Republican state Rep. Alex Kolodin said he used ChatGPT to write a subsection of House Bill 2394, which tackles AI-related impersonations of people by allowing Arizona residents to legally assert they are not featured in deepfake videos.
“I used it to write the part of the bill that had to do with defining what a deepfake was,” Kolodin told NBC News. “I was really struggling with the technical aspects of how to define what a deepfake was,” he said. “So I thought to myself, ‘Well, why not ask the subject matter expert, ChatGPT?’”
The bill was signed into law by Democratic Gov. Katie Hobbs on Tuesday. The legislation allows Arizona residents to obtain a court order stating the person identified in the deepfake video is not them.
Kolodin said that the portions ChatGPT created were precise.
“In fact, the portion of the bill that ChatGPT wrote was probably one of the least amended portions,” he said.
Hobbs was not aware of the portion of the legislation being authored by ChatGPT.
“I kind of wanted it to be a surprise once the bill got signed,” Kolodin said, noting that it was part of the plan.
The problem is GTP is learning from our existing knowledge base. If legislation is trying to amend a broken system, we don’t want AI to be modeling that system. This case seems fairly harmless, an AI takeover isn’t what we should be worries about.
Something like institutional racism being replicated in a more insidious manner is the concern. Relying on these closed systems potentially gives the type of people who implemented the discrimination being modeled to turn around and say, ‘See, we were right all along!’ If results are held up on a pedestal and AI is integrated into our political and legal systems, it may make changing society for the better much harder.
We shouldn’t universally condemn tools like ChatGTP being used in this way, but we should tread very carefully when it comes to large scale societal changes.