Puppets of a Digital Brain
Puppets of a Digital Brain is (will be soon!) a short film that explains the technology behind current generative AI systems, in particular Large Language Models (LLMs) in an accessible and attractive way, and explores and explains the good and bad sides of current AI technologies. The goal is to make this information accessible to the wider public, such that they are better informed to hopefully make more informed choices when using such systems, and shape the development of future AI systems that respect ethical principles in development, training and usage.
Synopsis
Driven by a noble desire to help scientists keep up with the never-ending tidal wave of research papers, Dr. Mal, a brilliant but socially oblivious researcher, sets out to build a language model. His creation, though trained using a rather simple idea, develops surprising capabilities. Before dr. Mal can throw on the digital seatbelts -- filters, guardrails, and safeguards -- the model slips into the public sphere.
Convinced by investors with dreams of world domination, and his own dreams of fame, dr. Mal tweaks the system to seem more human. It talks like us. It seems to think like us. It even refers to itself as I. The public eats it up. It spreads like wildfire, powerful, persuasive, and dangerously unfiltered. The dark sides of the system -- its unethical training, exploitation of severely underpaid workers, the effect on the compute requirements on the environment -- are hidden from the end users.
As the model becomes part of daily life, the consequences start rolling in, some inspiring, others deeply troubling. The system, after all, can’t be held responsible. The burden falls on those behind it: the idealistic scientist who skipped the ethical fine print, the financiers who pushed for their own goals, and the users who bend the tool for harm, deliberately or from ignorance.
In a world transformed by his own invention, Dr. Mal undergoes a transformation of his own. Once detached and naive, he becomes a vocal advocate for digital literacy, urging society to use these powerful tools wisely.
The film tracks this dramatic journey, reminding us that when it comes to AI, the conversation is not just technical, it’s human. And we all have a part to play.
"Trailer"
(The film will be made with puppets, whose "skin" is a collage of papers that reflect the character's personality and hobbies)
Dr. MAL: I swear, every time I read one paper, it sprouts five more like a hydra.
Dr. MAL (presenting a proposal to the commitee): We stand at the threshold of a transformative breakthrough! For the first time, we can build a system that will navigate the overwhelming expanse of scientific literature better than our brightest minds! It will find patterns, connections and insights that would otherwise remain forever hidden, accelerating discovery across disciplines! The question is not IF this will happen, but WHO will lead it! Esteemed committee, this is our chance to lead, to define how this technology shapes the future of research and the trajectory of human knowledge!
Dr. VAL(a bit surprised): Victor! Wow! You scared me a bit... that was terrifyingly compelling!
Dr. MAL (sure of himself) Just suggest they get the chance to shape the future. Works every time.
Dr. GRANT: Dr. Mal! This has the potential to be BIG! but it is your responsibility to make this a big GOOD, and not a big BAD! Don't forget this!
Dr. Mal is watching Pinocchio on TV:
GEPPETTO: Birba d'un figliuolo! Non sei ancora finito di fare e già cominci a mancare di rispetto a tuo padre! male, ragazzo mio, male!
Dr. MAL: Oh GEPPETTO... you just gave PINOCCHIO legs and he used them to kick you and then run away...
GEPPETTO: Me lo merito!... Dovevo pensarci prima! Ormai è tardi!
Dr. MAL (laughing) Next time think BEFORE you build a runaway puppet...
Dr. JORDAN (coughing): What happened here?
Dr. SANJANA (coughing): Looks like Victor's project took a violent turn...
Dr. JORDAN (coughing): Where is he? Is he hurt somewhere?
Dr. VAL (coughing): I know where he is... Come!
PhD student LEE (eager for knowledge): How will you train it, Dr. Mal?
Dr. MAL: Ah! with a brilliantly simple idea! I will give my system text fragments and train it to predict the next word. It sounds simple, but to guess the next word, it has to learn many different kinds of patterns.
Dr. VAL (worried): Have you been using texts from the web to train your system, Victor?
Dr. MAL (sure of himself): Obviously. It’s a goldmine! All free!
Dr. JORDAN: Victor! Big news! Your system came up with a new solution to the bin packing problem. Lots of real-world applications. Big impact.
Dr. MAL: My system is doing maths?
Dr. JORDAN: It basically solves 3D Tetris with many more shapes! How cool is that?
Dr. MAL: I didn't expect that just from processing text it can learn patterns to solve math problems!
Dr. SANJANA: Victor! We trained your system with data on chemical compounds interactions, and voila! dual-target drug discovery!
Dr. JURGEN (over her): Still needs real-world testing, but early results are very promising.
Dr. MAL (angry, talking to the phone from which he is reading an article): My system doesn't just "parrot language"! Okay, sure... I don’t know exactly how it works... But does it really matter? Have you seen the output? It freaking works! We don't need planes to flap their wings! "it’s not connected to the real world” It doesn’t need to be! The real world is overrated.
Dr. MAL (starting calm, and getting excited): Look, the system wasn’t meant to write marriage vows optimized for tiktok, or cheat on college essays. I built it to help scientists cut through the flood of papers, parse data, fight disease. It was supposed to amplify brilliance.
TV PRESENTER: Amplify brilliance, or shortcut responsibility?
Dr. MAL (frustrated): It’s a tool. Tools reflect their users. A microscope doesn’t care if you cure cancer or invent a better bomb. And yeah... I got swept up, too. The system worked better than I’d dreamed, and I—I let it do my job for me. I gave my student feedback... AI-generated. Didn’t double-check a word. I failed her.
Mr. MURG (at the launch of a new system): Good morning! We're thrilled that millions of you are with us, in person or online! The energy is awesome! Can you feel it? Yeah!!! Today is an exciting day, because we are here to launch our new system: Vill...
Motivation
Artificial Intelligence (AI) is the topic of important debates, and the cause of major changes in our working environment and society. These debates are conducted at a very abstract level, with a focus on the evolution of AI systems, while the technology behind AI systems is disregarded, it is not questioned, and does not inform the discussions. The fact that these systems are called "intelligent" leads the public to make assumptions based on their own understanding of the term and their experiences.
These assumptions are dangerous, because they influence users to endow these systems with capabilities they do not (yet) possess, and as a result, to trust them implicitly. This attitude is not justified and, more importantly, can have negative impacts from a small scale – using erroneous information in a homework or report – to a very large one – influencing the financial, juridical and social decisions. These are relevant and urgent problems that can, and already do, affect many lives at present, unlike the abstract discussion about reaching AI singularity -- when AI systems become more intelligent than people – at some hypothetical point in the future.
Our project aims to bring to the public information about the technology behind generative AI systems, like chatGPT, in an attractive and easily understandable way, through a short film. It is very important that people understand this technology, the ethical, moral and ecological implications of building and using them, their potential for appropriate but also missuse, such that they can take informed decisions about whether, when and how they interact with these systems, and how much trust they should place in their outputs.
Selective bibliography
Making this film takes advantage of the more than 20 years of first-hand experience in the fields of natural language processing and machine learning of our scientific consultant, and uses as inspiration numerous published interviews on AI of prominent researchers in the field, and on non-fiction publications from respected researchers in the field of natural language processing, machine learning and AI, and journalists.
Books
Stuart Russell: Human Compatible
Emily Bender and Alex Hannah: The AI Con
Karen Hao: Empire of AI
Interviews
Yoshua Bengio: The Catastrophic Risks of AI -- and a safer path
Munk debate on AI: Yoshua Bengio & Max Tegmark vs. Melanie Mitchell & Yann LeCun
Geoffrey Hinton: 60 minutes interview
The team
SciFilmIt is collaborating with Persona Association from Cluj-Napoca to make this short film. Here is the enthusiastic team that is working hard to make this film happen:
Project director and scientific consultant: Vivi Nastase (researcher in Natural Language Processing and Machine Learning, with more than 20 years experience)
Producer/Director: Rareș H. Stoica
Director of communication and production, screenplay assistant: Silvia Hamza
Screenplay: Vivi Nastase, Rareș H. Stoica, Asis. Silvia Hamza
Scenography: Cătălin Mocan, Asis./ Graphic Designer, Karin Vanessa Biriș
DoP: Andrei Giurgiuman, Rareș H. Stoica, 1Ad. Vlad Șandor
Music: Cristian Mihai Popescu, Asis. Karin Vanessa Biriș
Executive producer: Asociația Persona (Cluj-Napoca)
Associate producer: SciFilmit (Geneva, Switzerland)
Production company: Persona Media Production (Cluj-Napoca, Romania)
Distribution: Festivals, educational events, media platforms