High Blood Pressure Control and Decreased Risk
July 9, 2024Ambient Neural Language Model
July 16, 2024Dr. Gloria Wu disscusses AI-mediated chatbots, referencing a New York Times article from April 8, 2024. The article, authored by Cade Metz, Cecilia Kang, Shera Frankel, Stuart A. Thompson, and Niko Grant, highlights how tech companies like OpenAI, Google, and Meta navigate legal and ethical gray areas to harvest data for their AI models. In 2021, OpenAI transcribed over a million hours of YouTube videos to gather conversational text, despite potential privacy rule violations. This practice underscores the ongoing struggle to obtain enough data to train larger AI models, leading companies to increasingly bend corporate policies and laws.
As data sources dwindle, companies turn to synthetic data, where AI generates its own training data. This shift raises concerns about AI models perpetuating their own biases and errors. Additionally, there are legal challenges from entities like the New York Times, suing AI companies for using copyrighted content without permission. The article also mentions researchers like Jared Kaplan, who emphasized the need for vast data sets to improve AI performance. This endless quest for data, including Google’s training of AI on trillions of tokens, illustrates the industry’s race to advance technology. Users should be aware of these issues and verify information from multiple sources for accuracy.
Check out Dr. Gloria Wu’s publications on ChatGPT, LLM, and Internet of Things:
Wu G, Lee DA, Zhao W, Wong A, Sidhu S. “ChatGPT: Is it Good for Our Glaucoma Patients?” Frontiers in Ophthalmology, section Glaucoma. 2023. https://doi.org/10.3389/fopht.2023.1260415.
Wu, G., Zhao, W., Wong, A., & Lee, D. A. Patients with floaters: Answers from virtual assistants and large language models. Digital Health. 2024. https://doi.org/10.1177/20552076241229933
Wu G, Satheesha-Krishna S, Sakai S, Jhangiani R, Kurniawan, S. Google Assistant and ChatGPT: Is it useful for non-medical professionals looking for information about stroke or glaucoma. IOT 2023: The International Conference on the Internet of Things, Nagoya, Japan, November 2023.
https://doi.org/10.1145/3627050.3631579