Cory casually adopts the latest slopaganda
It's possible for some of the code produced by an LLM to be slop
Cory casually adopts the latest slopaganda
It's possible for some of the code produced by an LLM to be slop
Technically true, in that it’s the only move he’s ever had
Just as AI and quantum computing that doesn’t suck is always just a little bit off in the future, data centres that churn through horrendous amounts of water are always a thing of the recent past.
Have I misunderstood what they meant by ‘superexponential’? I thought they were doing exp(exp(t)) but I am inferring that it’s somehow even sillier than that?
I dare say that’s why they shouldn’t be confused.
effective journalists
I have gazed too long into the abyss, and interpreted this as some kind of effective altruist
Effective journalism: always be sure to talk up the billionaire class for the dividends to society these titans of industry yield. Wouldn’t that be crazy.
Not literally an asymptote, but once you presuppose something as farcical as nested exponentials, your model is doomed to blow up immediately after the last data point.
I expected the beneficiary would be some EA outfit, but it’s actually an entirely new company (and not any kind of charity or nonprofit) called the “Anthropic Public Benefit Fund”, which I’m sure will not at all be managed by EAs (and who would definitely not determine that the most effective altruism is sending money to Anthropic).
I like this:
Subject to stockholder approval, 5% of Anthropic’s shares will be committed today, with an additional 2% added at the board’s discretion for every 50% increase in the company’s value above a set baseline.
You like the public benefit, don’t you? Make sure you invest strongly in Anthropic and pump that valuation even higher, because there’s no other way we could choose to offer more support.
Anyway here’s their blog post on the matter, strangely not linked in the article as far as I can tell: https://www.anthropic.com/news/public-benefit-fund
That ‘superexponential’ thing annoys me so much. Just the most blatant ‘the line goes up because I really really want it to’, while all their rationalisations for it are just describing exponential behaviour.
We talk about the importance of humanities but it’d be nice if they learnt a bit of maths too.
I am saying that the anti-fascist communities ended up […] uninterested in using LLMs
:O
Some descriptors for ‘democratised’ software that people can slop for themselves:
Slop as single-use software
Slop as fast fashion
Once again, slop machines only work ‘well’ when you can externalise the pain and costs onto someone else (in this case, millions of stressed-out sysadmins and webmasters)