12 Comments
User's avatar
John Rux-Burton's avatar

Fascinating, and like the most reasoned arguements, blinding obvious ONCE made

Exploring ChatGPT's avatar

Thank you very much John, this means the world to me!

Neural Foundry's avatar

The feedback loop mechanisn you describe is fasinating. When AI optimizes at scale, individual outcomes stop being random in the traditional sense. Its like the diferrence between rolling a die naturally versus rolling one where the surface has been subtly warped by previous rolls.

Exploring ChatGPT's avatar

Thanks Neural! Precisely, once an AI is shaping the setup, the “randomness” stops being pure chance. It starts drifting in whatever direction the system keeps reinforcing. The die still rolls on its own, but the table isn’t the same table anymore, the surface keeps shifting based on past patterns.

Cyphram's avatar

Let’s call this what it is: machine bias. The assumptions of the designers will shape model priority.

Exploring ChatGPT's avatar

I agree though not entirely, the burden of proof that the assumptions of the designers will shape model priority is on those who believe this to be true, and unfortunately this is impossible to prove unless you are the designer. The corporations working on this behind the scenes know the truth.

Chris's avatar

I just asked ChatGPT what I could ask it to create synthetic luck, and it replied, “Given my current situation, what are the highest-leverage asymmetries available to me right now?” Great information!

Exploring ChatGPT's avatar

Thank you for sharing this Chris! Could you expand on the conversation, specifically in regards to these “asymmetries”.

CJ Pace's avatar

Does your hypothesis apply to health insurance?

Exploring ChatGPT's avatar

Thanks CJ, that’s a great question! That’s a great question. In a way, yes, the same logic can apply to health insurance, because insurance is basically a system that tries to smooth out luck across a population. It takes individual randomness and turns it into shared risk.

But the key difference is that insurance doesn’t generate “synthetic luck” the way algorithms do. It manages the luck you already have. Whether or not health insurance companies are already using AI systems to optimize their policies and pricing is a great question that I haven’t looked into.

So anyways, the overlap is there, but the mechanism is completely different.

CJ Pace's avatar

Thanks for your response. I think the way I think is “synthetic luck” is new(thanks) for me and represents variables unknown connected with an outcome. Maybe that’s what is sometimes called “hope”. Or is it something created and real?

Exploring ChatGPT's avatar

That’s a solid point. I think looking at it as hope is an interesting outlook. At the same time hope is something created and real, so the two definitely overlap.