agreeing with the point the question makes here; the game theory of global politics does not work with the same morals that we prescribe individual people

Personally, I think everything should be hackable, however...

Limiting the ability to _easily_ modify what's running on a system is more about public cyber-health than the individual's freedom. Viruses + malware much more easily infect systems when they are running outside of a sandbox.

"We got GTA III on dreamcast before we got GTA VI"

"Don't make me tap the sign."

The sign: "Games are enterprise software now, with all the bloat that implies"

^ This and we need to be continually learning on an energy budget similar to how much a human spends per hour.

The main reason why we can't do that now is because we require models to be digitally reproducible (IMHO, but also read Geoffrey Hinton's mortal computing).

The energy cost come from error correction as much as training algorithms.

Can we extract the latent understanding it has?

Would be really cool to convert it's predictive model into a computer program that predicts written in like python/C/rust/whatever, and I think that would better serve our ability to understand the world.

We don't need to; dynamical meteorology is an incredibly mature field and our understanding of the fluid dynamics of the atmosphere grossly exceeds the resolutions and limitations of coarse, 0.25 degree global numerical models.

Agree with this choice. DNT didn't work, and now it's just a signal that could _help_ organizations track you.

I think this is funny, but reading the code also makes my head hurt.

A+

From the title, I wanted to dislike this, but Dan McKinley drops another banger slide.

Giving people the keys to the car is both 1. how you make a happy person and 2. build systems that understand and operate with the bigger picture

I think they spit out human-readable code, because they've been tried on human authors.

But you make an interesting point: eventually AI will be making for other AI's + machines, and human verification can be an after thought.

This reads solely as a sales pitch, which quickly cuts to the "we're selling this product so you don't have to think about it."

...when you actually do want to think about it (in 2024).

Right now, we're collectively still figuring out:

  1. Best chunking strategies for documents
  2. Best ways to add context around chunks of documents
  3. How to mix and match similarity search with hybrid search
  4. Best way to version and update your embeddings

(post co-author here)

We agree a lot of stuff still needs to be figured out. Which is why we made vectorizer very configurable. You can configure chunking strategies, formatting (which is a way to add context back into chunks). You can mix semantic and lexical search on the results. That handles your 1,2,3. Versioning can mean a different version of the data (in which case the versioning info lives with the source data) OR a different embedding config, which we also support[1].

Admittedly, right now we have predefined chunking strategies. But we plan to add custom-code options very soon.

Our broader point is that the things you highlight above are the right things to worry about, not the data workflow ops and babysitting your lambda jobs. That's what we want to handle for you.

[1]: https://www.timescale.com/blog/which-rag-chunking-and-format...

Points 2-4 are clear pointers to a real database as the home for vector data & search.