novelty

@d6e.io

manic pubmed dream girl Location: cascadia

I really wish #nix had algebraic data types... Doing logic on strings is rough and it's really hard to know what values are accepted or not without building the derivation

What evidence do we have that intelligence actually scales exponentially and not logarithmically?

Bullying on social media seems to be a big problem. A small number of users seem to enjoy bullying. It would be neat if there was a labeler that did sentiment analysis on the last X replies a person made and measured how inflammatory they are. Then people could choose to auto block or hide them.

From my experience in Tokyo, if a location on Google Maps has more than 50 reviews, and the total number of reviews with less than 5 stars is less than 10% of the 5-star reviews, it’s likely that the reviews are manipulated or fake.

So like one cool thing we can do now: if there’s a class of statistical errors that occur in scientific papers, we can have bots that test for that. Prompt: “you are a p-hacking detection bot, read the following paper and assign a percentage score for p-hacking likelihood”

Recursively self-improving intelligence was based on the assumption that we lived in the Chomskian universe where AI is a very clever rule based algorithm. So far it seems we actually live in the Norvig universe where AI is a statistical model consisting of obscene quantities of compute and data

It would be really cool if there was a bluesky feed generator that used an embedding to generate feeds and you could specify a vector and get a custom feed of content close to that vector.

One of my big pain points with shell scripting is how it much it is implicitly dependent on the state of the file system and environment variables. The industry really needs some sort of static analysis tool which can predict the state of the file system and warn if a state mismatch would occur

One of the magical things about Japanese cities is that you can be in an area for days before noticing that there’s an entire supermarket beneath your feet

When Ai was starting out, we expected an intelligence explosion because technological development tends to be exponential. But what we’ve found in practice is that it takes exponential resources to get linear improvements in capabilities. This does not seem like intelligence explosion.

I actually wonder if mind uploading might be a flying car technology. An idea that is exciting for people but ultimately not practical and kind of bad in implementation.

I’m sitting in a cafe in #japan #tokyo waiting for my order and a guy came up to the clerk carrying a duplicate of the cafe’s custom doormat. The clerk gave him some money and then he swapped the old mat with a clean mat and left. Is this some sort of specialized custom mat service??