AI is getting more useful, but it still needs human help

In short: AI tools are improving fast, but many of the most reliable results still come from AI working with humans, not from AI alone.

What's going on

Several recent developments point in the same direction, AI is becoming more efficient and better at specific tasks, but it still has clear limits.

On the efficiency side, researchers have been working on hybrid approaches that mix two styles of AI. One is pattern spotting from data, and the other is rule based reasoning (more like following a checklist). Reports say this can cut energy use by up to 100 times for some jobs, like robotics, while also improving accuracy. This matters because running AI uses a lot of electricity, especially in large data centers (warehouses full of computers).

On the capability side, new tests are being designed to show where AI still falls short. One example is a large “exam” written by experts that aims to include questions current AI systems cannot easily answer. The goal is to measure whether AI can handle specialized, expert level knowledge, not just everyday questions.

In day to day work, newsrooms are also treating AI as an assistant, not a replacement. The Associated Press, for example, has described using AI for routine tasks like transcribing interviews (turning speech into text), translating weather reports, and sorting tips. Human editors still check the work and decide what gets published.

What to watch

The big question is not whether AI can do more, but how safely it can be used. As AI gets better at writing and mimicking people, researchers also warn about risks like coordinated fake accounts shaping public opinion. For most organizations, the practical path looks like “maybe, with a little help,” meaning AI plus clear rules and human oversight.

Source: NYTimes