NVIDIA's CEO Jensen Huang came to Japan and left behind deals that touch many parts of its tech industry. What matters here is not a single contract, but a bundle that spans several sectors. When demand for AI computing rises, demand for data centers, power, and chips grows with it. This visit looks like a sign that Japan is being pulled into that supply side. From a long-term view, the base that supports the hype can be as interesting as the stars of the show.

Talk about AI investment often centers on GPUs, the powerful chips used for AI computing. But the real bottleneck on the ground is often power and land, not chips. Large data centers use huge amounts of electricity and need water for cooling and a stable grid. Japan has high power costs and limited land, so these limits shape how fast investment can move. In short, where you can build and whether you can secure power become the hidden battleground.

Next, a topic closer to home. When you apply for a job, an AI, not a person, increasingly reads your résumé first. Researchers have long known that large language models—AI trained on huge amounts of text—pick up human bias from their training data. New research reported by MIT Tech Review says these models can also form their own biases, not just inherited ones. By design, an AI can build unexpected habits of judgment during training. Readers involved in hiring should treat an AI's verdict as one input, not the final word.

AI is moving beyond short tasks into work that runs on its own for long periods. OpenAI shared lessons from running such long-horizon models. The longer a model runs, the more likely it is to drift off course or stack up unintended actions. The company says it improves safeguards through iterative deployment—releasing gradually and learning as it goes. When you hand work to an AI, a practical step is to keep a way to check its progress rather than leaving it fully alone.

Finally, a note for those who create content. Adobe's experimental camera app, Project Indigo, has been updated with generative AI tools. It was first built for a natural, SLR-like look, but now adds features like removing backgrounds, The Verge reports. What stands out is that it does not rely only on Adobe's own generative AI. This hints at a trend where generative AI is quietly built into apps and operating systems. Even without opening special software, we may touch generative AI more often in everyday shooting.