pyrodiscus

@noampomsky not a perfect fit but high fidelity maybe

incredibly niche ask but I have a memory of a young adult novel I read as a kid, it must have been written in the 90s or early 2000s, where Bruce Springsteen was a major figure. I think the main character was a girl and her dad was obsessed with Bruce Springsteen or something like that. I tried asking AI and it's not "Dear Bruce Springsteen" or "So Lyrical"

@SereneDesiree @jashvira It sometimes acts like it can see web results when it can’t, if it actually can there’s a UI icon thing

@SereneDesiree @jashvira Show full conversation? 👀

@jashvira I personally release packages after adding significant work so it feels like a fine metric to me. Maybe they're helpful for throwaway scripts but I find it encourages me to write more scripts when I really want just one that's reusable

Not sure if I see a strong correlation between pypi releases and programming productivity. Haven’t checked the distribution of kind of packages in a while, but I assume you’d expect a lot of hobbyist single script like stuff? Might not be optimal, but I for one, get codex/etc to one shot a huge throwaway scripts over and over, which a few years ago I’d wanna make a package for.

People tend to bullshit or overstate their confidence when they think you won't notice. This makes them untrustworthy on precisely those issues where you most want their input. For example, Daniel Kahneman on the fraudulent theory of priming in Thinking Fast and Slow:

DeepMind does have this to some degree (SynthID :))

It's been three years since Scott Aaronson invented a cheap and simple watermark for AI content that doesn't degrade capabilities at all, and which OpenAI and Meta (Instagram) choose not to use. scottaaronson.blog/?p=9333

youtube.com/watch?v=dHBEQ-… jasoncollins.blog/posts/re-readi…

This kind of untrustworthiness is a major characteristic of Claude and other coding agents, and one of the reasons why we see almost zero uptick in PyPI package releases despite tech twitter's fever dreams. It doesn't know how unless you do, but if you don't it will pretend to

Change in AGI timelines of AI researchers presenting at top conferences, 2022 to 2023, vs. Metaculus forecasters, 2020 to 2024. Over the 2022-2023 period Metaculus drops from ~18 years to ~10 years and the AI researchers drop from ~35 years to ~20 years (both ~43% drops)

"I think the community of people working on transformative AI are likely underrating types of work that need five or more years in order to pay off."

"Oh, so you are saying you have no idea when it will happen — it could be next year, or it could be 6 presidential terms from now. And you’re saying there is a 1 in 5 chance it isn’t even in that range."

(MacBook Neo)

Is Apple going to do this once their ultra low cost high processing power macbooks are rolled out? Do we need an open source effort?

Is this important? "Individuals need to be equipped with locally-running AI that is explicitly loyal to them, to help balance out dark patterns and threats [on the internet] . . . and there are natural markets to test these ideas further (eg. protecting people from scams)."

@Scholars_Stage A classic way to prepare for in-person exams is to memorize a pre-prepared essay. I would put those essays through pangram.

A professor tells me that over the last few years student writing—even in blue book exams which she has long administered—increasingly sounds like AI. She says this is especially true for ESL students. Increasingly *humans* are trained on AI writing, not the other way around.

I would like to encourage everyone to develop/enjoy peptides, relationships, the natural world, meditation suitable for lay (or nebulously lay-ish)people, exercise, learning and evolving, pace, R&D, mdma but as safe as coffee, peptides but for your brain, and japan

pausing AI would induce artificial scarcity and induce resentment in those who know what's at stake. accelerating AI could produce radical abundance that could potentially cash out in spaciousness and thoughtfulness

". . . radical abundance can’t eliminate competition, but can blunt incentives for existential gambles . . . "