rhymes but his criteria was your performances will be at academic seminars with experts in Darwinian evolution across all kinds of social domains and psychology and everything so you need to make sure you represent the theory accurately which means you have to run your lyrics by me so I can check them for scientific veracity which I thought was a fair criteria for that kind of context so I sent him my lyrics he gave me some revision suggestions and then when I got up when I was set to perform the premiere of my Darwin show in February on Darwin Day, February 12th, 2009 the scientist his name is Mark Palin he was a bacterial geneticist he had this great intro typical British tongue and cheek style he was like, I have fact checked the lyrics and you're about to witness the first rap show in history that's peer reviewed I was like peer reviewed rap that's a great idea totally marks coinage but then right away the wheel started spinning like what other topics could you apply the peer reviewed rap lens to so I did my evolution rap show for several years it was called the Rap Guide to Evolution it played off Broadway won an award at the Edinburgh Fringe Festival and then other scientists started tugging me on the sleeve like, oh the Darwin raps are cool but could you do wilderness conservation and threatened species ecology could you do climate change could you do neuroscience and various great collaborations came from that so now I have a whole series of rap albums rap guides to various topics and my encounter with the active inference community and research base was when I was working on one called the Rap Guide to Consciousness and Professor Anil Seth is one of the I think main proponents and contributors to the Bayesian brain active inference research paradigm today and he was one of the main science advisors on the project so he was giving me Bayesian stuff to read including Andy Clark's What's Next highly technical review of all the literature which I slogged my way through as a non expert and ended up working Bayesian ideas quite in a detailed way into the Rap Guide to Consciousness show and what really resonated for me honestly I had never really paid attention to Bayes theorem or Bayesian reasoning at all but it took me a while to get my head around it but then I was like okay this is actually just a great paraphrase of the scientific method and I was already a massive proponent of the scientific method and I was geeking out on the degree to which the brain may be a Bayesian engine at the microscopic level producing a rational inquiry process that replicates the exact same Bayesian algorithm at the level of interactions with society so I did one song in the show called Good Bayesian let me show you how to be a good Bayesian change your predictions after taking information and if you're thinking I'll be less than amazing let's adjust those expectations and I was right away thinking about like how is a rap performance itself an instantiation of a Bayesian algorithm with the audience's expectations being revised in real time based on the rapper's performance and the rapper's expectations when creating the script of the performance about what reaction they'll get from the audience which is a prior belief about the effectiveness of the performance but when it's only a script don't actually know what's gonna work and then you do the performance you see what gets the reaction you revise your priors and you also revise your script and that was a concept that I had already been rapping about in my evolution show I have a show, a whole song and that called Performance Feedback Revision which instantiates the kind of evolutionary natural selection algorithm but at the level of creative writing and performance you write the script, you do the show you see how people react you go back and change the script and that's kind of like the genome being revised under phenotypic variants and how they end up surviving or not surviving, reproducing or not reproducing so this is, I just started out rapping and trying to master the craft and each scientific topic I get into I start going, oh, this actually helps me understand what's going on in rap much better and especially actually one of the examples I play within the show which is a lot of fun is like call and response expectations of the audience like a perfect kind of demonstration of a Bayesian moment when you go, double take, look twice calculate the odds, roll the and when I do that the whole crowd will be like dice but the expectation of what the rhyme couplets gonna be is a game rappers play all the time like finish the bar, right and that is entirely around a setup of an expectation and the fulfillment of it but then sometimes you can play with it and be like, yo, perceptual basics Asmorphias, we live in the and I usually get the crowd being like matrix back at me, but then other times I'd make it more difficult like, yo, perceptual basics ladybugs like to eat and then like maybe one person out of a hundred would be like aphids and I'd be like, yeah that was a person that had the correct scientific Bayesian priors to conclude my matrix basics aphids rhyme pattern but usually only a few get it and then everyone else is like ah, there's the nerd over there or whatever so, you know, it's become like actually informed my creative process in terms of how I try to set up and violate expectations in real time psychologically with an audience when I'm performing and I will be honest that my close reading of the research was several years ago at this point the rap guide to consciousness came out in 2017 at 18, I was performing it off Broadway 19, I was touring it around theaters in the UK and still performing it off Broadway and then 20, of course all performances ceased altogether and I had to think about what rappers do in pandemics when they can't get on stages and see crowds so, yeah, we can talk during this interview about my new project which is a custom rap agency you can see me wrapping it right here event rap, which could be fun to talk about in terms of the event inference elements of active inference and Bayesian analysis could be you know, I think some of the aspects of what attract me to events what attracted me to events during the pandemics and still does as a sort of catalyst for good art and for rap might have to do with active inference expectations and experiences but I don't know this literature very closely at this point so, yeah, that's basically brings us up to the present hopefully that was a good summary Amazing, just great to connect and to be able to share this moment with someone who's seen it from all sides in front of the stage, on stage, behind stage inside the mind, real, recognized, real so there's a lot out there and just like as if we were talking about a paper Namaste, real, recognized, real, namaste If we were talking about a paper we would read between the lines and reference out and similarly, there's so much that can be referenced out and people can listen to the original creations and see what that primary text looks like and the fact that it's an orally performed text for much of human development and history was our real so whether something is written is increasingly fluid these days with speech to text and text to speech so there's gonna be so many great avenues let's just start with event-based cognition so what is an event from the cognitive perspective? I mean, I was trying to find that definition with Google about 20 minutes ago so I didn't sound like an idiot in this podcast so I think it's a, you know it's a perceptual construct, right? It is the perception of an entity that there is some kind of instantiation of action that is circumscribed in time that has a beginning and an end and if it's not perceived as having a beginning and an end by some cognitive entity, historically organisms but in the future also probably AIs then it's not treated as an event which means an event could be a thousand years or five seconds or whatever it's actually interesting how the kind of the scientific definition of it in active inference research really is like a narrative definition in a way, right? It's like it's a story something you can tell a story about something you can construct a story around is an event is that pretty close to how scientists think about it? Yeah, I mean, already so much there you're right, it does not have to be tied to a given amount of clock time which we might call chronology parking back to Kronos that sort of decimal point time but action takes place in the understanding of the actor or the observer in this time scale of action or Kairos and that time scale of action may be narrative and that is what you invoke with the beginning middle and end, rising and falling, cultural patterns that's not necessarily happening on the clock just cause you can run a stopwatch while somebody is reciting Canterbury Tales doesn't mean that it'd be twice as exciting if they spoke twice as fast or that you could map in some linear way what was happening on the watch face to what's happening in the mind of audience who have different familiarity with the material or a historical study context as well it just made me think of like if I do a 10 minute long wrap of The Wife of Bath's Tale then that's an event I performed The Wife of Bath's Tale there was an audience that witnessed it you could circumscribe it that way but you could also, so you could do a scientific study on what occurred during that event, right? But you could also choose to study the event of humanity's retelling of the story of The Wife of Bath's Tale which is a story that takes place in King Arthur's Court so now we know where the mythology has to have started which is around forgive me medievalist but let's say like four or 500 AD and then whenever the historical King Arthur's Court existed and then there was an oral mythology that adhered to that over the course of a period of time as people told folk tales about it and then in the year 1400, Jeffrey Chaucer well I think he perished in 1400 so he was born in the 1330s but he writes The Wife of Bath's Tale at some point as a written representation of the oral narrative and then that vellum manuscript is passed down into college reading lists and I read it in 1998 and write a rap version of it so any study of humanity's engagement with the story of The Wife of Bath's Tale has to really start in the year 400 and the last time I rapped to my rap version of The Wife of Bath's Tale at the Edinburgh Fringe Festival which was a couple of years ago and no event that accounts for the phenomenon of the telling of that story can leave any of those instantiations out, right? Awesome, it's like the earlier events are our prior or our generative process which is our niche that events happen in our generative model, our cognitive model exists within the niche, the generative process and those prior events quite literally are our priors and there's no last chapter because the observations keep rolling in and whether the observation is exactly as expected whether that performance of that symphony is exactly as expected whether that performance of the hamburger or of the portrait even if it's exactly as expected it still is a new observation and then what is as expected? Well, that's in the encultured cognitive mind and that is this sometimes implicit reference point that novelty is contrasted against like, well, I knew I was gonna get a rap summary but I didn't know it'd be that funny or I didn't know I was gonna get a rap summary. These are like different levels and types of novelty about the story of a moment. Or look at how he changed the story from the archival version because there's definitely a completely different experience of people that are intimately familiar with Chaucer's version and ones who are hearing the story for the first time and encountering each of the characters. This is the kind of expectation of how accurately will the story be recounted in this new version. Also, let me just expand on something that you said. You said there's no last prior because there's constant re-instantiations and updating, right? But there is a last plot prior from the perspective of a temporary analysis of an event. You can set a sort of artificial beginning and end for the purpose of doing an analysis and that's I guess where you're able to apply mathematics to it but you always have to keep in mind that it's a chosen beginning and end point within a larger unfolding. Yeah, and that's very resonant with my experience as a researcher which is like there is a final data point in any time series. At some point, the fMRI machine or the ant foraging watching machine, you just turn it off for the day. So there is empirically something that's very addressable but in the general case, it's not the case that a time series algorithm says what the last observation is. And we might even say that that's a deflationary perspective on history because rather than debating whether somebody has released the definitive American novel or the last word on this symphony, it's like, well, the last note of their symphony was non-controversially the last note of their symphony but the question about the discourse and the community, of course, there's no last word in this open-ended space and art is always like surfing on that open space. One of the things I remember from my comparative literature courses is this idea of the anxiety of influence and how poets have their, I think it's maybe Harold Bloom that wrote that but poets have a sense of needing to create something new. I think this general applies to all artists, right? Theater rappers, this is like imagination conjures but of course, great artists produced on the shoulders of a lot of influence in the past and one of the measures in an English literature course of whether a poem rises to the level of worthiness to be studied at university is the degree of impact it had afterwards and that can sometimes be measured by how many imitators it generated. So one thing I always find fascinating is you read 14th century literature and Chaucer stands out as the main poet, right? You barely read any 15th century literature in an English lit course or at least when I was in school in the 90s that was the case. I've just gone all blurry, haven't I? I'm gonna see what's going on with my camera here. There we go, refocused. The camera has its own priors. So like Chaucer shut down English literature for like a hundred years and when I was in an English survey course, I'm like, why do we read Chaucer in the 14th century and the next poet we read is the 16th century? What happened to the 15th century? And my professors were like, well, every poet that wrote poetry in the 15th century was seen as so much of a Chaucer knockoff and derivative because his style was so influential that none of them rise to the level of having contributed anything distinctive until over a hundred years later. I was like, wow, now that is impact, right? That is like as much impact as you, but ironically, it's like negative impact. Like the most significant poem shuts everybody up because they don't dare to do anything except for imitated in a totally derivative way but creating something new out of it, that's what creates the most fecundity in art, right? Like you create something and a lot of our people are like, oh, I love this. I'm gonna take it and adapt it and run with it. That's one of the things I love about hip hop is it's so much about sampling and flipping and remixing and it's not afraid of any other genre because its attitude is you can take any genre and make it better by adding beats and rhymes but really effective art shuts it down. I think Andrew Marvell was also known as like the last Carpe Diem Poem writer because he wrote such a perfect Carpe Diem Poem gather ye rose buds while you may that no one really bothered with that genre anymore because nobody could do as well as Andrew Marvell and he wrote the last great garden poem as well English Garden Poem and it's like, how do you measure your success as an artist by the degree to which you've created something that's so intimidating for everybody else that they don't even bother with the genre anymore and now you're measuring your success by the deletion of genres from the cultural milieu which creates space for other people to invent more genres but this is an interesting way to measure success. Wow. I'm going to pick up on that pattern and bring it into attention and the Bayesian perspective on attention which is sometimes equated with precision. So if we're paying a lot of attention to a sensory input a lot of attention, high precision we're really putting a lot of credence in that information it's overriding our priors a lot we're updating our beliefs to basically be what we're paying attention to and then the other case would be we're not paying attention we have low precision or high uncertainty and we're not updating our beliefs at all that's what we basically conversationally mean when we say we're not paying attention to that like I couldn't tell you what it was it didn't leave an impact and when you were talking about this cultural evolutionary process I thought about cultures paying attention to genres and works because if there was one work that was the object of attention it would be in the spotlight but attention is like a spotlight and if it's a laser on one thing it's not on everything else and if there's no work that's rising to attention then there's this proliferation of hypotheses cultural endeavors but no one of them will necessarily be shared and so art that resonates with communities and with your friends it's like in this gray zone where it has to have enough attention on it so that there can be a culture of hip hop or the Grateful Dead or whatever it may be but not one that is so attended to that it gets burnt by the laser yeah it's like if everyone's only paying attention to one thing then that's actually very stifling for cultural flourishing because you don't get all the variants you get these totally dominant moments you can see that at the macro level in history but it also just happens in social media all the time too right everybody's talking about one thing and then it's like what if that thing is not interesting to you or what is the upshot of that dominance of attention in active inference we sometimes talk about regimes of attention and there have been some great recent papers on cultural regimes of attention and that kind of takes that basal Bayesian idea of attention as precision to stimuli and it puts it into the dynamical space where the regime of attention might be like the sequence of eye saccades across a page so that is a complex pattern with attention moving and we can just call that broadly like a regime of attention or attention and action and in some sense that's what culture is or can be described as is regime just the kind of summary of where you happen to look or is the intent that it's like there is some kind of intentional control of attention that is inherent to the work that's being attended to like you can expect people to look others in the eyes when you're seeing a video or a photo and the regime is that like there must be some implied like maybe an inbuilt psychology of things that we tend to find more fascinating is that what's implied by it or? Oh well a lot there the word regime does have a slight connotation of something that's imposed whereas it is in the neutral sense describing merely the dynamics and then what those shared regimes of attention might be like we're going to approach each other this way making this kind of eye contact and make these kinds of gestures that's how we're gonna be reducing our uncertainty about who we are and what this event is and without those shibboleths secret words and implicit or explicit processes and rituals it's hard to lock on culturally and there's so many ways to go from there but I'll leave it now I'm always whenever I see shibboleths and norms being acted out I'm always trying to I mean obviously there's scientific studies that are trying to unpack this as well but my favorite thing to speculate about is to what degree is this culturally constructed in an open-ended way and to what degree is the cultural aspects of it relying on an underlying cognitive psychology that is inherited from evolution like for instance if you're having a conversation with somebody one-on-one and you never look them in the eyes that's inherently suspicious like what's wrong with this person they won't look me in the eyes are they lying to me do they have something to hide that response that you get emotionally and in terms of somebody's disposition towards you I would find that very unlikely to be 100% culturally constructed like could you imagine a society that was egalitarian where they never looked each other in the eyes I can imagine societies where aristocrats prohibit people below them in the hierarchy from looking in the eyes because it's seen as a sign of disrespect which of course is entirely based on primate behavior when you're like challenging an alpha if you look at them for too long then they'll bite you and you gotta like cast your eyes down to demonstrate your submissiveness or whatever like humans have replicated some of these very deep mammalian instincts that we have right but like are we malleable enough that a society could ever evolve with the norm of nobody ever looks the other person in the eyes or would that swim upstream so strongly against our instinctive ways of measuring our degrees of trust and affinity towards each other that that culture would just inherently fail and not manage to be represented in the monopoly of cultures going forward I'll try to give a thought on that calling back to this attention discussion so the movement of the eyes does provide information whether it's deceptive or not it is an outlet of information from the person who you could be looking at so it's in the... Including attention itself, right? It's where you look to find out if someone's paying attention The eyes are the window to the soul maybe that wasn't just about the cornea attention as the window to the soul It was about game theory And so to neglect this fountain of information you have to have a cultural clamp on it like a secondary reason to not look because it is providing so much info on a millisecond by millisecond level in ways that aren't always even expressible like pupil diameter has a lot to do with arousal and psychologically people may or may not know about that empirically and I doubt it was shown empirically until we had cameras and measuring devices and millimeters and all of that but it can still play a role implicitly in behaviors and heuristics and strategies and so it's like if we're paying attention to our embodiment, certain things will arise and in situations where the inherited prior has to overrule or reinterpret those first principles takes those are like scenarios with an attention to the past for better and worse that is shading our experience interpretation versus the prima facia situational awareness that convergently evolves to a few attractors like looking at informative sources of information Yeah, and just because there's a lot of swimming upstream of sort of instinctive cognition necessary for cultural norms to play out doesn't necessarily mean that they're doomed to fail or a waste of time or whatever you could almost say like the whole scientific enterprise complex mathematics, even reading none of these things are instinctively evolved they're all very recent and we're like building them on top of our older cognition and architecture and instincts but they're gloriously culturally constructed and I would say like worthy of the effort of swimming upstream the keyboard Q, W, E, R, T, Y oh, well the keyboard doesn't have to be that way well, yeah, but we don't need keyboards and we haven't always had them and we won't always have them and left to right reading it could be right to left or up to down or it could be any but we haven't even had inscription technology so maybe let me bring it back to your experience with hip hop how has the generative process the extended cognitive niche of technology supported the remix culture and how does that manifest? Well, I think I'll start with some of the good ones I used to go to a studio and sit with producers and hear their beats and like, let's go with that one and then like try lyrics I had pre-written or sometimes write the track it was really expensive, fun, interactively I do miss the in-person studio experience but all the previous rap records I've made for the past 10 years were all made remotely with Dropbox so the rappers, you know, send or the producers send me a folder full of beats I choose the one that I like the best I write the rap I record the raps at home I send them to the producer and I've got like a track where the producers based in Seattle the session musician is playing keys is in Vancouver the vocalist is doing the song from his place in Toronto and I'm in Long Island like taking all the different parts and editing them and then sending them to the mix engineer and the master, you know it's like this great distributed collaborative process and that's even by today's standard that's analog, right? That's like sharing files in a Dropbox and throwing them into your local studio and sort of like measuring each input against the expectation a song to sound ultimately and giving people feedback and asking for revisions but I think there are now AI sort of web three platforms where you can all be in the same virtual room and the latency's worked out in your collaborating musically I haven't used those yet but I've been, you know I've been scoping them out so that's been a very I think that's generally been positive it's brought the cost down and it's allowed for real time well, not quite real time but like ongoing collaboration with creatives across different areas but then the kind of parallel revolution that's happened is from selling CDs selling tapes, selling CDs and then Spotify takes over and you're getting like fractions of pennies on the stream and your music is going into this massive archive of every kind of music and you're only getting paid based on how much your music is listened to compared to every other music that's on there they're partitioning their subscription fees and it's really, it's become much harder I think for a lot of musicians to make a living from the sale of music and so over the past 10 to 15 years that process really pushed a lot of artists into really just basing all their income on live performance like set up gigs and rock shows and sell merch at shows if you want any expectation of income from music sales you better be there meeting the fans shaking their hand and selling them a CD in person then COVID shut that down for a couple of years anyway it's coming back but my estimation during the pandemic was even as it comes back it's gonna, there's gonna be people that are more reluctant to travel to shows and do that have the in-person experience even when public health notices say it's safe because people have different degrees of paranoia and anxiety and they won't feel as comfortable even if it is objectively safe so pivoting to this event rap game was partially based on an analysis of what's happening with the industry and how hard it is to make it as an artist and part of that has been a technology driven process. Wow, also many ways to go let's kind of return to those expectations and I think there'll be some ways that your experience with expectations might map or resonate to the Bayesian brain so say a little bit about what are your expectations when you're coming out to a performance whatever setting that you choose to think about and then how does that interface with or to what extent do you even preempt or pregame the audience's expectation? Do you just call it like you see it or are you looking to finesse your performance based upon what you rightly or wrongly believe their expectations to be? Well, there's an ambient characteristic that I kind of just take for granted which is that I'm not very famous but I have a hardcore boutique fan base of the like science nerds and science rap nerds that so if I'm in front of a crowd most of the time more than half of them don't even know who I am and I've got to introduce myself to them for the first time and some portion of them have seen a bunch of my shows and they're fans and they're like, oh, it's his next show but I don't really write to address an audience that knows me. Generally, my assumption is that they don't and I'm constantly reintroducing myself and finding new fans that way. And let's be honest, like you see me walking down the street you're not gonna be like, there goes a rapper. The Bayesian priors of my ability to perform as a rapper are low and they have been since I started I was this bookish, suburban Canadian, woodsy teenager, like I wanna be a rapper but no one's gonna believe I'm a rapper so I better get really good at rapping and also just, it is a bit of a choice not to like dress overtly to hip hop or talk overtly to hip hop. It's like there is a hip hop persona that a lot of artists adopt maybe it becomes natural to them and they could say this is who I am there was either a conscious or unconscious decision at some point in their growth where they became more seemingly hip hop maybe to reduce the cognitive dissonance of the lack of expectations of them being able to rap from their audience maybe for other reasons, maybe peer influence maybe they just, none of that analysis really matters and they just felt it, they were passionate so they started adopting the lingo and I kind of always partitioned rap a little bit like when I'm performing I'm a rapper and my profession is I'm a rapper but I'm not gonna like make myself a cup of tea in a rap style saying it's like I do distinguish a little bit between the domains of my life and what that partially what that means is people don't mark me and expect me to be good at it or to be able to deliver and then I get to, there's a cost benefit there for one thing like the skepticism might lose me potential opportunities to perform or potential members of the audience who like see me setting up the mic and they're like that guy I don't think so and then leave but on the other hand the transformation I get to take an audience through as they realize what they're experiencing and feel their expectations being overturned is actually I think really part of the joy of the experience it's an expectation shattering experience when someone that you don't expect to be able to deliver in a certain art form just really does and I kind of play with that in the comedy and tease the audience for having low expectations of me and stuff like that but then part of this is contingent on me just never getting that famous if there was a point at which I had achieved as much as much of a fan base as M&M my like writing style about you don't think I'm gonna be able to rap do you while watch this would of course sounds pointless kind of like he stopped writing about being poor after his first album and first line of his second album was they say I can't rap about being broke no more just like naming the critical reception of his transition from album one to millionaire album two so I think there is a kind of interplay between artists and their audiences and I stay niche and I write to the expectations of that niche we're all in our own niche so I hope that part doesn't change but very interesting when we attend we don't even need to say pay attention and use this economic modality but when we attend to our experience and update our priors just statistically that's called learning and I think that speaks to this synthesis with the language the lexicon the slang or as we might call it the active inference ontology speaks across a continuum from the experiential and the folk psychological the day to day it's the same terms with the same authentic relationships that are increasingly becoming embodied in our intelligence algorithms again not even calling it artificial intelligence algorithms because there are extensions into the niche from us so these cognitive models that are being deployed AI artificial intelligence or large language models they are cognitive entities and they're participating in these ecosystems of engagements with us so I had a conversation with the niche and I had a new chat GPT driven Bing engine last night I just got added to Bing the AI driven version it's interesting the first thing I did when chat GPT came out was I said write me a rap song in the style of Snoop that explains how to change the oil in a Honda CRV because that's exactly the kind of project that we do with EventRap EventRap is like a team of skilled professional 10 year veteran artists that'll make you a song on any topic and either record it as a video that you can share as a kind of public engagement or explainer or come in live to one of your events virtual or in person and perform but it's a topical rap agency all the rappers have their own catalogs but they don't perform them at the events we book they always create all new stuff and chat GPT's Honda CRV rap oil change rap in the style of Snoop was uncannily on point it was very... it had the Snoop DOG lingo it was an accurate description of the steps that you would have to take to change the oil in your CRV it was like a cogently... from the scripting perspective I felt like we had a threat on our hands chat GPT has definitely shook a lot of people's professions, marketing, copy professors, academics it's writing papers the economics of this thing have not even begun to dawn on people yet but the question in my mind was how soon is it coming for the rapper's job and there was a sophomore cartoonish kind of kindergarten flavor to the way it's constructing bars so chat GPT and Bing by extension I think it's the same layer that's running it it writes raps like someone who's never rapped before so if I told you write me a rap, you could do it but there's layers of proficiency of master level stuff of syncopation and internal rhythm within the main stresses of the bars and especially and this is what I was chatting about with Bing last night is polysyllabic rhyme so polysyllabic rhyme seems the construction of multi-syllable rhymes out of parts of other words and I was like asking Bing can you do polysyllabic rhymes and Bing was like yes I can and I said okay give me some examples from mainstream rap of polysyllabic rhymes and it pulled an example from MF Doom which was an excellent example it was like the girls are crazy schizo but she's got amazing tits though so you know you can hear the pattern crazy schizo amazing tits though it's like multiple words constructing a longer pattern of syllables I was like Bravo Bing you found a perfect polysyllabic rhyme now write me a rap that explains your internal data architecture or your algorithmic architecture in the style of MF Doom emphasizing polysyllabic rhymes couldn't do it all monosyllabic like check me out I'm Bing I do a verse I am artificial intelligence I don't even need to rehearse and I was like Bing those are not polysyllabic rhymes you started out quoting MF Doom complex five syllable rhyme pattern and now you're rhyming rehearse with verse like don't insult my intelligence and I was like dissing the AI and I was like I'm sorry I'm a new model and I will get better at polysyllabic rhyme but basically that is the thing that's got our job safe for now because if rap is an elite art form that takes a decade or more to master the techniques of flowing live freestyling polysyllabic rhyme and rhythm patterns and also just a confidence to perform are the things that we're still getting paid for even if ability to write a basic like competent rap is now easily achievable by AIs I'm sure even outside of the rap profession these conversations are happening they certainly are in research and education previously the artifacts that we exuded into our niche were mnemonic almost at best it was probably transformative when some of these oral traditions were able to be written down and that may have also been seen as threatening to the underlying basis of the skill and the transmission and now we had just a brief window of these inert externalized artifacts these mnemonic devices have only existed a.k.a. writing for however long they've been in the niche and now it's like they've come not quite to life but certainly to activity and now these off sourced and outsourced cognitive entities are not just receiving and holding but doing and that transforms who and what we are in terms of our perception cognition and action loops they're going to be different and that is what any given particular thing is so any technological innovation that changes how we interact with each other and with our art forms yeah and from the technical side we use the word thing in active inference to describe the blanket and the internal states so we take a particular partition we cleave off the thing we're talking about from the environment the generative model of the thing is the blanket and the internal states and so that thing is defined in terms of its perception, cognition internal states and action so when new sensory modalities or sense data appear when new action modalities affordances appear and when our thinking is different it's a different thing it isn't just like well we mixed this one thing with another thing and they've both retained their identity it's like there's been a variation and a learning and we're in the middle of some fractal dimension of it yeah one of the I think this is like a radical break from previous technological changes one of the things I argued in my thesis when I was a master's student was that poets conceived of their audience and their ability to connect with people orally during a master's time reciting your poetry for a listening crowd but then the invention of the printing press with Gutenberg changed that ability to imagine an audience because it was so much easier to imagine our silent readership of 10,000 people reading your pamphlet but making 10,000 vellum manuscripts by hand was always prohibitively expensive so no poet could even imagine ever having a fan base through text Gutenberg allows for the possibility of having a fan base through text which changes the whole art form suddenly rhyme is extraneous because they're going to be reading quietly anyway so who cares what the poetry sounds like the main thing should be carefully constructed syntax and story and semantics in a concentrated way but you see then this rise of free verse poetry is sort of like shorn of all of its ornaments and my hypothesis was that they were just basically irrelevant because people were reading quietly and not reciting out loud Edison invents recorded sound which creates the possibility of an audience and fan base that's just as vast from the sound of your words as it would be from the silent readership and that paves the way for the possibility of hip hop because this is like people that are writing things rhyming verse and of course I call this the rhyme renaissance right poetry was all rhymed then none of it was rhymed and then if you think of rap as poetry now it's all rhymed again because who cares about poetry rap dominates like all of our verse consumption compared to every other kind of poetry out there right and rap is rhymed and I think people instinctively resonate with rhyme and just took a break from it when they were reading silent pamphlets published by poets who were imagining their readership but each of these examples is like a technological innovation that mediates the creative process with its ability to connect with an audience but this AI thing is something completely different isn't it it's like competing with us for in our production of the creative things and right now I'm taking some comfort in the fact that it's competing ineptly it's not as good at writing a rap as a good rapper but you know I'm probably a subset of professions that can still say that and lots I mean it's passing the LSAT it's passing the all the master's entry exams and you know which rarefied skill sets will be left maybe I could comfort myself to say the rapper will be the last to fall to AI because of the combination of creativity and semantic cogency and humor and all the things that go into it that the AI is going to have to work each of those out like maybe stand up comics too are safe for a while because jet GPT can write some jokes but they're corny jokes still right like original jokes it writes aren't that good compared to you know a Dave Chappelle so but what are we talking like ten years of safe jobs as artists or less you know like the speed of this last six months has been a little mind blowing who is the deep blue who is the Gary Kasparov of performance what is the Turing test and for you to again calling back the real recognize real I listen to that as being like kind of a Turing test is you have cognitive entities in this moment of reflection which has been called theory of mind and in active inference sometimes is called thinking through other minds TTOM because there's this cognitive emulation of if I do this then they'll do that and so thinking through other minds used to be a defined set of types of minds we had children we had adults like us we had adults in different cultures we had squirrels and then the workshop is open and the composability is extreme especially with algorithms that can construct other algorithms and all of a sudden that ecosystem of shared intelligence thinking through other minds we don't even know if it's a sports car or a horse and buggy or a spaceship and someone's like do you want to get in and it's like where are we going what is it going to be like all those expectations that we might want to have about a journey and an event and who can say what with any precision things will be like so what do we do then except it's more like a spaceship the pilot is like do you want to get in if you don't all these people are going to die and so it's your choice whether you want to get in but lives on the line because obviously people that work in all kinds of medical fields and scientific fields will point directly to the amount of human suffering that they're on the verge of alleviating using these tools and the applications to wrapper job stealing are merely you know spandrels of the main focus of it which is I think mostly very laudable and very necessary like let's not get on the spaceship means we continue to put up with all kinds of genetic diseases and levels of poverty and public health crises and things like that that the AI is being applied to and is likely to do a great job in solving yeah and someone says look if your or your old still wants to do art by hand we don't need to stop that but we just want to use it here that's going to be something that's very challenging to take a long term perspective on because it's going to seem like a reasonable compromise in the moment just like everything always does and what compromise was made when those oral traditions were written down what compromise was made when that sheet music became recorded and that became canonical and distributed and so what is the nature of the compromise and who's at the table in 2023 yeah that I think that's exactly the right level of analysis and you can say on balance that each of the compromises we made for some new technology to you know give us some but also benefits it was a totally understandable and reasonable one you know from the perspective of like when I look at an individual poet giving up on rhyme because they want to do pamphlets and reach more people I'm like well they got a family to feed you know that's the way that they can achieve the notoriety and acclaim that is their path to success as an individual and it has cultural ramifications overall and that you know that's I think you can put that like people can opt out of any technology and the people that decide to opt in have totally defensible personal choice reasons for why they're doing so usually to get a competitive advantage that was the case with you know tools going back to the stone age right I only throw rocks I don't use bow and arrows alright you do you I'm going to take down this antelope over here you know we'll give you a piece if you're lucky you know that's just that's that's the AI tool choice that everyone's facing out to you it also makes me think about art always playing this role of like the tip of the spear that kind of penetrates the veil we had the discussion about NFTs as if they were just a question about art but of course digital certificates are a lot broader and more significant than just visual images and there's a discussion about whether certain kinds of art are being replaced by generative algorithms long ago six months ago 12 months ago when it was pre chat GPT but there was a lot of the stable diffusion and midi journey type algorithms then the conversation was more about graphical design and professionals but of course that is a broader discussion that isn't just about the generation of images that you are paying a creative professional to make and it's like art as this frontier and mediator with the open-ended will people be looking for open-endedness and liminality in their events in 50 years I think they will be that niche the marginal niche of art as well as the central role of art I think it's fair to say we'll both exist now whether that cultural attention on the central artifacts will be about artifacts made by a person an algorithm or something in between if somebody designs a 3D sculpture but then it's printed or if they think about a drawing and then they use a pen that was produced by a machine who's to say in these extended scenarios whether one or the other will see it but in the margins it's where we're always going to be exploring in whatever people call art and people who are in the middle of the spotlight saying that's not art it's not their art but every single genre has had to have this mitosis or birthing process with people who weren't part of that process saying it wasn't art the right way to do this kind of science or rap is not music as a hip hop partisan I'm very familiar with defending against the rarefied tastes of people that want to gatekeep what constitutes art and what doesn't but one thing I've always loved about that aspect of it not being gatekeep by elitists if you want to be a professional ballet dancer like Billy Elliot's or he's not withstanding you've got to try to get into Juilliard you have to go through the process of rarefied judges that are telling people who can and can't get into the next one there is no backdoor to success in some art forms classical music opera but with hip hop there's no frontdoor to success it's like you've got to get good at it on the street level there's no degree that'll get you a job in it you just have to find crowds at the level of your neighborhood and then your city and then your state and if you survive the levels of competition on the way up then you get to the world stage or whatever but it's very like pick up a mic and show what you got kind of thing and it's mediated by the audience and the fans if they like it it's right and if they don't like it it's wrong and MC has moved the crowd and there's a fairly objective measurement about whether someone's doing rap well or not and it's how the crowd is responding there's no valid argument that everybody's booing you but you're actually awesome because they just don't understand your art no you failed to understand the context in which your art was meant to be consumed so whenever we're looking at these trade-offs often it comes down to producers and consumers each new tool is good for consumers but bad for producers and this is like the taxi guilds trying to fight off Uber Uber is great for people that want to ride but it's bad for people that are going to try to control the means of production and you're seeing this exact process play out with chatGPT now which is like and also with stable diffusion and the other examples you gave like people want to consume art and they don't necessarily care who made it as long as their reception of it gives them the feelings that they want from the consumption and I think that's the case with music and with poetry as well of AI generated music and poetry and comedy because it's not sophisticated enough to produce the highest level stuff although I think with classical music it is and with graphic design and art it is like the best AI art is actually competitive commercially or from a consumer perspective like taste test wise with the best rarefied professionals which is devastating for them but it's not unique to do for centuries is protect their status as the elite producers of a thing from the ability of just about anybody to consume it for as cheaply as they want and that's what technology keeps delivering and in general I'm actually a fan of that process I think it's democratizing it's good for everybody on average even though it perpetually upends the guilds but we shouldn't give guilds control over our ability to consume excellence and if excellence is delivered better by AIs then let's not be curmudgens about that and I can actually but there is still an individual challenge like okay what am I going to do to make a living now it does really create a lot of potential unemployment and economic devastation which is why a lot of people have thought that there's just no escape from universal basic income because the jobs are going to fall and society will descend into chaos with that much unemployment and frustration and resentment and anger but even if rappers are the last job to fall the analysis has to apply to all kinds of other jobs first and active inference comes into this because how does a rapper entertain the crowd by having a mental model of what the crowd is going to do if they perform action X what it will be the effect of X on Y Y is the crowd X is the intervention and the intervention is a choice of a bar I'm going to rhyme this with this I'm going to make a song about this topic I'm going to try this joke out in my flow as a punchline and you have a mental model oh yeah people are going to think that's really funny you know and then you deploy it and you see the degree to which your expectation has been fulfilled and I don't think AIs that good at that yet because it doesn't model expectations of audiences very well yet does it it doesn't have deep theories of mind of like perceptions of perceptions of perceptions it just has averaging abilities to recreate patterns that it finds in vast troves of human created art which includes pixel patterns and graphics and patterns in text semantic and syntactic patterns but like performance art with an audience requires a mental prior of the audience's disposition and attitude and that's why I say rap's going to be the last to fall but and that's also why it's really hard to get good at you know it takes years of performing dying on stage and sucking and being booed before you really build up this mental model of like I know what people like and I know if I do this they'll do that and even I've been doing this for 25 years I'm wrong all the time oh yeah that line fell flat all good all good you know either you like carry on or you go fix it that's where the that's where the AIs have to get to right wow so much one thought was rap will be the last to fall and the first to rise around that campfire again it'll arise situationally you mean after society descends into chaos and all the AIs are resting and we're back to spheres rap will be the first to arise again first in last out sorry go ahead what was the next thing I mentioned taste makers and I thought what is the ultimate or even the proximate taste maker it's our generative model of taste it's every person eating that dish the taste maker isn't who told them to go to the restaurant that's basically believing the hype and using the paradigm of advertising industry the taste maker is the taster that's who is generating the experience of taste taste isn't just in the dish before it is actually eaten by that person it's experienced and they're going to have some previous taste in their mouth or they're going to have a context or a cultural understanding or memory a trigger so the taste is generated authentically and situationally and yeah there's going to be a lot of volatility as taste making becomes actually understood to be personalized and that may look like the personal generation of media which has not even been plausible when you had only 33 rpm records then you have somebody selecting which ones to distribute but when you have algorithms it takes it one level further which ones you like to listen to people who like this also listen to that and this is like as far as we can see on another level of like you like this we're going to regenerate that story so that the person's voice is like that that takes it to the next level but also with enormous tradeoffs it brings it closer to how things really are well that makes me think of something we spoke on earlier which is the degree to which cultural norms are able to reorganize or swim upstream against instincts so when you say the taste maker is the taster the counter examples that jump to mind are rappers who are like JZ where he's like and before this song ends I'm going to start another trend like if you want to say the tasters of the taste makers I'm looking at people who merely suggest something is cool and it instantly becomes cool and everybody is able to post-hoc rationalize reasons why it has intrinsic qualities that they now recognize as amazing over and above what the taste maker told them to do but we are highly susceptible primates to cool factors so I would suggest the taste maker is not the taster, the taste maker is the best predictor of the average response of the tasters and when you get very very good at understanding instinctively how people will react to the taste using taste as a metaphor for all kinds of art then you're able to be treated as predictive on taste because you're right so often about what people will like and that power of like vested expertise from your fan base then allows you to start trends that are not just predictive you say something's great and everybody instantly believes it's great and experiences it is great as authentically as if they had discovered it for themselves and decided it was great because they've recognized your actual real average ability to predict taste across groups awesome I thought okay again who is the taste maker it's the chef and not everyone's in the kitchen because a chef is in their position because they've been recognized to repeatedly be able to construct dishes that in the local taste making diners like or maybe just a few diners like and it's like you can't do that with molecular gastronomy but some so it's like it's the chef and it's the social influence network and it's local what framework can we use to think about all of these interactions within and among cognitive entities within and among their niche we use active inference as just a way of thinking about it well what is active inference it's there in the title action and inference it's involving action and doing cognitive modeling and that gives us the composability and the ability to reconfigure and remix to capture all of these diverse scenarios not to close the book on them but to be like oh wow what an amazing question there's 50 PhDs of research that you could go into so it's a question generator and it's a lens by people who are even just beginning to learn it who review phenomena that they've already been told or experienced in one light all of a sudden they're seeing how it can be reconfigured for example seeing what we prefer in terms of what we expect rather than what we prefer in terms of what's rewarding which is the reward learning reinforcement paradigm which proposes well maybe you know the value of different states and your maximizing value it resonates with some contrast that with you have this complex generative model and you're always referencing to it and you're taking actions and learning to reduce your surprise bound your surprise of certain kinds which could include increasing your narrative surprise and you want that page turner or the horror film so it just kind of reflects back and then one last point on active inference as a research direction that you're interested to hear your thoughts on is it is happening in a time that has never been seen before in research and academia so you said there's certain careers or genres where there's no backdoor to success and then you said in the app there's no front door so traditionally even if we could say there was no backdoor to success in research the reality is there was a side door and people were getting kicked out and being brought into the party through the side door and a bunch of people were waiting at the front door they didn't even know that the show was not happening that way we're developing a framework in community at the speed of preprints and so there's transdisciplinary collaborations with philosophy historians of science cognitive researchers mathematicians, physicists so there's a transdisciplinary conversation and one with practitioners and one that's happening with everybody with view access to some large parts of it which wasn't even the case like break down the paywalls everybody should have access to the PDFs which is great but it turns out that having access to 200,000 PDFs having access to every single PDF that Carl Friston has ever been an author on it's not even going to take you 10% of the way unless you're going to engage in this total monastic life to unpack them and the details aren't even in what's necessarily those papers it's beyond that community it's in the software in the inaction so of course wait and see but there's so much with the way that active inference is developing on the influence of previous frameworks Bayesian brain and Bayesian statistics causal inference statistical physics happening in this sort of wrapping in front of the crowd way synchronously with the preprints synchronously with remote presentations and recordings that it's hard to think of another area that has had that kind of a developmental niche you're saying active inference is the hip hop of the scientific community basically you said it not me live on the mic I love it I think it's super exciting research and collaboration happening and it's very very unknown to the public you know if you try to explain what this conversation is about you know people understand the idea of expectation and priors and belief upgrading the very sort of basic version of the Bayesian algorithm but all the applications of it and the cross-disciplinary side and everything it needs more science communication I've made two Bayesian songs and I need to make 20 more because this is the applications of the research are vast and especially in the AI era people really need to know what these algorithms are achieving and the ways in which they are similar or different to our own cognitive processes and what we can expect them to achieve and how we can expect them to help and what they're threatening all that stuff goes through active inference doesn't it awesome segments would you like to do something musical or how do you want to proceed yes indeed I'd love to I'm very happy to do a demonstration of one of the last art forms that AI is not yet able to model and that is a freestyle rap so if we have anybody watching on the stream or Daniel you're welcome to do it as well if you want to pop some phrases into the text I will look at the text that you're displaying and also look at the comments and the chats on YouTube and anything that I see pop up I will visually process and convert into rhymes in a freestyle rap track and just to give a little insight into this process it is a violation and like expectation setting and expectation violating process like the active listener to a freestyle rap is trying to figure out the semantics and figure out where the rap is going to go and the rapper is also in real time figuring out the semantics because it's completely unscripted and at the same trying to set up expectations of the audience and violate them I think a lot of the fun is the surprise factor of like oh I didn't expect him to go there but oh now I see and like it's kind of like a game of chase like the rabbit and the fox so I see Sean Kelly's drop topology in the in the chat of YouTube I'm going to start there and I'm going to kick a little rap for you guys yeah this is hip hop tell me what you think this is it's Boba Brinkman rocking now an act of inference that's the inference looking for where info is but I feel it's kind of intimate yes indeed hip hop what you think this is is Boba Brinkman here on the active inference podcast I'm looking for where the info is so I can update my priors and get into it and maybe drop a beat Sean Kelly said topology yes right now I'm rocking hip hop ology can't you even see that's how I adjust the bass and treble topology that means checking all the levels and looking for the map and the different levels of impact I'm spitting rap yes I'm looking at the chat that's how I kick it yes how I go with rapping flows it is me kicking it with Mr. Daniel Friedman how I'm speaking yes I'm always just deleting variants every time I'm sometimes just tweaking through the speakers I just do this on a Tuesday or a week and indeed active inferences how I'm speaking I'm sneaking past all your expectations like a rat and a maze because my raps will amaze and yes I'm polysyllabic to a maze I'm just having to blaze yes I'm actively playing all the different rhymes on my brain like a piano changing the channels like I'm rocking gangsta Tony so friend oh yeah that's okay this is how I just anger the listeners sometimes cause raps are danger gotta give a shout out to Tucker Cahill Chambers who rocked this with us yes working the angles like bass theorem can you even hear them when I'm near them I start to see it as this mass scientific theorem that I could apply to every single single thing as Boba Brinkman gets all your neurons jingling it's how I attacked it I attracted a massive I'm active when I'm doing inference in my rap and indeed this is how I kick it when I enter my rap is directed it's never non-censored as I just try to construct it mistakes I make sometimes I'm run DMC and sometimes I'm Drake except I'm not quite as good of a singer with the melodies but then again I've got the freesties that kick a lot of active inference Bayesian styles free styles that's trying to make the people smile as to listen in yes that's how I'm straight rock and even a mod most off he says yo wow awesome with explosion emojis Boba Brinkman coming with the dope beats people are like hopefully he's one of those MCs that could rock straight up as ancient hip hop like rock him and Eric B I never do it terribly I just free style that means it's spontaneous my style is the zaniest I'm rapping for the Long Islanders plus the Canadians people listen and blow their brains when I'm blazing it yeah it's how it goes when I'm speaking on the basics I'm in Long Island but Daniel he's up in Davis California I put it on you it's active inference Boba Brinkman when I'm speaking it I'm getting into it looking for where the info is that's how I'm rocking this was like a four-minute continuous talking demonstration yes indeed I make it easy let's see you do this shit chat GPT-3 you can't do it you can't freestyle over this music and use a brain to actively infer that's so fluid yes that's a rap that I make up and invent the last four minutes has been an event prescribed in the time that's the freestyle rap that I did on this concept that's how I can adapt yes indeed that's a rhyme that I had to make I can do this my brain makes me a primate but I don't make mistakes okay sometimes I do but that's just mutation yes that's how I move all the mistakes get integrated my work is huge thank you Ahmad Mostafi I appreciate you okay now I'm starting to wind down and trying to just find a new direction for the rhyme sound this is how I just get into it it's Boba Brinkman rocking with active inference active inference this is how I'm never pretending I'm trying to infer some way for the rap ending but wait a second this is how I do it hip hop the easiest solution is just turn the music off alright that's freestyle rap and I really appreciated this chat it was a ton of fun and super interesting I'll give a closing remark and then any last words from you it brought a smile and a laugh to my face and body because another individual was engaging in the shared moment I've seen language models output incredible x in the style of y and every single time it brings a chuckle or less and it's like okay I kind of expected it to do that I want and I'm disappointed if it can't fulfill my request because I would have asked you this because I expected you to do that when there's joyful spontaneous engagements that is reflective of a shared experience it's the real thing and when that lens is applied it's going to be real recognizing real and I don't think humans will be out of a relationship role for each other for a long long time amen to that and I just want to say that the phrase that you just coined joyful spontaneous experience or joyful spontaneous engagement that's reflective of a shared experience that is basically the entire value proposition of my event wrap company is find a topic or an event or something that creates collective attention of a group and have that be the locus and the catalyst of a wrap because the wrap game is entirely predicated on make a thing that is very widely appreciated and goes viral and is hugely massive and making general purpose art is what most rappers are focusing on and I really thought it would be interesting to apply a rapper skill to this joyful shared experience and spontaneity challenge and thank you for giving me a platform for it and I hope your audience will now imagine every other possible application of a shared experience that a rapper could add value to because if such things occur then my business will succeed and lots of rappers will get lots of gigs rapping in weird places like this and I think that would add to the joyful variety of the world and the genre amazing so thank you so much Bobo always welcome to come back give us updates when that rap guide with inference drops alright I'm on it thank you thanks Daniel peace