联邦政府今年7月发布的报告评估指出,会计、文员等认知型工作面临最高的人工智能自动化风险,而技工、护理等体力劳动岗位受影响程度最低 1。尽管截至今年2月,人工智能尚未对澳大利亚整体就业市场产生显著影响,但澳联储近期发现,年轻工人在那些高度暴露于人工智能技术的行业中正遭遇求职困境 1。这一趋势在部分行业已现端倪:律师事务所开始削减初级律师职位并将低阶任务交由人工智能处理,Qantas航空公司则宣布将850个工作岗位外包至印度,同时加大了对人工智能技术的投资 1。
面对技术能力的快速跃升,高等教育体系亟需调整应对策略。研究显示,从2023年到2025年,人工智能系统在澳大利亚大学刑法考试中的平均得分由52.5%提升至76.3%,且2025年的模型在批判性思维能力上平均超过普通大学生 1。此外,OpenAI的一个模型最近推翻了自1946年以来几何学领域的一项长期猜想,进一步证明了其在复杂推理任务上的潜力 1。然而,人类与机器协作并非易事,2026年的一项试点研究发现,医护人员接受了半数故意设置错误的AI建议,并忽略了约六分之一的正确建议 1。因此,文章作者Armin Ashenberg建议,大学不应仅停留在传授学科知识或简单技能替代上,而应重点培养学生识别AI错误及优化其输出的能力,以实现高效的人机协作 1。
The federal government’s July report assessed that cognitive roles, such as accounting and clerical work, face the highest risk of automation by artificial intelligence in Australia, while physical labor jobs like trades and nursing are considered at lowest risk 1. Although a government assessment stated as of February this year that AI had not yet significantly impacted the Australian job market, the Reserve Bank of Australia has recently noted difficulties for young workers seeking employment in sectors with high AI exposure 1. Specific industry shifts include law firms reducing entry-level attorney positions and delegating lower-tier tasks to AI systems, alongside Qantas announcing the outsourcing of 850 jobs to India while increasing its investment in AI technology 1.
Academic performance data indicates that AI’s capabilities are advancing rapidly relative to human students. Between 2023 and 2025, the average score of AI models on an Australian university criminal law exam rose from 52.5% to 76.3%, and the 2025 models surpassed the average critical thinking ability of typical university students 1. Furthermore, a model developed by OpenAI recently disproved a long-standing conjecture in geometry that had persisted since 1946 1.
In response to these developments, experts suggest universities must shift focus from merely teaching disciplinary knowledge or simple skill substitution to cultivating students' ability to collaborate efficiently with AI 1. This includes training graduates to identify errors in AI outputs and optimize their use 1. However, effective collaboration remains challenging; a pilot study conducted in 2026 revealed that healthcare professionals accepted half of deliberately incorrect AI recommendations and ignored approximately one-sixth of correct ones, highlighting the complexity of integrating AI into professional workflows 1.
评论
还没有评论,欢迎留下第一条。