AI Boosterism in AI Training

I had some spare time so decided to get on with the AI training the Australian National University is encouraging staff to take. This is the usual online training module with some quizzes and exercises. It wasn't too bad, but as I know a bit about the topic, I was up to part 5 of 8 before it started telling me anything I didn't know. 

 At this point I was instructed to get AI to generate a LinkedIn post about AI & personal learning. While the previous modules did have some warnings about checking AI output for incorrect information and hallucinations, it was all a little too positive. At this point I used Copilot to generate the required LinkedIn post. This had only positive points on how AI could help with personal learning. As instructed I refined the result, but on my own initiative by asking for negative as well as positive points. That worked fine, but what struck me was that had I not asked, it would have all been positive.

Here are the negatives added:

"Negative impacts

• Over‑reliance on AI, which can weaken critical thinking and problem‑solving if learners default to AI-generated answers.

• Quality inconsistency, as AI can produce confident but incorrect explanations that mislead learners.

• Equity gaps, where those with better access to technology gain an advantage over those without it.

• Privacy concerns, especially when learning data is used to train models or inform performance decisions.

• Reduced originality, as learners may unintentionally mimic AI-generated patterns rather than developing their own voice or ideas.

These risks highlight the need for thoughtful integration rather than blind adoption."

Generative AI For Higher Education

Greetings from the Australian National University where I am taking part in a Generative AI Workshop Day. Professor Adam Bridgeman and Professor Danny Liu from the University of Sydney are taking ANU staff through how to Microsoft Copilot to create rubrics, multiple and choice questions. The also demonstrated using other AI to create a tutor chatbot to help e with the topic and even give feedback on their writing. One point emphasised is that experts have to test what the AI is producing. In February I attended a " Team-Based Learning Collaborative Asia Pacific Community Symposium" at USyd and was impressed with what they are doing. The state of the art has advanced since 2018, when I wrote a tutor bot to handled applications for assignment extensions: it always said "no". ;-)

As an exercise I had Copilot write a summary the ANU VC's PhD thesis, in HTML:

Telling Stories Out of School: Remembering the Carlisle Indian Industrial School, 1879-1918

Overview: "Telling Stories Out of School" examines the Carlisle Indian Industrial School, the first government-funded, co-educational, secular, non-reservation Indian School in the United States.

Key Points:

• Historical Context: Carlisle operated from 1879 to 1918 and influenced other non-reservation boarding schools.
• Goals: The study explores state policies, student experiences, and post-school impact.
• Student Population: Approximately 8,500 students from 75 Native American Nations attended Carlisle, learning academic and vocational skills while navigating their identities.

Australian Government Trial of Generative AI for Law, Education, Health, and Aged Care

The Digital Transformation Agency has announced 50 Australian Public Service (APS) agencies are conducting a 6-month trial of Copilot (Microsoft's implementation of Chat GPT), ending June 2024. Staff first take learning module. Agencies have also been given some general guidance, on Accountability, Transparency, Explainability, Privacy, Fairness, and Wellbeing. Agencies where AI could be controversial include Attorney-General's, Education, Health and Aged Care, Home Affairs, and the National Disability Insurance Agency.

A previous Australian Government got into difficulties with relatively simple technology in what was known as the "Robodebt Scheme", which resulted in several deaths. Careless application of AI has the potential to cause human misery and casualties, on a much larger scale. It is to be hoped the APS applies the technology with thought, so there is no need for a Robo 2.0 Royal Commission.

ps: As it happens I will be running ANU computer project students though some Generative AI exercises using Copilot, over the next few weeks. Some of the students are studying programming AI in depth, but this is more general on how to use it.

Australian Universities Accord Final Report: According to Generative AI

For the fun of it I asked Microsoft Copilot to summarize and critique the Australian Universities Accord Final Report. Here is what it came up with. ;-)

Summary:

The Australian Universities Accord Final Report, released by the Hon Jason Clare MP, Minister for Education, contains 47 recommendations aimed at reforming higher education to meet Australia’s future skills needs. Here are the key points:

1. Reform Goals: The report sets a vision for the next decade and beyond. It aims to enhance the higher education sector’s effectiveness and responsiveness.

2. Workforce Qualifications: The report recommends that at least 80% of the workforce should have either a vocational education and training (VET) or university qualification by 2050. Currently, this figure stands at 60%1.

3. Long-Term Plan: The Accord Final Report provides a long-term reform plan to address the evolving demands of Australia’s workforce and economy.

For more detailed insights, you can refer to the Summary Report prepared by the Department of Education2. It outlines the themes and issues identified in the full report.

Remember, this is just a concise overview, and I recommend exploring the full report for comprehensive information.

Critique:

Let’s delve into a concise critique of the Australian Universities Accord Final Report. Here are some key points to consider:

1. Ambitious Goals: The report sets ambitious targets, such as aiming for 80% of the workforce to have either a vocational education and training (VET) or university qualification by 2050. While this goal is commendable, achieving it will require substantial effort and investment.

2. Implementation Challenges: The report outlines several recommendations, but translating them into effective policies and practices may face challenges. Implementation across diverse universities and regions could be complex.

3. Equity and Access: While the report emphasizes workforce qualifications, it’s essential to ensure equitable access to higher education. Addressing barriers faced by disadvantaged groups, including Indigenous students and those from low socioeconomic backgrounds, is crucial.

4. Funding and Resources: The report doesn’t delve deeply into funding mechanisms. Sustainable funding models are essential to support quality education, research, and infrastructure.

5. Industry Collaboration: The report highlights the need for stronger collaboration between universities and industries. However, striking the right balance without compromising academic autonomy is critical.

6. Digital Transformation: The report acknowledges the importance of digital skills. Universities must adapt swiftly to technological advancements and prepare graduates for a rapidly changing job market.

7. Research and Innovation: While workforce qualifications are vital, universities’ role in research, innovation, and knowledge creation should not be overshadowed. Balancing practical skills with intellectual growth is essential.

In summary, the Accord Final Report provides a roadmap, but its success hinges on effective execution, stakeholder engagement, and ongoing evaluation. Policymakers, educators, and industry leaders must collaborate to turn these recommendations into meaningful outcomes for Australia’s higher education sector.