There was a time when Eliza was the latest and greatest in AI but that was 30+ years ago. Today we have LLMs but do not think that this is some kind of AI endgame, far far from it

We will look back at this period from a more advanced AI position built on better technology

LLMs will be seen as the darker, earlier, more formative times

It's just another stepping stone - isn't the destination

Generative AI is great at generation, funnily enough.... but it ain't so good at analysis

Maybe AnalyticalAI arrives, supported by different underpinning tech and provides a much better, more valid experience than asking a pattern matching matching pretending to be human via a natural language interface....which is turning out to be a reliable misinformation generator

When you're being creative and generating, any hallucinations are good. They're creativity.

When you're doing analytical work, any hallucinations are bad. They're fairytales.

AI one size fits all isn't going well. Fine tuned models or just models that aren't born to guess has to be where we're going - it's just messy as hell right now trying to fit one version of the technology to every single potential use case out there.

Better, more appropriate models for the task at hand will mean finding those use cases faster.

Happy Sunday! I'm heading to the airport shortly, two boxes of freshly baked pastel de nata in my luggage 🤘🤘😎