“What if artificial intelligence becomes autonomous beings?”

When and how did your project start?

During the 2020 lockdown, my friends and I started a virtual book club. The pandemic renewed our sense of urgency about climate change and led to many readings and conversations on how we might adapt as individuals and “a people.” The implications of these writings altered my faith in techno-solutionism and drew me to investigate the role of technologies in climate change. During my research, I read about how tech industry “thought-leaders” routinely listed artificial intelligence as a more severe threat to humanity than Climate Change. These statements seemed comical given the current state of general AI versus climate researchers’ dire projections.

What if artificial general intelligence becomes autonomous beings? Though “AI” manifests itself as software machine learning applications, it depends on a ubiquitous hardware infrastructure that requires constant human maintenance and temperature control. In order to reach a general level of intelligence, an intelligent system will need to understand how it’s situated in the environment. I wanted to speculate about a future where an AI agent has an experience of embodiment and tries to negotiate its existence with the surrounding ecosystems.

What is the profile of the project collaborators?
I’m a programmer and technologist born and raised in Qingdao, China, currently based in Brooklyn, New York. This is primarily a solo project, but throughout the process, I got a lot of feedback and support from my friends and community here in New York. My pandemic book club friends are working on projects in similar domains, and I hope to collaborate with them as an extension of my work soon. 

Which stage is the project in?
I wrote a few scripts that collected 270k+ climate-related articles over a few days, then trained text/image generation models based on them. I’m now working on a series of content guided by my curiosity, where I prompt topics, phrases, and keywords as inputs for the AI to produce content. The content made by this speculative AI will be displayed as a video screening and installation at the October event. 

In this phase of your research, what are the two most important aspects of the project?

1. What interests me the most about the idea of artificial general intelligence is its ability to question and demonstrate what human intelligence means to us. Machine learning is the best data visualizer of the Anthropocene, as it shows us what we chose to write down and amplify as a collective. By collecting and generating data, the scripts begin to reveal a baseline for “Climate Change” in human terms. 

2. I’m fascinated by the hardware infrastructure that supports the software work I do on a daily basis. Where are these infrastructures located? How do they operate? I feel removed from this whole process as a software engineer. This project provides an excellent opportunity to poke into this and see my day-to-day work from a different lens. 

What are the next steps in the research?
Going deeper into the embodiment concept, I want to speculate more on how this AI agent might reconcile its distributed existence on the hardware level with its objective to compute and optimize. What kind of choices would it make? What would it prioritize? There is so much to take into account from this AI’s perspective. I want to investigate these potential narratives and see where it takes me. 

How can AI help to solve the climate crisis?
The tech industry is notoriously at defending its expansionist mindset with technological solutions. When it comes to combating climate change, however, the industry’s characteristic systematic optimism and relentless willingness to fail might help the cause. I’ve seen many encouraging signs of applied machine learning in crisis prevention, energy efficiency, and biodiversity studies. The current form of machine learning is extremely good at data analysis, so it can help us navigate climate change when large amounts of data are involved. 
That said, disaster intervention, biodiversity, and climate predictions are situations that crave nuanced solutions, but machine learning tends to make generalizations that can amplify some of our worst mistakes. It’s crucial to have experts in the loop to ensure these systems do more good than harm.