I believe Google’s AI Principles are admirable but ultimately aspirational. So many of these systems are based in unethical and oftentimes inaccurate sources, such as written text or art that has been published on the internet, but not given consent to have that work used in the generator’s learning algorithms. Additionally, while the principles generally have the right idea to create a more ethically and accurate AI system, it does not address the environmental/physical impact of the systems. One huge issue with AI systems is the massive amount of energy consumption each generation prompt and query uses. The average Google search pre-AI used 10x less energy than what it currently uses when the AI answer function was added in (https://www.reddit.com/r/technology/comments/1dsvefb/googles_ai_search_summaries_use_10x_more_energy/). This also isn’t just an issue with Google, ChatGPT uses 15x more energy than a Google search query pre-AI (https://www.reddit.com/r/aipromptprogramming/comments/1212kmm/according_to_chatgpt_a_single_gpt_query_consumes/).
Google’s AI function has also been shown to give inaccurate information, from simple misinformation about program procedures (i.e. finding a function in a digital program) to incredibly dangerous misinformation regarding such topics as conspiracy theories (i.e. “Barack Obama is Muslim” - a completely facetious theory). Google AI overview function just is not developed enough, nor ethically sound, to be implemented as deeply as it has been so far.
No comments:
Post a Comment