![\"\"](\"https:\/\/world.edu\/wp-content\/uploads\/2024\/09\/AIrisks.jpg\")
<\/p>\n<\/p>\n
The Australian government this week released\u00a0voluntary artificial intelligence (AI) safety standards<\/a>, alongside a\u00a0proposals paper<\/a>\u00a0calling for greater regulation of the use of the fast-growing technology in high-risk situations.<\/p>\n The\u00a0take-home message<\/a>\u00a0from federal Minister for Industry and Science, Ed Husic, was:<\/p>\n We need more people to use AI and to do that we need to build trust.<\/em><\/p><\/blockquote>\n But why exactly do people need to trust this technology? And why exactly do more people need to use it?<\/p>\n AI systems are trained on incomprehensibly large data sets using advanced mathematics most people don\u2019t understand. They produce results we have no way of verifying. Even flagship, state-of-the-art systems produce output riddled with errors.<\/p>\n ChatGPT appears to be growing less accurate over time<\/a>. Even at its best it\u00a0can\u2019t tell you what letters<\/a>\u00a0are in the word \u201cstrawberry\u201d. Meanwhile,\u00a0Google\u2019s Gemini chatbot has recommended putting glue on pizza<\/a>, among other comical failures.<\/p>\n Given all this, public distrust of AI seems entirely reasonable. The case for using more of it seems quite weak \u2013 and also potentially dangerous.<\/p>\n Much has been made of the\u00a0\u201cexistential threat\u201d of AI<\/a>, and how it will lead to job losses. The harms AI presents range from the overt \u2013 such as\u00a0autonomous vehicles that hit pedestrians<\/a>\u00a0\u2013 to the more subtle, such as\u00a0AI recruitment systems that demonstrate bias against women<\/a>\u00a0or AI legal system\u00a0tools with a bias against people of colour<\/a>.<\/p>\n Other harms include fraud from deepfakes of\u00a0coworkers<\/a>\u00a0and of\u00a0loved ones<\/a>.<\/p>\n Never mind that the\u00a0federal government\u2019s own recent reporting<\/a>\u00a0showed humans are more effective, efficient and productive than AI.<\/p>\n But if\u00a0all you have is a hammer<\/a>, everything looks like a nail.<\/p>\n Technology adoption still falls into this familiar trope.\u00a0AI is not always the best tool for the job<\/a>. But when faced with an exciting new technology, we often use it without considering if we should.<\/p>\n Instead of encouraging more people to use AI, we should all learn what is a good, and not good, use of AI.<\/p>\n Just what does the Australian government get from more people using AI?<\/p>\n One of the largest risks is the\u00a0leaking of private data<\/a>. These tools are collecting our private information, our intellectual property and our thoughts on a scale we have never before seen.<\/p>\n Much of this data, in the case of ChatGPT, Google Gemini, Otter.ai and other AI models, is not processed onshore in Australia.<\/p>\n These companies preach\u00a0transparency, privacy and security<\/a>. But it is often hard to uncover if\u00a0your data is used<\/a>\u00a0for training their newer models,\u00a0how they secure it<\/a>, or what other organisations or governments have access to that data.<\/p>\nAI risks<\/strong><\/h4>\n
Is it the technology we need to trust \u2013 or the government?<\/strong><\/h4>\n