[-] artificialfish@programming.dev 2 points 1 week ago* (last edited 1 week ago)

RE: The biggest problem about this is the API definition. Libraries have APIs. But in a completely different way, webservers have APIs. If I say “we are going to a conference on API’s” what do you hear? That we are going to talk about REST, GraphQL, or gRPC, or web server APIs of some type. This is common phraseology. However, one time I was invited to such a conference, and it ended up being about C# design philosophy.

In that way API is an adjective (REST API, C# API, …) AND a noun (webserver API). That’s a problem.

And as an adverb, there is some justification to replace it with Library and others (REST Endpoint, C# Library)

Because otherwise all public functions are API’s, which doesn’t seem necessary to me. Saying a Library has an API is somewhat redundant. Saying a server hosts an API is not, many servers run jobs or databases. Many websites don’t host APIs. Etc.

[-] artificialfish@programming.dev 2 points 1 week ago* (last edited 1 week ago)

I know it’s difficult but just call a friend who is on facebook in the group for updates.

Enough people start doing that and the group will realize there’s a demand for off Facebook communication.

Even mastadon would be sufficient. But a signal group or even sms is enough.

[-] artificialfish@programming.dev 2 points 2 weeks ago* (last edited 2 weeks ago)

It’s not just a visual indication of if it’s encrypted. SMS sucks, truly, compared to apps like iMessage, WhatsApp, etc. so it’s actually annoying to message people with green text. Now that Apple does RCS it’s not a big deal, but in the USA there’s no default internet messaging app like WhatsApp, and to the extent that there is one it’s iMessage.

[-] artificialfish@programming.dev 2 points 2 weeks ago

That just looks like the field around a point charge…

[-] artificialfish@programming.dev 2 points 2 weeks ago

Oh I didn’t get that “no” element. Clever.

Old jokes can be hard to follow. I was expecting more idiom.

[-] artificialfish@programming.dev 2 points 3 weeks ago

And I was agreeing with you

[-] artificialfish@programming.dev 2 points 3 weeks ago

I was under the impression that’s just because of the relative surface area of the ocean vs arable land

[-] artificialfish@programming.dev 2 points 1 month ago

Actually now that I think about it, LLM's are decoder only these days. But decoders and encoders are architecturally very similar. You could probably cut off the "head" of the decoder, make a few fully connected layers, and fine tune them to provide a score.

[-] artificialfish@programming.dev 2 points 1 month ago

All theoretical, but I would cut the decoder off a very smart chat model, then fine tune the encoder to provide a score on the rationality test dataset under CoT prompting.

[-] artificialfish@programming.dev 2 points 1 month ago

Well I think you actually need to train a "discriminator" model on rationality tests. Probably an encoder only model like BERT just to assign a score to thoughts. Then you do monte carlo tree search.

[-] artificialfish@programming.dev 2 points 1 month ago

Nah, o1 has been out how long? They are already on o3 in the office.

It’s completely normal a year later for someone to copy their work and publish it.

It probably cost them less because they probably just distilled o1 XD. Or might have gotten insider knowledge (but honestly how hard could CoT fine tuning possibly be?)

view more: ‹ prev next ›

artificialfish

joined 1 month ago