Why did I choose Siegfried and Roy? Because it makes sense. They made magic shows where the audience paid to be wowed, even if they perfectly knew it's all just smoke and mirrors. They *wanted* to be fooled. And happily paid for the experience. That's how GenAI feels to me right now.

And my problem is that the marketing hype around the Big 4 in GenAI starts to make life bad for the honest people that have worked on Machine Learning and other forms of truly good solutions that are now summed up as "AI". The well is getting poisoned. Le Sigh.

On Wednesday I will lead a session with experts and journalists on this very topic. I will introduce an, IMHO, better way which I call #SwarmAI. Instead of the One Big Model To Rule Them All, let's have domain specific, small models that can be ML (Machine Learning), GenAI (Generative AI) or something else. Let's feed those small models with verified data. And combine their inference in a weighted way to deliver results. A swarm instead of an oracle.

This will make for a far better approach with far better results, I'm convinced of that. And a lot less energy will be wasted. "Big AI" has this weird philosophy of "If I feed all the shit I can get hold of, it will produce golden eggs". It doesn't. But domain specific, small models can deliver truly impressive results, in my experience. Combining those ist the better way, again IMHO.