tipsytoad

I couldn't find anywhere in the article, how long does it take to add all 10,000 embeddings to the index? That seems quite important for many applications

Dot product is equivalent to cosine similarly for normalized vecotrs

It's not one of the other. You can definitely use a mouse _and_ vim. I wouldn't recommend easymotion to anyone, a mouse is just so much better suited to that sort of behaviour. But there's plenty of motions much better suited for vim bindings, such as change the text up to the next bracket with ct( or changing the contents of a string in ci" and just e and b instead of alt arrows. I think people can just get carried away with "my hands will never leave the keyboard" but it's fantastic having the best of both worlds with vscode + neovim extension

But it don't tho?

https://chat.openai.com/share/3172fff1-db33-425e-8553-9bad0e...

I think largely because swift and mojo have the same creator, chris lattner. He was a guest on lex fridman last week

https://youtu.be/pdJQ8iVTwj8

I think it's disingenuous for the article to provide these prompts as evidence for their argument, yet use a subpar model with no comparison to gpt4. This article is only convincing in that it shows how far behind bard is compared to gpt4.

If you can give examples of tricking gpt4 in any of the ways in the article I'd be happy to hear it

Except none of these "tricks" work on gpt4...

https://chat.openai.com/share/419a7454-841f-4555-8188-40dbfc...

Why Not Just: Think of AGI Like a Corporation? https://youtu.be/L5pUA3LsEaw

Lol. Except the huge library of plugins, tight copilot integration and majority market share

I feel like this is like saying I don't use autocomplete because I still have to type...

I'm no expert, but I think it's so that the model can learn the relative position wrt other tokens.

They use indices for models like vision transformers with a fixed number of patches but for variable length context I think it's more beneficial to use encodings that can also capture the relative distance.

I think I misrepresented the point there a little. It's not the fundamental question per se but something he asks every one of his guests, and comes off alil 3am teenage philosopher.

I don't mean to bash him too much though, I think it's hard interviewing experts in areas they have spent their whole lives in and not to feel less informed in comparison.

Ignoring the arrogance that, the network effects of his podcast is incredible. I don't know of another podcast that gets the kind of guest list lex does. But you're right, his questions are often quite (pseudo-intellectual?) he's got a thing for "are we living in a simulation"

Aside: I usually use direnv to activate the venv (or poetry) when entering a dictionary

https://gist.github.com/tom-pollak/8326cb9b9989e0326f0d2e19f...

Nitpick: get started and pricing buttons don't collapse the burger menu

Also I assume I need an openai API key for this? I can't find that anywhere

Harmonic on android. Best by far from what I can find

I just found contexts and it's fantastic, it's just like alt tab but also comes with a spotlight esq quick switcher and supports vim jk by default. It's also really useful for switching between desktops by turning "reduce motion" on.

https://contexts.co/

I don't know why people think this is such a bad take, if we look at just gpt over the past few years gpt-2 -> 3 -> 3.5 (2019,2020,2022) is absolutely a 2x gain, and I don't think anyone's in a better spot to determine what improvements gpt4 will bring.

Humans set a high bar on intelligence but it seems inevitable that even with no significant changes we'll be overtaken, at least in breadth of knowledge

It's coming!

https://blog.rust-lang.org/inside-rust/2022/11/17/async-fn-i...

Although I'm still not sure how they return the unsized Future without the whole pin box method

Also, leaking memory with reference cycles from shared ownership

I'd convert that last tuple to struct, unlike the rest of the type it's undecipherable without some kind of documentation.

For normal people maybe, but for illicit activities an untraceable currency is incredibly useful. So not entirely useless :/

Lol true, and I'm currently working on a project leveraging a clip model which is why my answer is largely skewed towards vision transformers. By no means a complete list :)

Came here to say the same thing, but a few off the top of my head

  - attention is all you need
  - image is worth 16x16 words (vit)
  - openai clip
  - transformer XL
  - memorizing transformers / retro
  - language models are few shot learners (gpt)

A few newer papers

  - recurrent block wise transformers
  - mobilevit (conv + transformer)
  - star (self taught transformer)

:update is like :w but only writes when changes are made.

While we're at it :earlier and :later are undo/redo but instead of using an undo stack they use change points in time, so your undo history is never overwritten :)

Can I state the obvious? They look dumb as hell

I jump between neovim and vscode with neovim every few months. Neovim can be faster and more scriptable but vscode just works.

If you start using Neovim know you'll be spending hours breaking and updating your config to reach "vimvarna". Let me know if you ever get there ;) you can enjoy the journey getting close to perfect just for something else to break the next day.

It is crazy to me people try to shoehorn jupyter notebooks and debugging into Neovim, if you haven't try the jupyter notebook extension in vscode, will make you reevaluate your life choices

Fantastic article! How does this compare to ANN for the use case?

https://erikbern.com/2015/10/01/nearest-neighbors-and-vector...

Map holding caps lock to control and esc on tap, you never use it anyway :)

What about this code from the article

``` if cx.path == "/tags" { return tags::serve_list(cx).await; } else if cx.path.starts_with("/tags/") { return tags::serve_single(cx).await; }

    if cx.path == "/settings" {
        return settings::serve(cx).await;
    }

    if cx.path == "/search" {
        return search::serve(cx).await;
    }

    if cx.path == "/login" {
        return login::serve_login(cx).await;
    }

    if cx.path == "/patreon/oauth" {
        return login::serve_patreon_oauth(cx).await;
    }

    if cx.path == "/logout" {
        return login::serve_logout(cx).await;
    }

    if cx.path == "/debug-credentials" {
        return login::serve_debug_credentials(cx).await;
    }

    if cx.path == "/comments" {
        return comments::serve(cx).await;
    }

    if cx.path == "/latest-video" {
        return latest_video::serve(cx).await;
    }

    if cx.path == "/patron-list" {
        return patron_list::serve(cx).await;
    }

    if cx.path == "/index.xml" {
        return cx
            .serve_template("index.xml", Default::default(), mime::atom())
            .await;
    }

```