peepeepoopoo114
peepeepoopoo114 on Hacker News
Re:Virginia passes law to enforce maximum vehicle speeds for repeat sp...
on Monday April 28, 2025 @05:45AM (#43819387)
It's a safeguard in addition to, not in replacement of, other measures.
Re:Virginia passes law to enforce maximum vehicle speeds for repeat sp...
on Monday April 28, 2025 @05:41AM (#43819358)
All democracies before the current era began as revolutions. Roman plebeians were well armed enough that the state could never become too abusive towards them. English democracy, and the entire modern idea of constitutional democracy itself, came about because the British public happened to be well armed enough with longbows, originally intended for times of war, that they could resist the tyrannical acts of the state and the professional military that it commanded. Some of the most peaceful and healthiest democracies in the world are also the most heavily armed: look at Switzerland for an example. The entire point of widespread civilian ownership of military weapons is that they can serve as a deterrent so that no tyrants, whether in the government or another private faction, can ever wield unassailable power over the masses, and that the weapons themselves never have to be used. Civil institutions can be captured over time by corrupt interests, but it's quite difficult to capture an empowered public.
Re:Virginia passes law to enforce maximum vehicle speeds for repeat sp...
on Monday April 28, 2025 @04:58AM (#43819105)
Just as mutually assured destruction brought lasting peace on the international scale, widespread civilian ownership of military weapons has also been a remarkably effective deterrent and safeguard against would-be tyrants across history. If both Karl Marx and the US founding fathers agreed that a well-armed public is important for a lasting civil peace, it's probably a good idea to listen.
Re:Inference-Aware Fine-Tuning for Best-of-N Sampling in Large Languag...
on Monday April 28, 2025 @02:09AM (#43818141)
Almost all of the efficiency gains have come from shedding bit precision, but the problem is that AI labs are now running out of bits to shed. The move to reduced precision inference has been masking the insane unsustainability of compute scaling as a model improvement paradigm.