I think the paper is really good and makes loads of valid points... and it's kinda terrifying.

Having super accessible machines that can make anything up and aren't held accountable run the world is going to break so many systems where truth matters.

▲

pizzafeelsright 5 hours ago | parent [-]

> Having super accessible machines

Having large bureaucratic organizations

> that can make anything up and aren't held accountable

that run everything and aren't held accountable

> run the world is going to break so many systems

run the world breaking up families, freedom, and fun

> where truth matters.

where truth is determined by policy

	▲	gbanfalvi 5 hours ago \| parent [-]
		> Having large bureaucratic organizations Yes, that's what the paper argues. Institutions at every scale (say, doctor's clinics, hospitals, entire healthcare systems) are very challenging to access compared to me asking ChatGPT. And it's not just the bureaucracy, but there's time, money and many other intangible costs associated with interacting with institutions. > [Large bureaucratic organizations]that run everything and aren't held accountable But they ultimately are. People from all types of institutions are fired and systems are constantly reorganized and optimized all the time. Not necessarily for the better -- but physical people are not black boxes spewing tokens. Individuals' choices are ultimately a product of their knowledge and their incentives. An MLM's output is the result of literal randomness. > run the world breaking up families, freedom, and fun There's lots of terrible institutions vulnerable to corruption and with fucked up policies, but inserting a black box into _can't_ improve these. > where truth is determined by policy The truth is the truth. Regardless of what policy says. The question is, do you want to be able to have someone to hold accountable or just "¯\_(ツ)_/¯ hey the algorithm told me that you're not eligible for healthcare"