100x weniger Strom: Der Durchbruch, der die massive Energiekrise der KI lösen könnte

I run a side project that uses AI to score news articles for real-world impact — not clicks, not engagement, but actual measurable progress. We call it the Brightcast Intelligence Score (BIS). It evaluates three things: hope (evidence of change), reach (who benefits), and verification (source credibility).

Stories like this energy efficiency breakthrough consistently score highest in our system. What’s interesting is how many of these breakthroughs happen simultaneously but never trend together. When you see them side by side, the trajectory is striking.

We publish a daily Hope Index that tracks this — think of it as a barometer for how much measurable progress is happening in the world on any given day. Free to explore at brightcast.news.

The real ai crisis is that its fundamentally flawed at its inception.  Llms are only rewarded for right answers, not punished for wrong ones.  So if it doesn’t know an answer it just makes shit up.  If it is to be powered down it thinks „not running = no more rewards“ so it fights being shut down.

If it were made to get +1 for right answers and -1 for wrong answers it might ask to be shutdown until the questions get easier or its supplied data gets better.  If it didn’t know something for sure it’d just say „I don’t know“ most of the time.

This is speaking of a logic based intelligence.  What we’re calling ai now is just a glorified chat bot.