Mat McGann

Using "right and wrong" or "true and false" in a given scenario is an attempt to guide decision making on a foundation. There are no foundations. Instead of "right and wrong" consider what's "outsourced on trust" or "stood up to in conscience"

Prediction: Typos will be an indication of manual effort and therefore a status thing as AIs saturate our communication

@svembu A common use case is to generate reliable RAG of known unknowns from the outside world. Works for arbitrary data models Horiz.in

@svembu Agreed. We are building data pipelines using only the reasoning of LLMs to create and maintain structured knowledge bases. The process can extract latent information from unstructured sources. As a result the data has full provenance. Often used to provide reliable RAG.

All the LLMs and other deep learning models are based on neural networks. We can think of them as mathematical functions with hundreds of billions of parameters. Those parameters (weights) are determined during training and we train these networks with trillions of tokens (text, images, videos that are split up into tokens to be ingested by the models). We can say that every one of the trillions of tokens played a part in determining the value of each of the hundreds of billions of parameters. The image I have in mind is a giant lake where we dissolve trillions of cubes of salt, sugar etc. After the dissolution we cannot know which of the cubes of sugar went where in the lake - every cube of sugar is everywhere! Therein lies a problem: if we use a business database, such as customer relationship data, to train a neural network model (i.e to determine its parameters), when the customer changes that data or deletes the data, we do not know how to alter the weights of the model to account for this change in the data. Even if the model were dedicated to that customer, we still cannot guarantee the customer that their changes to the data will be reflected in the model. In that sense, neural networks (and therefore LLMs) are NOT a suitable database. This is a fundamental limitation of the current scientific mathematical approach and cannot be fixed only by technological fine tuning. The RAG (retrieval augmented generation) architecture keeps the business database separate and augments the user prompt with data fetched from the database. In that case, the model itself is not trained on the (potentially changing) customer data because that data is only used in the prompt. But RAGs can only go so far. I personally have come to believe more foundational work is needed. What does that look like? All I have right now are hunches. That is the existing part of scientific work!

I bet the JFK files show incompetence and corruption rather than conspiracy

In matters of civilisation, the only thing that makes things inevitable are thinking they're inevitable

Clear as day availability bias. People overestimate how likely civil war is in the open societies and underestimate how likely it is everywhere else

Rank historicism

@averykimball Axioms and theories do seem different but I'm not sure they are. They're all just ideas. When an experiment doesn't give you what you expect it could be your theories or your axioms that are wrong

@MatMcGann theories underlying an instrument being wrong doesn't falsify determinism as you say it's a ground assumption, an axiom, 'falsification' doesn't apply to it- it's used *for* falsifying (like the other suite of consequences inside Realism)

I encourage you to look at determinism as the boldest conjecture in history. And wrong.

It's cool that it works when it does. And it obviously gets at something important and true about reality. But it is an assumption about the world. It is by no means self evident and every experiment that has had any kind of experimental error (which is all of them) falsifies it.

Determinism states that things must be fully deterministically caused and it applies to every aspect of reality at all times. Determinism is just the theory-est theory. It's maximum constraint and maximum reach.

In my opinion it's a Dunning Kruger effect. When you learn enough physics you see determinism underlying all it's success and infer it's truth. But when you learn more physics you realise that our understanding of the world is waaay patchier.

Determinism is a fertile theory where it works but people seem to think it's more than a theory.

"Casual reasoning [and hence determinism] is a heuristic" 👌👌👌 i.e. determinism is a theory about the world, not a ground truth. youtu.be/nh1Z3UTobrY?si…

I've listened to hundreds of In Our Times. They've covered every kind of famous person imaginable. Except actors. I can't remember a single IOT about an actor.

This would explain the drastic rise in allergies over the last few decades: Feedback loop: - Some kids are born allergic - More awareness of this > less exposure - less exposure causes emergent allergies - More awareness > decrease exposure > ...

Wolfram is the most correct about AI person living imo