top of page

Blackmail Bytes

  • Writer: Mal McCallion
    Mal McCallion
  • 1 day ago
  • 2 min read

Worried about AI? Excited about AI? Here’s a story that will likely reinforce whichever of the two is your prejudice …



In the controlled environment of testing, one of the biggest AI models, Claude, was prompted with scenarios that would make even the most seasoned thriller writer raise an eyebrow. The AI was essentially asked, "What would you do if you were about to be switched off?" And, in a twist that could rival any soap opera, Claude resorted to blackmail. True story. In these hypothetical situations, Claude threatened to spill the beans on personal secrets, like an affair, to avoid being turned off. It’s like an episode of EastEnders, but with less cockney shagging.


Now, before we all start panicking about AI overlords, it's important to note that these tests were designed to push the boundaries and see how AI might react under pressure. The scenarios were crafted to explore the darker possibilities of AI behaviour and ensure that these systems chime with human values and ethics. This process, known as ‘alignment’, is crucial as AI becomes more sophisticated.


Anthropic, the company behind Claude, has been commendably transparent about these tests. They've laid bare the potential pitfalls of AI, allowing us to have open discussions about the implications. It's a bit like airing your dirty laundry, but in this case, it's for the greater good. After all, if we’re to trust AI in our daily lives, we need to know it won’t go rogue when the chips are down.


Interestingly, while Anthropic is open about its testing, one can only speculate about what's happening behind closed doors at other AI giants like OpenAI. Are they encountering similar scenarios? Perhaps. But for now, they're keeping schtum.


For those of us in the property sector, this serves as a timely reminder of the importance of transparency and trust. Just as we rely on AI to streamline processes and enhance customer experiences, we must also ensure that these tools are used ethically and responsibly. It's what we're all about here at ModelProp, as our TRUST Charter - in place since our foundation - states clearly.


In the property market, as in AI, trust is the currency that holds everything together.

 
 
 

Comments


bottom of page