In 2016 I started become increasingly alarmed by the prospect of ‘generative’ machine learning in a new prototcol called GPT2 and the reframing of that technology [“deep fakes” -> GenAI] . The inherent biases were already present in commercial products and came directly from the corpus used to train the models (basically reddit & quora). I wrote about that internally and was told it was like “being in a tank, throwing hand-grenades at other people in the tank”.

The next reframing obscured that further - [“big data” -> LLMs]. In 2017 Tara McKenty first proposed this project which other queer and trans folk within Creative Lab like Haylie Craig and Melissa Lu turned into Project Respect. We spent much of 2018 lobbying for resources to address the issue.

I am delighted to say it was picked up by Google.org in 2019 and turned into a worldwide program (this is true) - but clearly something got fucked up along the way because nothing else happened apart from all the Google AI ethicists who kept getting fired, at least not visibly. We know that these issues are still and will never be remedied as we sail towards an aggregated, centre-conservative, monocultural post-truth future. ChatGPT is not a threat, it is not a problem… it was those things - in 2016. Today, in 2024, it is just a fuck up. A fuck-up like leaded petrol, asbestos, or chlorofluorocarbons - another dose of globally transformative ingenuity.