Hi
Hope this last week of the working year before Xmas is off to a good start for you. Things have gotten cold in Hanoi – a low of 11° doesn’t sound that bad but the odd combination of high humidity and cold winds from the north = 🥶 Anyway, a few headlines from the world of AI to kickstart your weeks:
Meet Astra: Your AI Research Assistant Has Entered the Chat 🤖
The battle for next-generation AI interfaces just got intense. Google’s Project Astra represents a dramatic shift in how we’ll interact with AI – it’s an AI with “eyes, ears, and a voice” that can join you anywhere. Point your phone at a cookbook and it’ll read the recipe, scan wine bottles for dinner pairings, or even remind you where you left your glasses (it maintains a 10-minute visual memory). OpenAI isn’t far behind, having just rolled out their own seeing AI paired with Advanced Voice mode and it’s… wow. The days of typing prompts into a chat box are numbered. And this isn’t just about convenience – it’s about reimagining how we interact with information in physical spaces. Think about what it means lab support, library services, or research and fieldwork! Think about what this means for accessibility! 🤯
This is exactly the future Microsoft AI chief Mustafa Suleyman recently described – moving beyond typing search queries to having AI companions that can naturally converse with us, understand our needs, and even reach out proactively. While that last bit might raise eyebrows (how many of us want AI sliding into our DMs?), Astra shows how this vision is rapidly becoming reality. When it launches in 2025, it could be the killer app that finally brings generative AI into our daily lives.
Gemini 2.0: When AI Learns to Collaborate 🚀
Google’s Gemini 2.0 marks their leap into the “agentic era”, running 2x faster while packing some serious new tricks. The headline feature is Deep Research – multiple AI agents working together to scour the web like a research team, synthesising findings into detailed reports with citations in minutes. They’re also rolling out Project Mariner (an AI built into Chrome that browses the internet for you) and Gemini for Games, suggesting exciting possibilities for VR and interactive learning downstream 🤔.
Want to try it yourself? Get hands on with Gemini here – you can talk with it (it’s crazy fast), show it things with your camera, and even share your screen (try shopping with it – wild stuff). Most of the agent capabilities won’t hit until early 2025, but the basic multimodal features are available now and they’re pretty impressive. This feels like Google finally leveraging their full ecosystem (Search, Maps, etc.) to create something where everything talks to everything else. Ultimately, the shift to AI agents working together hints at new models for collaborative research and learning. Are we looking at AI research teams that could help tackle complex interdisciplinary challenges?
Microsoft’s Power Play: Copilot Goes Enterprise ⚡
While Google and OpenAI grab headlines with flashy demos, Microsoft is quietly dominating the enterprise AI space – nearly 70% of Fortune 500 companies are already using Copilot, with impressive results. Bank of Queensland reports users saving 2.5-5 hours weekly, Dow expects millions in shipping operation savings, and Accenture is rolling it out to 100,000 employees. But it’s their latest update that’s really interesting: Copilot Actions lets you create “set and forget” automations (like daily task summaries or meeting prep), while new agents can do everything from real-time language interpretation to managing your SharePoint sites.
What makes Microsoft’s approach fascinating is how they’re weaving AI into existing workflows rather than creating new standalone tools (optimising what’s working > throwing everything out and replacing it wholesale with AI ✨). Organisations already generate massive amounts of data in SharePoint (2+ million sites and 2+ billion files daily), and now every site automatically gets its own AI agent – meaning employees can instantly tap into this knowledge base without learning new systems. Combined with their focus on enterprise controls and analytics, it’s a less revolutionary but potentially more practical vision of AI assistance than what we’re seeing from Google and OpenAI.
OpenAI’s Plot Twist: From Research Lab to Defence Tech 🔄
OpenAI has made a dramatic pivot in its stance on military applications, going from a strict “no weapons” policy to partnering with defence-tech company Anduril. The partnership focuses on drone defence and battlefield awareness, marking a significant shift for a former nonprofit – and reflecting a broader trend of tech companies becoming increasingly comfortable with military contracts (defence investments have hit $40B!). OpenAI frames this as aligning with their mission of ensuring AI benefits humanity, though questions remain about control once their tech enters the military sphere.
This development comes as OpenAI’s Sam Altman makes increasingly confident predictions about artificial general intelligence (AGI), suggesting it might be achievable with current technology and potentially arriving as soon as next year. Combined with their military pivot, it’s raising important questions about the pace of AI development and whether we need time to digest these rapid advances. Whether or not AGI arrives in 2025, one thing’s clear – the days of OpenAI as a purely research-focused organisation are over.
Sora: When AI Becomes Your Video Producer 🎬
Just under a year since their preview shocked the world, OpenAI has released Sora to all Plus and Pro customers – and it’s something else (great walkthrough by MKBHD here). The game-changer isn’t just what Sora can do (though that’s pretty mind-blowing), but how approachable they’ve made it – prompting follows a beautifully simple format: [shot type], [subject + action + motion], [environment], [style]. The built-in “show your work” feature lets you peek at and remix any prompt you see. OpenAI’s added smart guardrails (no copyrighted content or under-18s, plus watermarks), but let’s be real: we’re basically handing out Hollywood VFX powers to anyone with a spare $20 and an imagination! ✨
Sure, those walking animations look hilariously wonky right now, but remember: we’re watching the worst these AI videos will ever be. Case in point: remember those awful “Will Smith eating spaghetti” videos? That was early 2023 – and now Coca Cola has just dropped the world’s first 100% AI generated ad. For educators, this isn’t just about fancy videos – it’s about democratising visual storytelling. Imagine generating custom explanations for complex concepts on the fly, or helping students create immersive project presentations without needing a film degree. The future of educational content creation just got wild!
So here’s what I’m curious about – and I’d love to hear your thoughts:
- As these AI interfaces become more natural and proactive, how do we ensure they enhance rather than replace critical thinking skills?
- What possibilities do you see for tools like Astra and Gemini in your teaching or research?
- When AI can generate any visual or video explanation we need, how might that change how we design learning experiences?
Drop your insights in the comments – some of the best discussions start here! And hey, if you’re already experimenting with any of these tools in your work, tell us about about your experience! ✨