Auckland Uni's Controversial AI Move

Kia ora! Welcome to New Zealand’s weekly AI news round-up.

As we head into an AI agent (a type of AI that can work with limited human supervision to accomplish tasks and solve problems) world, we're witnessing systems operating with increasing autonomy.

• In a fascinating use case, streamers have set up Claude AI to play Pokémon Red live, demonstrating how AI can operate without human intervention.

• In a development that's frightening and fascinating at the same time, Meta engineers unveiled two AI chatbots chatting to one another in English, then switching to “GibberLink mode”, their own non-human language, for "more efficient communication". This breakthrough highlights both the potential and risks of agentic AI—systems not only acting independently but also developing communication protocols beyond human understanding.

Enjoy this week’s read ✌️

Highlights from this week’s edition:

• 🎓 Auckland Uni AI Tutor Debate

• 🌋 AI To Predict Volcanic Eruptions

• 💬 Auckland Council’s AI Assistant

• 🥝 Kiwi Haka Co-Written by AI

• 🐄 AI Innovation In Dairy

• 🎭 AI Recreates Gabby Petito’s Voice

• 📚 AI Terms Simplified for Everyone

EDUCATION X AI

🎓 Auckland Uni's AI Tutor Experiment Sparks Backlash

📣 Word On The Street: NZ's largest university is rolling out three AI tutors instead of lecturers in a marketing course, causing student uproar over educational quality while administrators defend it as vital preparation for tech-driven careers.

🔍 Zooming In:

• Starting Monday, three AI tutors will supplement weekly in-person tutorials in the Digital Marketing 304 course, with the university claiming it's preparing students for "non-negotiable" industry requirements.

• Students have taken to social media arguing AI is "constantly incorrect" and questioning why they're paying full course fees.

• The university maintains students still receive "highly interactive two-hour tutorials in person every week" and that AI skills are essential for marketing graduates in "a rapidly changing professional landscape".

🏘️ Our Take: This pioneering approach at UoA represents a critical inflection point for NZ's education-to-employment pipeline. While preparing students for AI-driven workplaces is crucial, the backlash highlights tensions between educational quality and AI adoption. UoA's implementation appears to lack clear communication and student enablement, potentially going too big too early with lecturer replacement rather than lecturer enhancement. Although, students shouldn't jump to conclusions too quickly—Harvard studies show AI tutoring is a resounding success when implemented to complement human teaching. Learning fundamentally requires human engagement—something COVID clearly demonstrated when online education struggled without personal connection. If AI tutoring succeeds, expect widespread adoption across NZ institutions; if it fails, we risk eroding trust in our tertiary system while graduates accumulate debt for lower educational experiences.

AI X ENVIRONMENT

🌋 Canterbury Uni Builds Volcano Eruption Early Warning System

📣 Word On The Street: Researchers developed a revolutionary AI forecasting tool to spot volcanic eruption patterns, drawing on seismic data from Whakaari and other volcanoes, to identify early warning signals previously missed by human analysts.

🔍 Zooming In:

• Machine learning model is trained on 41 eruptions across 24 volcanoes (including three in NZ) to identify warning signals applicable to any mountain.

• Research led by Dr. Alberto Ardid targets the safety of 29 million people living within 10km of active volcanoes worldwide.

• Tool will be shared via open-access policy with volcanic observatories worldwide, making cutting-edge forecasting available even in developing regions, particularly in Southeast Asia and Central America.

🏘️ Our Take: This breakthrough puts Kiwi science on the global map. With active volcanoes dotting our landscape, GNS Science adds another layer of protection for communities in volcanic zones. Beyond safety benefits, there's real economic value for regions like Rotorua and Taupō that rely on geothermal attractions - better prediction means better business planning and public confidence. The tool's pattern recognition approach could revolutionise how we handle other natural disasters too, from flooding to earthquake aftershocks. The team's commitment to open-access sharing ensures that communities worldwide, especially those with limited resources, can benefit - positioning NZ as both a scientific and global cooperation leader.

COUNCIL AI

💬 Auckland Council Goes Digital First With 24/7 AI Helper

📣 Word On The Street: Auckland Council's Google-powered AI assistant will offer round-the-clock access to council services and information.

🔍 Zooming In:

• Handling over 1.5 million yearly calls, the AI assistant will significantly reduce the effort for users to find information across multiple council platforms.

• Google Cloud and Deloitte are fully funding this initiative, creating one of New Zealand's largest public sector AI implementations.

• Currently in prototype phase, the next step involves testing with actual Aucklanders later in 2025 to refine the technology based on real customer experiences.

🏘️ Our Take: This AI assistant will modernise how we interact with council services, offering 24/7 help for everything from rates queries to finding local attractions. With plans to support our three official languages and eventually more, it reflects Auckland's multicultural makeup while creating a model that could benefit communities across Aotearoa. However, we should remain mindful of potential data privacy issues and whether the system can truly deliver accurate information across the council's complex services. For those without digital skills, additional support pathways will be essential to ensure no one gets left behind.

🥝 Other Kiwi Bites

🌀 AI Meets Haka. Auckland kapa haka group Angitu made history by performing a haka partially composed by ChatGPT. This pioneering fusion of mātauranga Māori with AI comes as Netsafe reveals 46% of Māori face harmful digital communications. Netsafe simultaneously launched crucial digital safety resources to combat the creative potential and risks of emerging tech in cultural contexts. 2-min watch.

⚠️ AI Scams Target Kiwi Finance Providers. Scammers are using AI to create convincingly fake websites impersonating NZ banks and investment platforms, prompting the FMA to issue more public warnings. Sophisticated scams have already duped numerous Kiwis. One alert identified 96 separate fake platforms. 1-min read.

👁️ AI Catches Diabetes Eye Disease. Health Minister Simeon Brown has launched a pilot where AI instantly analyses retinal scans to detect vision-threatening diabetic retinopathy for 26,000 overdue South Aucklanders. This tech-forward approach shifts screening from hospitals to local clinics, potentially revolutionising how we prevent diabetes-related blindness nationwide. 2-min read.

🎨 Artists vs Algorithms. Professor Alex Sims from Auckland University reveals 52% of AI image prompts specifically request artist styles, like Banksy, yet courts are ruling that simply training AI on existing works doesn't constitute infringement unless the output directly copies specific original pieces. 18-min listen.

📈 Kiwis Lead AI Race. NZ business leaders are storming ahead with 40% ranking AI as their top priority - double Australia's 22%. The Feb 2025 Salesforce/YouGov study of 487 executives found NZ firms more likely to have fully integrated AI strategies (38% vs 21%). Smaller firms focus more on training (56% vs 45%), while larger ones prioritise integrating AI into existing workflows (58% vs 49%). 19-page deck.

🏢 Corporate Spotlight

🐄 AI Fights Cow Germs. DairySmart's AI tech can identify mastitis bacteria in just 24 hours, helping kiwi dairy farmers target antibiotic use more precisely. This innovation supports global efforts against antibiotic resistance, which the WHO ranks as one of its top three health priorities. 2-min read.

🥛 Fonterra Embraces AI Revolution. Fonterra is harnessing AI to wrangle its overwhelming data footprint, including 29,000+ SharePoint sites. Caroline Wishart, Head of Information Management at Fonterra, outlined three specific AI implementations: mandatory email classification, an R&D driver enabling scientists to search across all research repositories, and integrating with their 4,000-user asset management system to achieve major cost savings by 2030. 3-min read.

🦘 From Across the Ditch

🔍 Robots Hunting Tax Dodgers. Australian tax authorities have unleashed 43 AI systems with over 100 data scientists to detect small business non-compliance. A concerning 74% of these AI models lack completed ethics assessments, despite their significant role in tax enforcement. 3-min read.

💰 Queensland AI Boom. Noosa-founded Springboards has launched with $5M in seed funding from Blackbird Ventures, while Toowoomba-based Tendl secured $2M to automate government tenders. These regional success stories show how Australia's AI ecosystem is expanding beyond major cities. 2-min read.

🔍 Centrelink's AI Trial. Australia's social security agency, Centrelink, has been testing machine learning tools for detecting welfare fraud and prioritising debt recovery, with humans making final determinations. Following their disastrous Robodebt scandal (automated systems wrongly issued over half a million debt notices to welfare recipients) welfare advocates are calling for public scrutiny of the AI models to prevent automated discrimination. 3-min read.

🌍 The News from Global

🤖 AI Safety Gets Security Upgrade, as the UK rebrands its AI Safety Institute to tackle national security threats like cyber-attacks, fraud, and AI misuse for weapons—shifting focus from bias and free speech. The Data Bill now regulates AI web crawlers, enforcing copyright compliance and data disclosure. 2-min read.

🎭 Gabby Petito's AI Voice Sparks Outrage, with Netflix's controversial decision to use AI to recreate the murdered YouTuber’s voice to read her own journal entries. Despite having parental permission, the choice sparked fierce debate about the ethical use deepfake tech. Is this a step too far for victims? 3-min read.

💰 Alibaba Bets Big On AI, as the Chinese tech giant announces plans to invest a whopping $53B in AI and cloud computing over the next three years. The massive investment comes just a week after co-founder Jack Ma was spotted meeting with President Xi Jinping, sending the company's shares soaring to three-year highs. Is this a sign that China is doubling down on the AI race? 2-min read.

🌏 Tech Updates You Should Know

• Anthropic: Releases Claude 3.7 Sonnet, its "smartest AI model" yet with improved coding abilities and a "Claude Code" agent tool for Devs. We’re still yet to see web search capabilities; 4.0 Sonnet release to include?

• Apple: Announces four-year $500B AI investment, including 20,000 US jobs and Texas AI hardware factory. Another of Apple’s false promises?

• OpenAI: Launches GPT-4.5 as its "best chat model" in the "classic GPT series", excelling at natural conversation and EQ over reasoning tasks, with significantly reduced hallucinations. Available first to Pro users.

• Apple: iOS 18.4 beta code reveals potential Google Gemini AI partnership alongside existing ChatGPT integration for Apple Intelligence, offering users choice between AI models as the company plays catch-up with competitors while key features like Siri 2.0 face delays.

📅 AI Events in New Zealand

We've Changed Things Up: You might notice that instead of our usual list of AI events, we're now featuring a select gathering to keep our newsletter concise, and capturing the full events calendar of AI happenings in NZ on a separate page.

This week’s featured event:

• AI ROI Workshop, 6 Mar - Auckland: SupaHuman & Justin Flitter (NewZealand.AI Founder) present a hands-on AI masterclass on how businesses can unlock generative AI's game-changing potential. No cost.

Want to see all the available events? Access the full events calendar here.

💼  AI Roles around Aotearoa

We've made a similar change to our jobs section, just like the events section above. Instead of our usual comprehensive list of AI roles, we're now featuring a select few positions to keep our newsletter concise.

Picklist of 🌶️ HOT 🌶️ roles in AI.

• VP of Artificial Intelligence - Trilogy, Remote

• Senior Data Scientist - Tracksuit, Auckland

• Data Consultant - Spark, Auckland & Wellington

Want to see all the available positions? Access the full job board here.

📚️ Levelling up with AI

There's a lot of new terminology flying around, so let's clear up the confusion with some simple definitions so that your nerdy colleagues' explanations don't go in one ear and out the other.

AI Terms Simplified

🤖 AI (Artificial Intelligence): Computer systems that perform tasks requiring human intelligence like seeing, speaking, and decision-making. Think of it like cooking, which encompasses various methods and techniques used to prepare food.

💡 Machine Learning (ML): A subset of AI focused on algorithms that learn patterns from data and improve through experience, without explicit programming. If AI is cooking, then machine learning is like different heated-cooking methods—a major subset with many specific techniques.

🧠 Neural Networks: A specific type of machine learning algorithm that mimics how brain cells connect and communicate. Neural networks are to machine learning what roasting is to cooking methods.

📊 Large Language Models (LLMs): AI systems trained on massive text data that can understand and generate human language. These include models like GPT-4o or o1 which can be built using different architectures. They're like specialised chefs who have mastered various techniques for creating elaborate language dishes.

🚀 Frontier Models: The most advanced, cutting-edge AI systems at the forefront of research. These include familiar models like ChatGPT (OpenAI) and Claude (Anthropic). While many frontier models are LLMs, not all LLMs are Frontier models. This category also includes advanced computer vision systems etc.

📝 GPT (Generative Pre-trained Transformer): A specific type of LLM developed by OpenAI. The name means: Generative (creates content), Pre-trained (initially trained on vast text before fine-tuning), Transformer (its neural network architecture). All GPTs are LLMs, but not all LLMs are GPTs—like all soufflés are baked goods, but not all baked goods are soufflés. ChatGPT is further refined with RLHF (Reinforcement Learning from Human Feedback) for better conversations.

🎓 Training: The process where AI systems learn from data by adjusting their parameters. Like a chef spending years learning techniques, tasting ingredients, and mastering cooking fundamentals.

🎯 Inference: When AI applies what it learned during training to make decisions or generate content. Like a trained chef actually preparing meals for customers using all their developed skills.

🤦‍♂️ AI Fail of the Week

We all love AI, but we all know it’s far from perfect…

Thank you for all of your support! Until next week 👋 Mike