0x008

so if it does not enforce a particular transport layer, it just defines a couple of methods on top of jsonrpc?

or is there more to it?

I think the idea with LLMs leading to AGI is more like:

Natural language is a fuzzy context aware state machine of some sorts that can theoretically represent any arbitrarily complex state in the outside world given enough high quality text.

And by reiterating and extrapolating the rules found in human communication an AI could by the sheer ability to simulate infinitely long discussions discover new things, given the ability to independently verify outcomes.

Natur is actually designed really well, it’s a just a shitty joke for our race and the ones who have to give way for nature to run its course.

I think what they mean is that the fundamental question is IF any intelligence can really break out of its confined area of expertise and control a substantial amount of the world just by excelling in highly verifiable domains. Because a lot of what humans need to do is decisions based on expertise and judgement that in systems follows no transparent rules.

I guess it’s the age old question if we really know what we are doing („experience“) or we just tumble through life and it works out because the overall system of humans interacting with each other is big enough. The current state of world politics makes be think it’s the latter.

This model is open source and Beats all proprietary models in benchmarks. How is this stagnant?

If you train a model on tasks in a similar format it's even less indicative of "AGI".

> OpenAI's new o3 system - trained on the ARC-AGI-1 Public Training set…

Sounds fishy to me

Why are we posting this now? I think uv, ruff, and astral have been posted here so often. Is something new here that I am missing?

Of course your frame of reference is biased because you have the best railway system out of ~30 European countries.

Germany’s system gets a lot of hate but it still in the top ~8 or so.

Is this the modern version of modifying the sys.path in a python script?

And the supposedly 24/7 defrag twitch stream

Apple forces you to use Apple Music though :/

They have the highest number of registered and concurrent users of all LLM providers, the best apps and they have deals with Microsoft AND Apple.

So I would say they have a solid grip on the market, and a good position to keep it with minimal improvements.

Most users don’t care about the best LLM, they will not switch. I mean they probably are not even aware they are using OpenAI because it is just integrated into Office or Siri…

The only benefit this could provide is time warrior integration.

You can use custom ranking functions (shown in the article) add weights to columns, etc.

What techniques are you talking about that cannot be implemented in Postgres?

But tsvector supports all sorts of different languages, at least western.

What is this non-package mode you are referring to?

I was talking about the maintainers not wanting to include features a lot of the userbase would like (like monorepo stuff), because they are saying their target audience is package authors, while in fact most of the users aren't.

I think its not always that people want to feel important. It's more like the subtle change perception: Whenever you are part of a discussion, you have the feeling that it's going better, just because you are there.

For me this sounds a lot like a distinction between being intrinsically motivated (“because it is fun for me”) and extrinsic motivation (“I want to achieve something in the outside world”)

Just to give you maybe another perspective: I think part of what makes the work interesting is that I can’t do it all day as I please. So when I have time to work, I am super motivated. As unsatisfying as that might be.

Whenever I have holidays and weekends or unlimited time and want to work on something I usually don’t follow through.

These pointless meetings are usually a result of the fact that people feel things are going better if they are a little bit involved, even if their involvement is not actually contributing to the solution.

Also it’s a lot about lack of trust in large orgs.

Poetry support for mono repos is really horrible and all the maintainers are saying is „it’s a tool indented to publish packages, not manage your dev environment“.

Offsetting carbon emissions - what does it even mean? There is no feasible way to remove carbon from the athmosphere as far as I know. Would you care to explain what you mean by that?

I think we need to be very careful making blanket statements about the usefulness of certain LLMs.

In my experience for every task another LLM is excelling and where one was good it might fail for the other task. They can do great things, but it’s not guaranteed and a lot of manual intervention and back and forth is still needed.

We are not at the point where using AI in the company ist just a blanket win for everyone involved. Companies are investing a lot but the return is hard to measure and not always guaranteed.

This is the problem with early technologies, they work sometimes but not guaranteed and we build our expectations on extrapolating their usefulness. We should not judge this technology by the current success rate, but rather by how much impact it will have once we get the success rate to higher and higher levels.

Still, what we can say is that for certain occupations it already helps them reduce their work by 15% (software engineers) and probably even more for some (writers, product owners, office warriors and alike). This is a great achievement in of itself, think how much this will make up in a company as large as MS or Google.

why do we have this sensationalist click-baity stock-market focused article on the HN front page?

this article must be written with the intention to troll HN

I can recommend to try poetry. It is a lot more succesful in resolving dependencies than pip.

Although I think the UX of poetry is stupid and I do not agree with some design decisions, I have not had any dependency conflicts since I used it.

probably they are hoping for some kind of statistical covergence as the number of samples in an area goes to infinity?

The Embedding model is only involved in the "R" part (retrieval). You need to use the same embedding model for indexing and retrieval, though, that is important.

The Embedding model is independent of the actual LLM used in generation. You can use any embedding model with any LLM. In fact ada is only one option and there are lot of really good embedding models, also ones you can easily run local, or ones which are a lot of better than ada, readily available. ada is not the best choice in all cases.

What can OpenAI realistically do about it except block the user account?

I think the TOS should have no effect on the license and copyright of the resulting model, since outputs of AI Models cannot be copyrighted (at least in the US).