👋 Good morning from a yum cha restaurant in Guangzhou!

I've just spent the last few days at the 11th Silver Industry China Expo in this beautiful city. The scale and pace of development were beyond anything I expected, from AI-supported tools treated as core infrastructure, to high school students pitching tech & aged care projects alongside government officials and aged care providers in the same rooms. The integration across the entire ecosystem, from concept to deployment, is simply remarkable. I'll start sharing some insights and takeaways from this week, over a few parts.

But before we continue, I still need your help!

In nine months, this newsletter has grown from fewer than 20 subscribers to more than 800. That's thanks to you, reading, sharing, and engaging with what I write each Tuesday.

Now I need to understand what's actually working. What do you find valuable? What's missing? What should I do more of, or less of?

I've put together a short survey (3-4 minutes). Your feedback matters, and all responses are anonymous. Tell me what you think.

What we cover this week:

CHINA INSIGHTS

After six weeks in China, including the Silver Industry Expo and several site visits, I'm starting a series of articles on what I'm observing.

In the first one, I focus on a policy that stands out: China directly subsidises technology adoption for older people with up to $5,000 AUD per person—no assessment required. The program covers smart monitoring devices, AI assistants, smart beds, toilets, home modifications, and more.

Crucially, family members can claim the subsidy on behalf of older relatives, reflecting collectivist decision-making patterns—perhaps something we need to learn in Australia, where our system prioritises individualist approaches despite our population coming from many collectivist cultures as well.

In my full analysis, I discuss the scheme and discuss some of the predictable objections—cost, privacy, "we can't just copy China". I believe Australia has the funding, tech sector, and ageing population to benefit from this. What we lack is a little imagination.

Read Part I of China Insights

AI IN CARE

AI translation has reached a turning point. Current systems achieve 96% accuracy across 133 languages, and in healthcare specifically, AI translation usage rose by 700% between 2023 and 2024. Admittedly, this report may be biased, but the numbers remain impressive even accounting for that. If you need to see peer-reviewed articles, see this one, or this one.

(Some critics rightly point out that many languages remain underserved, as only 20% of languages are currently supported by major platforms. But Meta's Omnilingual ASR recently expanded speech recognition coverage to more than 1,600 languages, including over 500 never before served by any system.)

Whilst I acknowledge this is terrible news for human translators (no point sugarcoating it), for us in aged care it means we can now access AI-powered interpretation and translation on any smart device with 24/7 availability. As I mentioned in a previous newsletter, I bought in-ear translation earbuds in China for $26, and they work. Families who previously relied on limited-English-speaking relatives to interpret can now communicate directly with care teams. And there are benefits for staff too.

The smart approach combines AI with human interpreters, but not for the sake of translation itself. AI is already as good as humans in many languages, and eventually it will become better than humans in all languages. But we'll still need humans for high-stakes conversations (such as advanced care planning, family meetings about end-of-life care, mental health discussions) that require emotional intelligence and a human in the room with us.

Stephen Fry and Gemma Whelan star in a new short drama exploring AI and memory. Fry plays a grandfather with dementia using AI to fill memory gaps. As his family reviews his life archive, they make a shocking discovery that raises questions about which memories are real and how AI shapes identity. Written by David Baddiel. (I saw this on Stephen Johnston’s latest newsletter. Highly recommended.)

The fastest-growing occupation through 2034 isn't in tech, but rather it’s care workers, adding 740,000 jobs (in the US market). Healthcare will account for 1.7 million new positions, driven by ageing demographics. These roles require empathy and connection that AI can't replicate.

In more bizarre news, a Disney Channel actor launched 2wai, an app creating interactive AI avatars of deceased relatives from minutes of footage. Currently free on iOS, it's drawn widespread criticism from exploitative to "demonic", raising concerns about healthy grief processing, and the psychological risks of forming attachments to simulations of the dead.

Continuing on the subject of translation, Synthesia provides an AI tool that translates and dubs videos into multiple languages. It creates new voiceovers, aligns lip movements, and keeps the speaker’s tone and pace. You can also adjust the phrasing (while the system ensures you are not altering the original meaning), which reduces the risk of inaccurate or misleading translations. I have used this tool before, and it’s one of the best ones out there.

The tool can be used to support communication with CALD clients (both marketing to new clients/communities, as well as sending updates to existing clients). It also offers an easy way to create training content for a diverse workforce. Watch this video to see how it works, or register here to try it yourself.

One of my newsletter readers recently pointed me to a claim online suggesting that ChatGPT has “secret slash commands” (things like /rewrite, /email, or /plan) that supposedly unlock special modes or built-in templates. I was doubtful, because OpenAI hasn’t announced anything like this, and it didn’t fit with how the model actually works. I tested the commands myself across several prompts and found no difference at all in the output. The slash doesn’t trigger anything; the model just reads the instruction normally, the same way it would if you wrote “rewrite this” or “write an email.” If you want real structured outputs, the reliable method is to state the structure explicitly in your prompt or use a custom GPT you configure yourself.

I’m not here to hype trends. I’m here to explore the changes shaping ageing—technology included—and to share ideas you can apply in practice. Whether you’re exploring new tools, rethinking services, or looking ahead to what’s coming, I hope you found something here worth your time.

Feel free to forward this to your network or share it with your team.

See you next Tuesday,
George