356-winter-2023/final
Contents
Final Project: "Escape from the Turing Trap (or at least make a start)"
Music and AI (Music356/CS470) | Winter 2023 | by Ge Wang
First of all, thanks to Graciela for commenting on the idea of the Turing Trap in her Reading Response #2:
One idea discussed in the talk was what Dr. (Pattie) Maes called the Turing Trap: the notion that the Turing Test had fundamentally sent AI researchers down the wrong path since the very beginning of machine learning. She suggested that the Turing Test established the goal of AI research to create machines that outperformed humans. Thus, researchers are perpetually taking tasks that humans are good at (chess, driving, making art) and making models that meet and surpass our own performance. By following this path, we find ourselves optimizing against our own interests as we build systems that fundamentally aim to replace humans rather than empower humans. Instead, she proposed that we try to optimize for what people + AI would be able to accomplish rather than just AI alone. —Graciela, "Power to the People" reading response.
Questioning the Premise
The idea of creating machines that are indistinguishable from humans can be traced back to long before Artificial Intelligence (think Mary Shelley's 1818 Frankenstein, a fiction about an artificial and intelligent creature; recall the original Mechanical Turk--a different kind of "fiction": a human chess-master hidden within the supposed chess-playing automaton). While there is potential value in exploring this all-too-human desire for AI to mirror and outperform humans, it has so captured our cultural and technical imagination as the de facto benchmark of progress, that it is almost impossible to think of AI has having any other overarching aims. The presumed goals are either "do what humans do, and do it even better" or its commerce-friendly version: the unrelenting march towards "convenience" (and the monetization thereof). In fact, try to think of any other overarching aims of AI. Go ahead. I'll wait.
The Turing Trap is called a "trap" because of its seductive power. It has long become the holy grail, something that is rarely questioned but pursued with fervor. How much is the premise of AI research questioned in other AI engineering contexts like courses, tech companies, product meetings (a lot? a little? not at all?)? While there ought to be room to explore human-like machines, the peril is that its tenets have largely come to dominate the conversation, impoverish the imagination, making uncritical nerds of us all in a staggering groupthink regarding the most powerful technology of our time. And like the original Crusades, it has the power to cause great human suffering. Sheeeeeeeet.
Making Things to Think With
In our critical making course, we have always asked a fundamental question, "what do we (really) want from artificial intelligence?". And instead of focusing on, say, "how do we make AI more clever or more powerful", the goal of this course has always been to explore questions like "how do we want to live with AI? And through what we design with AI, how do we want to live with one another?" In the spirit of this critical making course on Music and AI, this final project asks you to reflect on these questions and design/extend a system to explore the idea of escaping from the Turing Trap. The truth is, of course, all of you have been doing this all along in this course, from your display of vast human creativity (especially as you grappled with limited AI tools) in poetry tools, beautiful audio mosaics, and playful wekinations of the world around you.
One final project (or one course) is not going to turn the tide on the Turing Trap. But here and now is, in general, as good of a place and time to start as any. Remember, as always, these don't have to be useful; in fact, whimsical is good! Absurd is good! Playful is wonderful!
Due Dates
- Milestone (webpage + in-class presentation): Tuesday (3/14)
- Final Deliverable (polished webpage): polished webpage due Wednesday (3/22, 11:59pm)
- Final Presentation (in-class presentation): Thursday (3/23) 3:30pm CCRMA Stage
Tools to Play With
- anything you want to use, including all the tools so far (ChucK/ChAI, Wekinator, word2vec, mosaic)
A Few Bad Ideas!
- Using camera input and sound output, create a day / night classifier (or sonifier)—I mean, who has time to look out the window?
- An instrument using one or more continuous input (such as mouse, gamepad, gametrak, Kinect, wiimote, VR controller with or without the VR) controlling a multi-parameter synthesis (like one of the STK instruments like Bowed)—try playing this for your roommate, or your cat (and film the performance and audience reception)
- using Wekinator's dynamic time warping, train a few audio or other gestures, and create a voice or gesture command system that does meaningless things—inject a little AI-mediated chaos into your day
- Use Wekinator to train an instrument for controlling/exploring your audio mosaics (the output would be an N-dimensional/dimensionally-reduced vector)
- for more ideas, check out these projects (some of these are large-scale projects; for this programming etude, remember to keep the scope of your Wekinate creations small)!
One Final Reflection
- write ~300 words of reflection on your project. It can be about your process or the products. Tell us about your attempt to deploy them.
Final Deliverables
- create a CCRMA webpage for this etude
- the URL should live at https://ccrma.stanford.edu/~YOURUSERID/356/final or https://ccrma.stanford.edu/~YOURUSERID/470/final
- alternately, you may use Medium or another publishing platform (but please still link to that page from your CCRMA webpage)
- your webpage is to include
- a title and description of what you made (free free to link to this wiki page)
- video recordings corresponding to your system in action
- all relevant code and data files
- brief report on how you created these
- your 300-word reflection
- any acknowledgements (people, code, or other things that helped you through this)
- submit to Canvas only your webpage URL
