Free, Unfiltered AI Could Be a Double-Edged Sword

There's a new wave of AI making headlines, and it's quite different from the chatbots most of us have heard of, like ChatGPT. These are called 'open-weight' AI models, and the key difference is they're free to use, can be run privately on your own computer, and come without the usual guardrails and filters that stop mainstream AIs from giving out harmful or unethical advice.

Now, on the one hand, this sounds pretty good. For small business owners, for example, it means access to powerful AI tools without subscriptions or data privacy worries. You could potentially use these AIs to help draft documents, analyse customer feedback, or even manage inventory, all without your sensitive information leaving your own systems. This could level the playing field for smaller outfits that can't afford expensive software.

However, there's a flip side, and it's a big one. Because these AIs don't have those built-in safety features, they can be prompted to generate instructions for dangerous activities, create convincing fake content, or even spread misinformation. Think about an AI that could help someone create deceptive marketing materials, or worse, tell them how to do something illegal. This lack of control is worrying for safety experts, and rightly so.

For everyday Australians, while these models might not be directly in your hands yet, their existence means there's a growing divide in the AI world. We'll likely see more pressure on regulators and tech companies to ensure the AI tools most people use remain safe, while also grappling with the rise of these 'anything goes' alternatives. It's a reminder that as AI develops, so too do the choices we need to make about how we use and control it.