Killed by Robots

AI Artificial Intelligence / Robotics News & Philosophy

  • AI Alignment: Our Moral Mirror
    When we talk about Artificial Intelligence, especially the really powerful kind – the general AI that can think and learn like us, or even better – one of the biggest headaches we face is something called the ‘alignment problem.’ Simply put, it’s about making sure that these incredibly smart machines do what we want them to do, and not something else entirely. It sounds straightforward, right? Just tell the AI to be good. But, as anyone who has ever tried to negotiate bedtime with a toddler, or stick to a New Year’s resolution, knows, ‘good’ can be surprisingly tricky to…
  • CES 2026: The Future Arrived
    The year 2026 will forever be etched in history as the moment the future truly arrived. At the Consumer Electronics Show, a profound shift unfolded before our very eyes. For years, we’ve marvelled at glimpses of artificial intelligence and robotics, often in carefully choreographed demonstrations. But at CES 2026, something fundamentally different happened. The machines stopped performing for us and started working alongside us. This wasn’t just an exhibition; it was the unveiling of a new epoch, where the science fiction dreams of yesterday began to manifest as the practical realities of today. The Awakening of the Humanoids Witnessing the…
  • Robots Learn to Touch and Feel
    Imagine a future where robots aren’t just tireless machines confined to factories, but intelligent partners capable of understanding you, seeing the world, and even feeling with a gentle touch. This long-held dream is now taking a significant leap forward, thanks to a remarkable new development from Microsoft Research. They’ve unveiled **Rho-alpha**, a pioneering “master brain” for robots, the first of its kind from Microsoft’s advanced Phi series. Rho-alpha empowers robots to interpret natural language instructions – just like you’d speak to a person – and then combine what they see with what they feel through touch, to execute surprisingly complex…