AI and all that Jazz
Meeting report 23 September.
– 23 Sep 2025: AI by Justin Flitter
Justin founded AI New Zealand in 2017, years before the generative AI boom. His background includes tech marketing and early work in natural language processing (NLP) — things like predictive text, sentiment analyses, etc.
He describes how large language models (LLMs) like ChatGPT work: trained on huge datasets to predict next words. Also notes that past “AI winters” were due to lack of data and compute, which the modern era has remedied.
AI is disrupting traditional workflows: research, consulting, reporting, presentation creation etc. Tasks that used to require large teams or hours of work can now often be done quickly via AI tools.
Preserving institutional/organizational knowledge is a major use case: e.g. transcribing meetings, capturing founders’ tacit knowledge, and helping companies retain what makes them unique. Also, using chatbots for compliance or on-site worker safety.
Emerging AI modalities: text-to-image, text-to-video, digital avatars. Interfaces shifting toward voice & vision rather than keyboard/mouse.
Caution & ethics: Justin refers to warnings from AI researchers (e.g. Geoff Hinton) about short-term profit motivations, misuse of AI, and risk of machines growing too controlling.
Practical examples: automating mundane tasks. E.g., making an online order via shopping list; generating architectural renders via AI; asking questions about past data rather than digging through documents manually.
Interactive content & voice/vision enabled tools: asking AI questions mid-podcast/audio, photo/vision apps to analyse landmarks, changing accents/voices in AI content, etc.
An audience member asked risk risks and concerns related to AI:
Hallucinations: Flitter explained that AI "hallucinations"—generating factually incorrect or nonsensical information—occur because the models are "incentivized to give an answer" even when there are gaps in their training data. This is a core aspect of how they're designed.
Job Displacement: AI is already replacing entry-level roles and some complex, senior-level work, creating a "punching pressure" in the workforce and making it harder for graduates to find jobs.
Threat to Business Models: The business model of consultancies is under threat because AI tools can now perform the research and report generation that previously required a team of people.
Misinformation and Disinformation: With AI-generated video and news, the lines between reality and fabrication are "going to become very, very gray very quickly." Flitter noted that this contributes to a bombardment of "quite convincing narratives" that manipulate what people think.
Safety and Privacy: Flitter addressed the concern that information typed into AI tools could become publicly available. While he stated that models typically don't work that way, he acknowledged this is a worry for many people.
Long-Term Existential Risk: An audience member raised the concerns of AI pioneer Geoffrey Hinton, who has cautioned that companies are developing AI for "short-term gain" and that it could one day start "controlling us." Flitter responded by saying "we're a while away from that" and that current models are not advanced enough to become an "artificial general intelligence" (AGI).
For a Video of the full meeting use the link below to view in YouTube
https://youtu.be/lMB55HyFcVE?si=N2qCdtrUqwSQ1I4N
Web Site for AI New Zealand