liminal

Why is structured output a safety issue?

Is <blank> the only ML paper title?

Fast, cheap or good. Pick two. It seems like they've been prioritizing fast (lots of planes) and cheap (low staffing, outdated equipment) and paying lip service to good (safety).

Publishers are using AI to generate pages. I want AI to extract the worthwhile bits. Why shouldn't users have tools too?

Will your device measure hydrogen sulfide levels?

Buck2 doesn't seem to support JS/TS, so it's not ready for our use case.

We're trying to hire an Angular design engineer, so now the recruiter is the person trying to get management to switch to React to make his search easier. The tech doesn't matter too much, but the market forces do.

An issue I've noticed with court imposed fines is that they seem to be set based on the cost of the incident involved, not relative to the finances of the org/person being fined. A pipeline is very expensive, so fines end up being very expensive regardless of the receiver's ability to pay. OTOH, a large company can destroy lives with impunity since the cost of a person's entire lifetime earnings is not material to a multinational corporation's bottom line. This seems backwards to me. Fines should be set in such a way as to properly disincentivize behavior and adjusted to the means of the person/organization paying them.

This is a good way to ostracize people on other platforms.

I'd love to see the training data open sourced for all models so we can be sure no copyright material has been used. Just kidding, we all know it's stolen.

We just moved from Elastic to Quickwit. Sigh. What other open source, object storage backed logging databases do people recommend?

Is 14B parameters still considered small?

Lots of people have mentioned Polars' sane API as the main reason to favor it, but the other crucial reason for us is that it's based on Apache Arrow. That allows us to use it where it's the best tool and then switch to whatever else we need when it isn't.

This has already happened. We won't use software, open-source or otherwise, from an adversarial nation.

This is very cool, but would greatly benefit from a perceptually linear color space. I think the author is using HSL -- HCL would be a better choice.

People here don't seem to realize the importance of Canadian content rules in maintaining a cohesive national identity and supporting local content producers. The overwhelming barrage of U.S. media necessitates funding the creation of Canadian content, artists and points of view.

A writer with a long history of published works probably had their works in the LLM training data. This would lead to LLMs duplicating their writing and thus their new work being classified as written by AI.

Really wish you could filter by dataset attributes like language, number of rows, number of columns, etc.

Interesting choice making Gaza so prominent and not including Jerusalem

I'd like to see a weight tax on vehicles. Lighter vehicles have less tire wear, so less pollution. They also require less energy to move, so less emissions. Plus lighter vehicles would improve pedestrian safety.

The launch looked like a hostage video: https://www.youtube.com/watch?v=th3vzKTE0O8

I don't see the point of this acquisition. The products don't compete, so it's not eliminating market competition. The products also don't 'synergize' (ugh) well: they target different market demographics, have separate learning curves and different business models. Why?

> If you normalize your vectors, cosine similarity is the same as Euclidean distance.

If you normalize your vectors, cosine similarity is the same as dot product. Euclidean distance is still different.

Does this exploit require the page with the specially crafted link to be on the host secured by Keycloak?

Qdrant is great vector DB ...with the strangest hero image on their release announcement. A robot crab with a galleon in the background??

1. Yes, that's reasonable and saves running another DB

2. You often can perform the embedding in the DB, but there are a lot of use cases where you want to manage your embedding models outside the DB. This way you aren't dependent on which models the DB supports and you don't duplicate them throughout your system

Shouldn't the person/org filing the take down notice be required to prove they own the copyright they allege is being infringed?

[delayed]

My main issue with pretty much all queue approaches is that they don't work across platforms. They are built for one technology stack, be it Python/NodeJS/etc. This is fine if you've only got one stack, but in a microservices world it doesn't work where jobs can span multiple systems. You might be able to find some abandoned library that supports that queue tech on the other platforms you need, but now you've basically become a queue tech maintainer.