Writing with Technologies Webinar Series - From Script to Screen: AI and Contemporary Storytelling

all right I think we should begin um just because we've only got an hour together so I think it would be a good time to start so welcome to this last session in our four-part webinar series um uh we're we're part of Basbar University's uh center for cultural and creative industries and the narrative and emerging technologies lab and through this series we've been offering an in-depth of um at AI's emerging influence across writing and publishing across multiple fields so we've invited many speakers and we've been looking at AI from from multiple perspectives we're also part of the wider my world program which is um leading on creative innovation in the southwest my name is Amy Spencer i'm a researcher at Baths University and I'm going to be um chairing this session uh so in this webinar we have four speakers we have Guy Gadney Sarah Gibson Yates Doug Spect and we have um uh I haven't made a note of Robin's surname on here robin we will introduce you in a second sorry my notes were a little scrappy um so in this session we're going to be thinking about the use of um AI uh in screenwriting so we've got researchers industry professionals together to talk about how how AI tools have the potential to develop complex characters develop dialogue and identify structural issues if you've got any questions during the session please add them to the Q&A and we will have um questions at the end so um I'm going to invite Doug to speak um first doug Spect is a reader in cultural geography and communication um and head of the school of media and communication at the University of Westminster his research explores how data and technologies influence environmental justice human rights and access to education with a focus on the production and codification of knowledge he writes for publications such as Guangh the conversation uh geographical and for the times higher education um he's a co-au co-author of the upcoming volume the student guide to the creative studio in the digital age so uh big welcome to Doug who's going to speak um on on this topic thank you very much well thank you and uh so it's always odd hearing your own biography being read out isn't it but uh thank you very much for the the kind introduction uh let me just start sharing uh some slides here can't do storytelling without some images when we're talking about script can we so uh hopefully you can see those uh slides without too much problem um so good afternoon um thank you very much for for the invitation to to speak on this topic i think um I'm probably going to start us off with uh some quite broad ideas around AI and creativity AI and and the interaction between how we we think creatively um which I hope will then feed nicely into the other uh talk that might uh delve a bit more into the the details of of script and screen and that kind of writing um well you'll see I'm starting here on my title slide with a a small clip uh from Content uh the film adaptation of Carl Sargon's book um and this is the moment that Jodie Foster's character uh is prompted with the the uncontainable wonder of uh the universe and in this scene uh I didn't dare try and someone's work across the internet even in this day and age uh in the scene as she beholds the size of the universe the glory of the universe uh she exclaims they should have sent a poet rather than an astronaut and I think this is a really interesting way for us to start thinking about artificial intelligence and thinking about uh the modern world that we live in because of course actually the birth of artificial intelligence the birth of the computer is itself wrapped up uh in poetry um the invention which the computer which is attributed of course to Adam Love Lace and Charles Babage um Babage has his own story but it's Love Lace who really leads us nicely into thinking about AI and what's interesting about Ada Love Lace uh you may well know that her father of course was uh Lord Byron um she didn't really know him he died when she was 8 years old she was raised by her mother uh who was very adamant that uh all of her father's influence should be erased from her life and so uh from the age of about four years old her her mother banned her from engaging with anything sort of poetic and instead uh took her down a route of maths and science and she became quite fascinated by mechanical engineering uh she wrote this the book that these uh pictures are from about constructing flying apparatus but all this time while she was doing these things she she felt that part of her the poetic part was being somewhat repressed um and she wrote to her mother in her teenage years "You will not concede me philosophical poetry invert the order will you give me poetical philosophy poetical science." This idea of poetry ran deeply through uh thinking u mechanically as well and through this Ada Loveace really appreciated the concept of the general purpose computing machine um she could envision that this could be something beyond just processing numbers but something that could also process symbolic notations including music and artistic ones and she saw a poetry uh in this idea and she set out to encourage others to see that as well but she was under no illusion about the limits uh here she said that the analytical engine has no pretentions whatsoever to originate anything it can do whatever we know how to per uh how to order it to perform it can follow analysis but it has no power of anticipating any analytical relations or truths she stated that no computer no matter how powerful would ever truly uh be a thinking machine and a decade or sorry a century after this Alan Turing dubbed this love lady lovelies objection and tried to dismiss this turing himself laid the foundations as we know for artificial intelligence his famous Cheuring test asked whether a computer could ever enjoy strawberries and cream or compel you to fall in love with it but Turing's ideas lead us to even more questions diane Aceman herself a poet asks whether uh when robots weep who will comfort them akaman notes that our relationship with nature is evolving rapidly but also incrementally and at times so subtly that we don't perceive the sonic booms literally or metaphorically as we're redefining our perception of the world surrounding us and the world inside of us we are revising our fundamental ideas about exactly what it means to be human and what we deem to be natural and to this end I would argue that we are reinventing the human race right now the pace of scientific progress is directly correlated uh to our alliance and allegiance with digital machines john Carol offers us an interesting idea when he states that we're all machines that think and the distinction between different types of machines is eroding we pay a lot of attention these days with good reason to artificial machines and intelligences ones constructed by human ingenuity but the natural ones that have evolved through natural selection like you and me are still around and one of the most exciting frontiers in technology and cognition is the increasingly permeable boundary between these two cats and I think that's where we then start to develop some really exciting but potentially unnerving moments in writing in script writing in film in television and in the creative industry uh in general where this becomes complicated is around our idea of what computation is uh the whole idea of computation is that once we have a complete step-by-step account of any process we can program it on a computer and we tend to think about the human brain that way computation is still the best and probably really the only scientific explanation we have of how an object like the brain can act intelligently but it doesn't really tell us how the brain works it has no way to really explain to us how we see creativity particularly the kind of creativity that we see in children it cannot be translated in that way and until we actually manage to understand that and create computers that think the way that humans think rather than emulating through step-by-step processes the largest and most powerful computers will still be no match for the smallest and weakest human beings i refer to the creativity of children because thinking is not computation it is cognition it is contemplation and this inevitably leads to imagination and imagination is how we elevate the real towards the ideal it requires a moral framework of what is ideal and in order to become script writers creatives filmmakers producers we need the imagination that comes through cognition and contemplation the most important thing about making machines that can think says Kevin Kelly the founder of Wired magazine is that they can and will think differently uh we believe that human intelligence is not is singular but it is not there is a possibility that there are other types of intelligence in the world and as we continue to build synthetic minds we'll come to realize that human thinking isn't general uh but is actually just one species of thinking the kind of thinking done by today's emerging AIs is not at all like human thinking but that's not necessarily when we consider it to be a companion to human thinking and much of the work around thinking about AI as a companion to human thinking looks at the idea of using AI to take over certain parts of human life the drudgery uh of the dullness of life what's wrong with turning over the drudgery of thought for high-tech marbles asked Standard Dennit well nothing he concludes so long as we don't delude ourselves and we somehow managed to keep our own cognitive skills but what we must also remember is that by outsourcing the drudgery we outsource some of those sources of inspiration and creativity uh out of adversity comes creativity of course if we look at the work of Rachel Carson from her text Silent Spring she gives us a glimpse as to where this might go she says "The future offers very little hope for those who expect that our new mechanical slaves will offer us a world in which we may rest from thinking help us they may but at the cost of supreme demands upon our honesty and our intelligence the world of the future will be an ever more demanding struggle against the limitations of our intelligence not a comfortable hammock in which we can lie down and be waited upon by our robot slaves and I think that really touches at the tensions between using artificial intelligence and harnessing it both as a companion or as a complete tool uh in the creative enterprise and just by uh way of rounding that up I want to uh share a debate um between a couple of uh well-known people um nearly two centuries after Ada Love Lace wrote the first algorithm uh Harrari wrote in his book 21 Lessons for the 21st century that artificial intelligence with its limitless potential and connectedness will ultimately render humans redundant in the workplace and we see that now entirely feasible to many people it doesn't sound overly feasible or at all but to many it does but what's more concerning is not the debate as to whether humans can be replaced uh I think is his assertion in this text that AI will be able to write better than humans can in his example he was talking about writing music specifically apply it to script to novels to any type of writing whatsoever he says we listen to songs to make us feel certain certain ways we can read books to feel certain ways we watch films to feel certain ways and that in future AI will simply map to the individual mind and create songs or texts that are tailored exclusively to our own particular mental algorithms that can make us feel with far more intensity and precision whatever we want to feel now this of course is something that Adela herself said could not be possible on the flip side of this argument and where I sit perhaps more closely is Nick Kay's response to this assertion because he says this isn't all that songs do and I would argue it's not all that writing does it's not all that film does nor television or scripts or any kind of creative endeavor of course songs do make us feel something happy sad sexy homesick excited or whatever he said but it is not all a song does what a great song does is fill us with a sense of awe there is a reason for this a sense of awe is most exclusively predicated on our limitations as human beings it is entirely to do with our or our audacity as humans to reach beyond our potential which is the awe of artificial intelligence but not the awe that artificial intelligence can mimic if we have a limitless potential then what is there to transcend and therefore what is the purpose of imagination ai we agree will have the capacity to write a good song but it will never write a great one because it lacks the nerve and just a little plug for a short article I wrote related to uh this um series of of talks i was looking at the brutalist and the way that AI was used there and how this made people feel uncomfortable it took us not into the dialogue more but out of it as we took away the awe and wonder of the human and replaced it with awe of what technology could do and many have talked about how that took us out of the film and uh caused this great controversy around the use of AI in the film the brutalist so I shall conclude by sharing that Alan Cheuring Adela Love Lace they laid the foundations of the perplexities of AI as a creative thing with their central questions around poetry and the machine and questioning whether a machine can think but it's impossible for us to think about the implications for writing whether that's text music film video whatever without building upon those ideas and asking can machines and can algorithms ever actually feel and I will leave it there thank you very much thank you um as a a great introduction to what we're speaking about today um thank you very very much um we're going to move on to our next speaker um we have uh Sarah Gibson Yates who is a senior lecturer in film media and writing within the CA Cambridge School of Creative Industries at Anglia Ruskin University um here in the UK her research reflects an interest in multiddisciplinary creative practice and um the uh ideas around uh film writing and creativity her current work reflects um the urgency to better understand the value and creative potential of working with AI technologies um other research includes AI's impact on sustainability and nature and the potential of entangled creative practices for understanding society and culture in perma crisis so over to you Sarah thank you very much sarah you're on mute yeah we can't hear you sorry okay sorry yeah so I just said a quick thanks sorry i just realized before I shared my PowerPoint that I needed to unmute and um put my camera back on so yeah I was just saying thank you for the invitation and thanks for the series which has been great so uh today um I'm going to just share briefly uh how I worked with on an AI um with AI on a recent screenwriting and film making project funded by the British Academy with support from my institution so I'm going to um introduce a summary of the main aims and insights that I got um from the project show a short clip from the film to kind of illustrate um the screenplay as it became on screen and then just a couple of brief words about some of the creative processes um creative decisions that I worked through with AI around that so um this is the project um a little of the heart uh was the screenplay that I I wrote a short drama film that made that made um that was made from an experimental screenwriting process using generative AI um particularly working with claude chat GBT and Google Gemini um so from early experiments following GBT's launch in November 22 I found the back and forth prompt and response critical reflective loop of working with chatbots very similar to the uh drafting and revision process of my own creative writing and academic writing um so I felt that there was something there interesting there going on that might be useful for developing my own uh creative practice uh and wanted to explore this further about the same time I also identified what I perceived to be a kind of a a situation arising within the film industry that seem to indicate two um kind of opposite or contradictory directions of travel one was around increased autom automation um so as an impact of AI particularly on screenwriting prompted by the screenwriter strikes during the summer of 2023 um uh so obviously this was raising urgent questions for everybody around authenticity bias um IP industry working practices and job security and then the flip side of that I kind of was noticing also uh around this recent establishment of a code of care um within screen production particularly around the formation of the role of the intimacy coordinator who facilitates and choreographs intimate scenes in film TV and um and theater ensuring actors um safety and consent so it seemed um these two movements seemed quite contradictory one around automata auto automation of creative working practices and the other one around a more situated human- centered um way of working so with this wider context in mind um uh to my own kind of personal interest in exploring the technology based on kind of work I'd done in my PhD looking at the languages and practices of social media i was already kind of very interested in uh technology and um storytelling and how those two shaped and informed each other i made um a British Academy funding application and and got a bit of money to support this project um so a little of the heart uh set out to investigate the entangled uh post-digital interactions that occur when human and machine um that human and machines uh interlocators in screenwriting and in the intimate moment of therapeutic exchange with AI chatbots um so this I actually came across this article by Evan and Ringrose um very recently last month and interestingly it kind of predicts the work that I was doing last year um I was aware of the uh their work in post-digital intimacy before this um but they've kind of really formulated um their ideas for this as a theoretical framework which I think is um quite interesting um and chimes a lot with um the work I've been doing and hope to do going forwards um so the work itself draws on competiatory literary techniques and traditional screenwriting processes try to bring those together and responds to calls for new narratives um exploring uh our everyday relationships with AI rather than looking at AI from the point of view of science fiction narratives I was looking at how do we uh bring narratives around the realist narratives around the everyday um looking at AI and uh the project uh aims to sort of posit new ways of generating contemporary screen stories about with and through AI um Clark's notion of creative amplication informed the iterative human machine dialogues that shaped the screenwriting process this process consisted in conversations about around ideiation log lines scene development synopsis writing and dialogue uh where I ran ideas essentially through chatbot dialogue mechanisms uh assessing the work via a combination of intuition aesthetic aims past experience of what works and what doesn't work for screen drama in my opinion and according to various other models for story structure and also dramaty um I looked at um and then added to what I saw as gaps or missing pieces or failings in those results um in various different ways so um which I also found this kind of idea of story construction through a process of filling in gaps um the idea that you have you know you start with an idea and then you kind of flesh out that idea and you build it up um through uh various different mechanisms um so essentially I've created space for inspiration to fill these gaps using the usual writerly resources such as um serendipity dreaming imagination free association and improvisation in addition to conversations with humans of course um cast crew friends and a script editor that I got involved um as a result of the funding I I received so all of these things are the normal way that ideas develop um and I'm kind of interested in how the genesis of ideas screen ideas particularly come come to play and not much has been written about this so in this way I was able to add the necessary detail plot points thematic development I needed um or wanted um these human collabor contributions are particularly effective at providing the screen imaginary and audio design that I wish to carry um much of the weight of the storytelling in the realization of a poetic realist style of film making um that I wanted to achieve so the resulting combatory screenplay uh which is a hybrid really of of human and machine input and that's very much the kind of the nature of the project i was trying to negotiate that kind of space between uh the affordances and limitations of what um those AI uh tools could um could provide anyway the story tells this uh the story tells um the tale of a young woman holding trauma from a past relationship who turns to AI for advice um an AI chatbot the ensuing film includes a real-time conversation with an AI chatbot and explores the contradictions inherent in asking a computational device for help with matters of the heart while also exploring AI's capacity to respond appropriately to the intimate personal embodied experiences of the protagonist um okay so I'm going to just show you um a short clip from the film um so I think this sort of illustrates uh let me move this over here illustrates what I'm talking about um should there be sound there were someone in the chat has mentioned they can't hear anything and I wasn't sure whether there was sound in this extract of the um yeah there there was we checked it before it seemed to work okay okay is it not coming through great question i can't hear it um trying to get someone you like to notice you can be a good way to get someone you like to notice you i have put it start a conversation as far as I can on my Yeah no we we can I can hear it now both like that you can connect on that's great do you like them too why don't you tell them you like this band you can also ask which is their favorite song that would be a great way to start a conversation that's so true it can feel really awkward trying to get to know someone for the first time but just imagine how great it could be to get to know someone new there are lots of ways to get over feeling shy time to make a memorable first impression okay I'll pause it there um okay so the rest of the scene so that's sort of um the Nia continues to share her anxieties in the rest of that scene it's a little bit longer than um than that it's not the the full end of the scene um so she shares her anxieties with the chatbot about starting a new relationship um and the scene occurs sort of midpoint really in this in the film which is 13 minutes long um after we have been introduced to Nia and uh and her love interest and marks the point where she must decide to speak to him or not the conversation with the chatbot provides a catalyst for self-reflection and self-standing she doesn't take the chatbot's advice directly in the end but uses this in between digital and human recursive space to reflect on what she is comfortable doing she is aware um that the chatbot suggestions for action do not share her embodied experience or hold the emotional trauma of a difficult past relationship and the doubt that accompanies thoughts of starting a new one the conversation on screen was improvised with a real chatbot snap u Snapchat in fact and prevented verbatim um or practically as you be aware chat bots tend to be very verbose and often um too too much too long answers so I ended up cutting down a little bit of some of the responses you'll also note that I adapted the textual conversation into a partial audio exchange so while I wrote this conversation in the screenplay as a text chat as you can see on the screen now on my slide um when it came to thinking about filming the scene and obviously my perspective as director knowing that I was going to um produce this direct and produce this film informed how I work with AI at the screenwriting process um I knew that that wasn't really going to work it was going to be too much text on screen with an undesired effect of slacking the dramatic tension and distancing viewers potentially emotionally from the uh from the scene so presenting the AI agent as a voice afforded me new creative opportunities when thinking about the visual and sound design of this scene specifically whose voice it should be we decided to use Nia's own voice um the actor who played Nia um's own voice adding a filter to mechanize its audio texture using the same actor's voice through interesting parallels between the therapeutic role of the AI chatbot and of Nia's internal processes of self-reflection effectively merging the two visually we included Nia's side of the chat in text boxes complete with emojis a dialogue trait of Snapchat bots um presenting AI as a character was a creative story solution um and a response to the sometimes limited characterizations that had resulted through the prompting and revision process I'd explored um with um with AI so to conclude um what I've described here is a way of working with AI that is highly personal but also aligns um with traditional screen story development and screenwriting processes using AI as a deliberate constraint i have gained a deeper understanding of the affordances and limited limitations of both sides of the experience that had tangible benefits for productivity within the context of the neoliberal university and late capitalism i hesitate to name productivity as a virtue as it's often overemphasized at the expense of other human- centered and caring practices however for me personally increased creative productivity has been a good thing and I will continue to work with it on AI on the next project um which is uh revising my PhD novel with and about and through so-called um AI thanks very much thank you Sarah it's um it's great to hear a practitioner's perspective to begin to understand all of this so thank you very much um we're going to move on to our next speaker um our next speaker is Guy Gadney he's C CEO of Charisma AI which is at the forefront of creative AI he has run three startups as well as working for international media and um other organizations so with Charisma AI guys transforming the use of AI in film television and video games focusing on the power of good storytelling and bridging the gap between the creative and technology industries guys um on the on Innovate UK's Bridge AI advisory board and on the board of the Oxford story museum and as a co-founder of the collaborative AI consortium where he researches the impact of AI on the creative industries so thank you very much for being here today guy and over to you uh thanks Amy hi everyone uh I will zoom through this quite quickly and uh hopefully help some put some useful um links in i won't go into too much background about me um other than I've known uh Amy and Kate for a little while now um and occasionally write for the writing platform which is what I've just put into the chat um so charismatic charisma AI both of them are the work that I uh that I spend time on um we started off looking at storytelling and and AI probably about 101 15 years ago uh yeah we were working in AI and storytelling that far uh back and about 5 years ago four years ago we started working with the University of Southern California to see if we could come up with a universal character model so looking at storytelling through the lens of characters rather than just uh narrative uh and we published uh we've published two papers already and the third one's uh just about to be published which is actually has a sort of GitHub link and everything that I can put into chat as well which is looking at how you might have a persistent character within a persistent narrative uh run by AI in essence it boils down to three things character conversation uh and context context is the bit that is the hardest um because there is so much of real world experience and demographic changes and uh geoloccation and culture that uh that we have in our biases around context is important um so you can't just throw the whole lot to an LLM I suppose is the is the uh is the short piece we've explored a lot quotes like Kurt Vonaggets there's no reason the simple shape of stories cannot be fed into computers i love this quote because there's one key word that leaps out at me which is the word simple uh we can we can feed the simple shapes of stories into computers but not necessarily complex ones and one of the key findings that we found around using LLMs for stories is that they are uh lousy at subtext uh they abhore a vacuum and they just want to fill it which means makes things like dramatic pauses quite hard uh but this is what we've been working on for the last little while um and when we then throw in things like interactivity a story can be incredibly complex it's not just this this single narrative that that uh has a beginning middle and an end but you all you will veer off up and down this sort of halfpipe that we've that we use the analogy of a lot um and that when you have user interaction with that for interactive stories things get more complicated as well and it's what we spend a lot of our time looking at so Charismatic uh was a project that started just over a year ago like two days uh over a year ago um funded by Innovate UK and with University of Falmouth and Creative Computing Institute at University of Arts London as our academic partners Channel 4 and Ardman Animations uh as our primary industry partners and it was really looking at whether or not we can feed in something as simple as a title and maybe a log line into a system and create in essence a visual and compelling uh story so what you're seeing here is an output uh that I did the other day for um a uh story uh the first step we do is that we write all of the characters all the stories the the the the context and everything that I talked about and then we visualize it in a sort of two two-dimensional what we call a read through mode so you've got the script playing on the right hand side you've got the story being visualized on the left and this is for writers to to to see how that story is manifested edit it go back change it and part of the view that we wanted to do here was to allow people um especially under reppresented writers who may not have had the privilege of going through film school Indian school at all to get the story out of their head into some form of uh of experiential uh piece very rapidly um we can then take that and put it into animated system graves outside what's the story there bro pets all pets so this is a a short animated horror because horror and in uh in in uh narratives are very difficult to do with standard LLMs um or we can do multiple different styles in animation or so from my perspective the challenge I give myself the whole time is how do we avoid looking at the future through the rear view mirror and how do we look at something which is which is sort of transformational in in stories um there are lots of inspirations that we seek in this but for us the the the idea of fluid stories of narratives that can change evolve with different people's 's interactions is critical uh and that is enabled by AI in ways that we we haven't been able to do before so uh in the interest of time I will leave it there uh but pick up any conversations in um in chat as well thank you thank you guy um I've spotted you've been answering a couple of questions in the Q&A which is great um if anyone else wants to add any questions in there we'll we'll get to the end but it's also great to um have answers to those um live as well so thank you very much and thank you for your presentation um moving on to our fourth presentation of this afternoon uh we have Robin Winfield Smith who is a stage screen and XR director based as the founder um of Liinal and development researcher at the Hatrick Lab where she specializes in the development of trans media spectacle IP universes that harness emerging technologies in real time virtual production spatial environments gen AI and more to engage audiences of the future through more um immersive interactive and embodied forms and experiences robin's 10-year background in theater has seen her work for the Royal Shakespeare Company the National Theater Studio and on international internationally touring live cinema productions in Europe since her claw leadership transform fellowship in 2021 she's collaborated with companies ranging from Epic Games Story Futures and ILM been commissioned by Screen Skills to develop specialist virtual production training jointly won a silver BEMA award for her work on an interactive um digital comedian prototype exploring the ethical implications of using Genai in the creative industries and been awarded innovate UK funding um with which to shoot a 180 immersive film testing new workflows for spatial content creation robert Robin is a regular keynote speaker and a consultant at cross- sector events so thank you very much Robin and we're really pleased to have you here and over to you very good thank you very much uh Robin needs to cut down her biog um I'm really pleased that everybody has covered all of the things that they have because each speaker has I think brought into focus really different elements of a of a conversation around use of AI to create to write um and I'm going to take yet another slightly different angle um and I'm going to try and go quickly because I'm aware of the time so I'm not going to kind of list tools that could be used uh for things like writing or development vis visualization that kind of thing um instead I'm going to talk about opportunities i'm going to do that through a case study because I think that audiences of the future while still of course engaging with story and wanting story and wanting practitioners to kind of help shape the entertainment experiences they have and of course have a sense of collective viewing i think people are wanting what guy started to talk about which is kind of greater immersion in worlds greater control over how they watch and engage and greater inter interactivity agency and even embodiment so happy to talk about things like that in the questions um and of course there's a question about the next generation whether they will be as interactive um as we think they will be as the digitally native generation alpha um but the thing that I think is going to keep drawing people whether for passive or active uh experiences is trans media storytelling and I think that is predicated on the idea that there is a world people want to keep coming back to and that there are characters they care about so the project I want to talk about um was something I worked on last year with both Hatchick and Liinal and it forms the basis of a larger um uh project that my company's taking forward this year um so the project was an innovate UK funded um sprint uh it was run through digital catapult and target 3D and we had a number of other partners including allseeing eye meaning machine Alex Horton design and others and we all collaborated across a fiveweek sprint to create a digital comedian prototype who would explore the ethical implications of harnessing generative AI in the creative industries now what was interesting about this project was that I as creative lead was of course being asked for a character a world a story a scene i was being asked for kind of creative decisions on the characters physical characteristics their movements their tone of voice their energy so on decisions about the environment in which we were going to meet them but crucially I was being asked for a script but of course a script is what was missing because the idea was that this character was going to create that script live um uh so instead I found myself writing a set of given circumstances and a character that would give rise to a story that would have the hand of the author all over it but was not written by the author and instead was kind of personalized to the viewer or the participant just like an interactive um immersive theater show might be um or an improvised standup show um so my background as you've heard um is as a director and so the creative challenge at hand here it it it seemed to me to be no different to working with actors uh whether scripted devising or improvising we figure out a character's objective or intention the actions they might play we understand the given circumstances dig into the character relationships and research into areas that we might need to know more about or think about particular themes um even think about like the stage craft of the performance to give the audience the best experience or prepare some anchors for the scene so that we do have a sense of where the story might go and we can put in some prompts that might help it um and this is you know something that we might then rehearse as well if it's not an improvised performance but something that needs to run again and again then repetition might happen in the form of rehearsal and all of this is exactly what we did with our composite character so made up of many different elements and many different artists to create uh the character that we called Kyle um who was this digital comedian um I'll talk a bit about how we did this but not in too much detail so we prepared these pieces of information that I've described as a single prompt injection that went into a pipeline that then allowed us to drive the character um again I can talk about stuff in the questions or you are welcome to contact me separately if you're interested in how that worked um and I'm glad that uh Guy is one of the speakers today because I was going to name check Charisma AI as being a company that is very good at understanding how do you put that information into the LLM because as Guy had already said it's not so simple as just putting it straight in so we worked with um a company called Mini Machine who prepared um a prompt injection interface for us so that we could plug it into the the pipeline that we were using um and what I was going to say about Charisma was you know they're fantastic at choosing which LLM you might use for which character in your story and so on um I won't say too much more about them because they can speak for themselves um but the reason um that I'm outlining this way of working during this session when we're talking about writing is that I think this is a really exciting form of writing that allows for emergent dialogue to take place between the character or characters and the and the participant and it could really revolutionize the way we think about not just TV but also games and even the relationship between TV and games and other forms beyond the two that I've just name checked um it might be useful to highlight some of the challenges um because they are the most interesting thing about any sprint like this um so of course you can't control what the character is going to say so the the prompt injection and the fine-tuning which goes on underneath that prompt injection process um it makes it much more likely that the character might say what you want them to say in other words you create the world the character the set of practical elements that will enable your story to emerge enable the dialogue that you kind of want um and there are lots of companies working on this exact idea that you want to create a sort of um a premium a premium and an authorled piece of work that allows for emergent dialogue between characters and audience that can remain intact to something that um you had in inverted commas planned um you need some mechanisms of course by which the character can learn the events that are happening during the interaction um as opposed to the facts that predate that interaction and you can potentially hear from what I'm saying here that the vocabulary that we were working with during our sprint was borrowed straight out of the the theater world so I'm using Katie Mitchell's um late Stanniskian approach of events um and looking at intentions things like that um LLMs are developing uh the ability to remember what has gone on in a chat and I'd recommend checking out Sesame AI for really interesting conversational AI um but that's certainly an area of interest and it gets more complex when you're writing a scene in which the topography and the physics of the world start to affect whether or not a character can see or hear things that are happening in that world um I'll finish here to sort of by saying that I I found the the creative challenges that were set by the technical limitations and we've heard this from other speakers today um these were infinitely freeing of my creative process um on the Kai project and I was working as a you know part of a much larger team of people many more people came together to create this one character uh which was an exciting way of working there's a lot of um emphasis um on LLM's as as research assistant as as kind of sounding board and a lot of focus on um the creation of images and videos using generative AI with with prompting but there's far less focus on this I think really exciting area of character development and what's most thrilling about it is that of all the things that I was experimenting with within the the kind of prompt injection the the dialogue instructions the emotional range of the character whether or not the character was uh being uh enabled to do an assign to an audience out of all of the things that could have made the real difference to the content of what the character said in the scene uh in its kind of real world setting the thing that made the difference was sample dialogue and that of course is the job of the writer i had to write an example of the style and tone of the character i had to create the sentences uh that the character might say and then sort of train that character on its own greatest hits gradually um because the most important thing here is the creative input you're giving the the character as author so people talk about AI itself will potentially only generate mediocrity right now but if you the writer can get your authoral hand all over the prompting and fine-tuning of that AI I think we can generate some really exciting new forms for audiences to get inspired by so I'm going to leave it there because I know the questions will be far more interesting than anything else I could say here uh so thank you very much and I hand over probably to Amy yes thank you i um yeah I really enjoyed um hearing about the challenges and particularly the characterization yeah it's given me a lot to think about um I think that's a good starting point for our first question um I'd like to ask around about um creative challenges that that any of you have faced during uh your work with AI um AI tools i think it's something that came up in in Robin Sarah and Guy's um uh presentations in particular so if you got any thoughts or any reflections on creative challenges that you came up against or you tackled you tackled or grappled with during your projects yeah it won't kill anyone uh it's a real problem because it's all been uh made so vanilla and through safeguarding that actually if you want someone like Darth Vader or any sort of strong character for a murder mystery um it won't let you do it so there are ways in which we we can get around it by uh fine-tuning our own models which we do but uh that's quite interesting you know the idea of having strong characters in a in a as part of a story um is uh is is one of the things we found so it just sort of um dulls them down and and you can't have strong characters that's Yeah it'll flag it for warnings okay okay we were working with Stephen Wolf which is a Warner Brothers character from the Justice League franchise who uh if you know your cannon is is uh work is working for dark side whose goal it is to wipe out the uh the galaxy and kept we kept having it flagged for genocide and it wouldn't um create any dark for us so we had don't want to get flagged for genocide no um it's the same with anything that's like um vaguely sexual or even using the word intimate asking it to write an intimate scene or describe something that has any level of intimacy where um that word it just reads as being sort of sexually explicit it doesn't kind of have any of the the sort of the nuance of the the kind of um the range of intimacies that humans um engage in um yes so then I I I got that classic response you know I'm a language model and I can't help with this question um which I always find very amusing i quite like it making it work so you get that response of it sometimes i mean I will echo what others have said um and talk a little about the vanilla but but also about um a particular recurrent sort of brand of humor that seems to come out of BLM at the moment um and whilst we were able to to move it somewhat away from that there was a consistent return to it so we often had to find new ways of um of fine-tuning the prompt itself the prompt injection to to deal with that um so I suppose yeah the the the vanilla and the uh repetitiveness of the content that was coming out i mean there's many more um practical challenges but they were all to do with the running of the the kind of body of the character so the movement of the character rather than looking at writing but I think this is such a a huge issue as we're hearing yeah thank you all um I'm going to move to the Q&A um questions we've been asked from the um audience um Guy has already given a wonderful answer to this one but I wanted to open it up to all of you uh could any of you talk more about ethical issues that you grapple with in your creative practice or your thinking around AI so entering the kind of thorny territory of ethics um in its broadest sense um could any of you speak around um kind of ethical issues of AI and script writing specifically i don't mind starting and and I will preface what I have to say with the fact that I probably can never cover all the things to say on this topic and one of the reasons that Limol and Hatrick got involved with the Kai project was because it explicitly was setting out to um expose everything about the process including you know problems so there were moments where the latency was awful and we wanted to expose that but it that included some of the ethical issues around the use of generative AI in the creative industries and I think what's challenging of course is that in order to get the voices of creative people people who work in arts and culture to get those voices into the space you must participate um and whilst participating you can learn things that you didn't know before about I don't know anything from this whether it's a sustainable act to be using an LLM as opposed to a search or whatever um that then make you question well should I be in this space but I think the longer that I as in the more time that I've spent experimenting with AI the more I've realized how important it is especially for artists to be in this um safe uh sandbox where we are understanding these tools and how to talk about them and what the issues are um I'll highlight quite aside from you can hear there's a philosophical question that will always arise and I think it's important to be involved so that the voice that we have um is connected to an understanding of the technology um but there are other other issues like um at the moment it's quite difficult to obviously track where artists data goes and we had um mocap artists who it's very hard once you've got movement data going into a system it's very difficult to track that so that's one of many examples of how do you um you know credit people appropriately how do you protect people's data who owns what um Kai as a character is frequently availability checked to speak at conferences like how does that happen who whose character is Kai so there are many ethical issues and I'm not going to talk about all of them i'm going to let other people step in where I've left off or reiterate other things that I've said thank you yeah do any of the other speakers have any thoughts around ethics in in its broadest sense around AI well I I would say I encountered in the way that I worked with it was quite obviously very different from the way that Guy and Robin have been working with AI working in a much more sort of traditional kind of writerly way um but uh stereotypes really I guess and sort of trying to avoid stereotypes so you know you're putting in a log line or an idea and you'll end up you know with um you know the women characters behaving in a certain way according to certain traditional stereotypes or um and I guess that comes out of the sort of comes under the category of quite generic or kind of and and one of the problems of pattern recognition isn't it is that they they are you know um producing uh outputs that um are a result of what's been put in um so uh you know data data in data out so it's um that was something you had to kind of I had to write over overrite um or write around I suppose um and uh yeah so that was kind of one of the main the main issues um that I came across in my practice I suppose biases trying to um conflate them or work against the biases that that it turned out and then your and then it also questions around how much of the AI content you know at what point in the revision and reddrafting process does the work become yours again so if you are taking ideas from AI you know um uh and issues around um authorship I guess and visibility you know so if AI does produce something that you quite like you want to use um how do you uh how do you acknowledge that um if you're going to publish it or use it in some way can I just call it mine just because I put the prompt in or what should I do about that i think that's still quite um uh quite a gray area um yeah so how how do you kind of credit AI's input in work that you want to publish thank you that's great um and I see in the in the Q&A guy has um signposted to a paper which I think is going to be really useful for this um I'm going to have to wrap up because it um is the end of our hour together which is a shame because I know we could have kept talking for for much longer but I'd like to um thank our four panelists it's been great spending time with you and thank you very much for your diverse um experiences and um background on this topic because I think that's that's what we need to hear from lots of different people um so thank you very much thank you for everybody attending it's been great to have you all um hopefully we gave you lots to think about and hopefully the webinar series may return at some point and we will um have time to speak again thanks very much everyone thank you yeah I know you've got a race off guy as well we're just waiting for the We've got a I've got a go to Amy but we'll catch up soon yeah yeah okay bye yeah bye great just got a couple more um participants in the room still and then just checking okay I think it's just us now all the attendees have have gone um yeah thank you very much uh both of you for for speaking it's been really great to hear what you've what you have to say yeah well thanks for the invite it was um it was nice to find Guy on the call as well or the webinar yeah yeah sorry that that was wasn't clear but I think No no it was lovely it's just funny to us that we end up you know in the same spaces a lot of course i bet I bet you're always in the same in the same room i'm in a meeting with him this afternoon you know he's all over all over my calendar you really are connected to Guy um I think what you said it worked really well it kind of connected well it's totally different to the academic approach

2025-04-15 10:57

Other news