If you haven’t already, subscribe and join our community in receiving weekly AI insights, updates and interviews with industry experts straight to your feed.

DeepDive 

AI Your weekly immersion in AI 

Imagine you get home after a long day at work, and your coffee mug greets you by name. Or your phone highlights your cat in a crowded photo with impressive certainty. Meanwhile, over on TikTok, a totally AI-generated video of an exploding cathedral made of jelly is racking up millions of views…created purely for mischief, without a single shred of utility. 

Welcome to the wonderfully odd corners of AI in 2025. While the headlines circle around enterprise copilots and scientific breakthroughs, the fringes are full of creativity, whimsy, and experiments that defy categorisation. 

Just to lighten up your week a little, here are three of our favourites.

1. Everyday objects get a personality: Talking Spell

It’s one of the most charming pieces of research this year, and the stuff of childhood dreams. Talking Spell is a wearable system that lets you assign a voice and persona to any object you pick up – your mug, your pen, your stapler. The system uses a combination of a wearable camera, vision-language models and a speech interface to make your objects ‘talk back’.

The researchers found that people used the system for entertainment, companionship, creativity and even (sort of) utility – like asking their water bottle to remind them to hydrate.

It reimagines the concept of the AI assistant. Instead of a new device or chatbot window, the assistant becomes an object you already own and use every day. We love the way this opens up questions about how we anthropomorphise tools…and what happens when our belongings start talking back.

2. ‘Find my mug, find my cat’: MIT’s personalised object localisation

MIT researchers unveiled a technique that teaches vision-language models to identify a specific object or animal, not just a generic category. Ask it to find ‘my striped mug’ or ‘my dog Fluffy’, and it can spot that unique item in a new image – even when it’s nestled inside a complex scene.

The team combined video tracking data with in-context examples so the model can reason across time: this mug is the same as the one it saw earlier, even if the angle, lighting or background changes.

This is a small step for AI models, but a big step for user experience. AI that recognises categories is useful; AI that recognises your actual stuff feels almost magical (or slightly intrusive, depending on your perspective). 

With development, the technique could power assistive technologies, memory aids for neurodivergent users, or smart home organisation tools. 

3. The glorious chaos of AI slop

Not every AI breakthrough is a polished marvel. In fact, one of the most culturally pervasive AI phenomena of 2025 is what the internet has dubbed ‘AI slop’ – low-effort, surreal, and often very silly generative content that floods social feeds.

We love to hate it. 

Think endlessly looping AI videos of bizarre events, uncanny product mock ups, and deepfakes that are more confusing than convincing.

Business Insider recently documented the rise of AI-generated prank videos engineered solely for virality – from chaotic fake scenes to impossible physics.

It’s all messy and unserious, and it’s unexpectedly influential. AI slop shapes online culture just as much as carefully crafted generative media. And as out-there as it might be, we like the way it shows that not all AI innovation is about efficiency or accuracy. Sometimes it’s just for fun. 

The deeper meaning 

These niche experiments do reveal a few interesting truths about where AI is heading:

• Play is a driver of adoption. People explore new tech most freely when there’s no pressure for utility.
• Personalisation is accelerating. From object voices to object tracking, AI is becoming more intimate and context-aware.
• Cultural impact matters as much as technical capability. Unserious AI content is shaping norms, expectations and attention spans.

Finally, a question for you

If you could give one object in your home a voice and personality, which would you choose – and what would it sound like?