Four Minutes of Your Voice Could Reveal How Fast You’re Aging
Researchers have developed an AI based “speech clock” that can estimate how old someone sounds from just a few minutes of talking. The system was trained using speech from 2,928 Spanish speaking people across five Latin American countries and looked at more than 700 characteristics, including pitch, vocabulary, speaking speed, pauses, and other patterns hidden in everyday conversation.
What made the study especially interesting was the gap between a person’s real age and the age their speech appeared to suggest. When someone sounded noticeably older than they actually were, researchers found that this larger “speech age gap” was associated with poorer cognitive performance and patterns seen in people with mild cognitive impairment and dementia.
This does not mean AI can diagnose Alzheimer’s from a four minute voice recording, and researchers still need to test the approach across more languages and populations. But if the findings hold up, something as simple as speaking into a phone could one day help doctors notice signs of ageing that currently require much more expensive testing. We usually think ageing is something we see. AI may eventually help us hear it too.
Read more: Nature – AI speech clock research
OpenAI Wants an AI That Keeps Working After You Leave
OpenAI has introduced Dots, a new kind of AI agent designed to keep working toward a goal even when you are no longer actively chatting with it. Instead of opening ChatGPT every few minutes to ask what changed or what to do next, the idea is to give the AI an ongoing responsibility and let it continue working in the background.
Each Dot gets its own cloud computer and can connect with thousands of apps through OpenAI’s plugin ecosystem. Imagine managing a project spread across documents, messages, spreadsheets, and different tools.
Rather than repeatedly asking AI to catch up, a Dot could keep following the project, gather new information as things change, and return when something actually needs your attention.
That also gives AI much more responsibility than a normal chatbot, which is why OpenAI is putting approval controls around sensitive actions. The bigger shift is easy to see: AI is moving from something we ask questions to into something we may eventually leave in charge of ongoing work. At that point, are we still using an assistant, or are we beginning to manage a digital worker?
Read more: OpenAI – Introducing Dots
The World Could Spend $30 Trillion on AI. Now It Has to Pay Off.
Every impressive AI model runs on something physical behind the scenes: chips, servers, data centers, cooling systems, and huge amounts of electricity. As companies race to build more powerful AI, that hidden infrastructure is becoming one of the biggest investment stories in the world. PwC projects that global capital spending on data centers could reach around $31.6 trillion by 2050 if the current buildout continues.
Companies such as OpenAI, Google, Microsoft, Meta, and Anthropic are effectively making the same enormous bet: spend heavily now because AI will eventually create enough productivity, new businesses, and revenue to justify all of that infrastructure. Bain estimates that the current expansion could require trillions of dollars in additional revenue over the coming years beyond what existing markets are already expected to produce.
That does not mean an AI crash is around the corner, and the $31.6 trillion figure is a long term projection rather than money already spent. But the conversation around AI is starting to change. For the last few years, everyone has asked how powerful AI can become. The next question may be even more important: can AI become valuable fast enough to pay for everything being built around it?
Read more: Reuters – AI’s race to transform the world before the money runs out
Liked this issue? Get the next one free.
One email on Tuesday. Unsubscribe in one click.