> Also let me ask you why we need better and better and models if what we have already can produce good output with 'all the tooling to verify its hypotheses'

“Good” isn’t “perfect” and even if it was, the ability to produce perfect output with all the tooling to verify its hypotheses could still be improved, in time and token efficiency, by better models producing fewer spurious hypotheses, rejecting those it does generate faster, and taking fewer unnecessary steps in confirming its good hypotheses.

hmm.. ok but i dont think trillions of dollars are being spent for the improvements you mentioned.

Love this line in the critical material quoted there:

“Writing correctly in STE is not an easy task as it requires a good command of the English language together with a good knowledge of the matter of the writing.”

This is simply what is required for good writing (in English) in any domain with or without STE.

> Today, humans convert their labor to capital. Capital holders need labor (humans) to acquire more capital. When the price of inference for these robots becomes less than the price of labor then capital holders don’t need labor.

Ignoring quibbling about inference not being the only cost and other issues and just accepting the proposed end state: this is very good if capital is effectively democratized, and apocalyptically bad if it remains highly concentrated in a narrow class.

> this is very good if capital is effectively democratized

Are there any indications to think it's possible in our world?

I mean, there's a time in history when the population successfully brought out the guillotines and made real change against centralized power.

A time when covert organization and coordination of opposition were possible, as opposed to the panopticon of today, monitored 24/7 in real time.

Mean personal disposable income being comparable to median household income isn't too surprising and doesn't necessarily indicate any error. Mean is much higher than median, individual is much lower than household, disposable is much lower than total. You are comparing numbers that are different in all three axes (in ways which do not point in the same direction), so really any relationship between them would be plausible without detailed knowledge of the exact distribution of all the relevant components.

In finance and VC-funded -startup worshipping circles it is derogatory, because it means you aren't trying to scale like a VC-funded-startup.

That space is often confused with “tech”, and obviously overlaps with it, but its not the same thing.

“Didn’t forbid”, if referring to the framers of the Constitution, is a bit of an understatement, given that they created multiple layers of protection to prevent dismantling slavery at the federal level into the future.

Franklin also supported universal male suffrage, wrote Pennsulvania’s 1776 constitution which adopted universal male suffrage, and condemned property requirements for the franchise, reportedly arguing:

“Today a man owns a jackass worth fifty dollars, and he is entitled to vote; but before the next election the jackass dies. The man in the meantime has become more experienced, his knowledge of the principles of government, and his acquaintance with mankind, are more extensive, and he is therefore better qualified to make a proper selection of rulers — but the jackass is dead and the man cannot vote. Now gentlemen, pray inform me, in whom is the right of suffrage? In the man or in the jackass.”

The founding fathers, the framers (a different and later group, though the two overlap and are often conflated) and original states had a variety of (and evolving in some cases between the Founding and the Constitution) opinions on the issue; some (Thomas Paine and Ben Franklin, notably, the latter also the primary author of PA’s 1776 Constitution which also adopted this position at the state level) were in favor of universal male suffrage.

Any data that is consulted in the course of deciding how to use the capabilities of a system are instructions (they are a different and less privileged level of instruction than the set of instructions which provide the capabilities, unless one of the capabilities of the system is updating those privileged instructions, in which case that separation disappears.)

> Quick version: “abliteration” (basically removing the direction in the model that causes it to refuse) is the go-to method people use to make open models uncensored.

Tru-ish (lots of people distinguish between abliteration and uncensoring, though.)

> Most people treat it like a clean surgical cut - it just kills the refusals and leaves everything else untouched.

Basically no one does this, its widely recognized that this isn’t how it works and it has for quite some time been common for makers of anliterated model versions to publish metrics for how far a particular abliteration (1) removes refusals (typical before/after refusal rate on a standard test set), and (2) diverges to the output of the base model (KL divergence), and it is widely understood that there is generally, in practice, a tradeoff between these two metrics, where more refusal reduction tends to come at the expense of higher KL divergence.

That’s not saying that it isn’t interesting and new to characterize the kind of divergence that occurs with abliteration in different model families, but there is no reason for a late-night informercial level of misrepresentation of the existing understanding to come along with that.

Fair point - I overstated it. Thanks for the correction.