Regarding the Oct. 31 news article “President’s executive order on AI marks ambitious regulatory attempt”:
Many international leaders believe data used by artificial intelligence must be regulated. I agree, but I think there are a few things we must consider.
First, some people think of AI as highly advanced, sentient robots that will eventually turn on us. This misconception is common because of social influences, including the Terminator movies. But the modern “AI” is just code that can take an input, then output something that you want. For example, researchers at DeepMind trained an AI model on microscopic protein structures, and it can now output the 3D structure of a protein, a game changer in biomedical science. To have functional regulations, both the input and output must be considered.
President Biden’s executive action primarily focuses on using unbiased data and approved sources as an input to train our AI, but it doesn’t fully consider the output, the uses of AI. Some AI models being used are unknowingly malicious, such as predictive policing, or intentionally malicious, such as spreading misinformation. We should regulate the output freedom of AI models by including human oversight to maintain productivity and increase accountability.
Anurag Perakalapudi, Fairfax
