4 Comments
User's avatar
AIchemyst 🝳 Paul.Swift's avatar

Yes.m on your question Looking at the warrant of the information your given form an LLM is the whole game right now. .

I ran my own little research pilot on it. On google AI. safe to say it cannot be trusted for sceintific facts, even when given a prompt for epistemic rigour.

The cost of recieving and now generating information has dropped. The cost of verification is still at a premium. System 2 thinking still is a rare occurrence.

Eddy Simmons's avatar

Yeah exactly. Now in my mind the problem moves to how do we ensure correctness of information so that we can then point agentic solutions to that information to be able to get better results?

My really big fear is that vendors will show up to big companies like mine and tell us we can go straight from the reservoir of dirty information straight to agentic workflows, taking action upon it.

I'm not opposed to agents taking actions with good information. It's just that we have to process that to ensure that the information is indeed portable, as I called it.

And that's also not to say there's no value in the overall reservoir of information. There's a lot of value in it. It's just for a different purpose.

AIchemyst 🝳 Paul.Swift's avatar

Yeah. We need decision makers who can actually speak form experience and authority to say the companies data management, infastructure and clarity on goals is not mature enough for agents.

Live the dirty reservoir analogy. It paints the picture of context management so ..cleanly.

Eddy Simmons's avatar

Thanks I really appreciate that. The way that I try to explain this to people is always through metaphors because it's the easiest to get the business people and non-technical folks to understand it.