leodriesch

GitHub Copilot X [0] seems to be exactly that, it is in technical preview with a waitlist.

[0]: https://github.com/features/preview/copilot-x

TypeScript is a superset of JavaScript.

I’ve also recently been able to create a Google account with my own email domain, from Germany.

This will probably also kill any R&D in that space, because why would you invest in searching for a better charging port only to have it denied by the EU.

I think the privacy aspect is also interesting, think about journal or second-brain type applications. These could benefit a lot from language model use, but you don’t want these types of information sent to a cloud provider in unencrypted fashion.

Is Uber genuinely gathering driver data? They're only managing a fleet, and it seems unlikely that each vehicle has cameras installed. Wouldn't the autonomous driving tech actually come from collaborating companies instead of Uber directly?

Be wary that some implementations of the API use server-side transcription and Firefox does not support it at all.

OpenAI also released the weights for Whisper. Some model releases, like LLaMa from Meta, just contain the code for training and the data. You can train the weights yourself, but for LLaMa that takes multiple weeks on very expensive hardware.

(LLaMa also released the weights for researchers and they leaked online, but the weights are not open-source.)

When a company releases a model including the weights, you can download their pre-trained weights and run inference with them, without having to train by yourself.

Amazing AI by Sindre Sorhus uses Apples port of Stable Diffusion to the neural engine.

https://sindresorhus.com/amazing-ai

Not an expert on the matter so take this with a grain of salt, but I’d say the compression is also about VRAM/RAM, which seems to be the more limiting factor over inference speed.

I'd say Inter is pretty close to Helvetica Neue or San Francisco, the system-ui font for macOS and iOS.

So I'd say the Neo-Grotesque style. But that uses Arial on Windows which I find pretty awful.

While AI gets better and better at creating what I would call "creative output", e.g. poems, texts of any form really, imagery and videos, I think the human skill it takes to produce these becomes less valuable.

In the future I imagine you'd no longer have to be good at writing poems, you'd just have to be good at distinguishing a "bad" poem from a good one. "Bad" is obviously highly subjective in this context.

So it becomes more and more important to have what I would call "good" taste, not the skills to do creative work yourself.

I think Chess is an easier thing to be defeated at by a machine because there is a clear winner and a clear loser.

Thinking, reading, interpreting and writing are skills which produce outputs that are not as simple as black wins, white loses.

You might like a text that a specific author writes much more than what GPT-4 may be able to produce. And you might have a different interpretation of a painting than GPT-4 has.

And no one can really say who is better and who is worse on that regard.

They show an example of this under "Training Recipe". They prompt Davinci to give them instruction following examples, 20 at a time. And that a lot of times. The context window is also 4000 tokens, at least that is the maximum amount you can pick in the OpenAI playground.

On the page under Developer Focus it says:

> Data submitted through the API is no longer used for service improvements (including model training) unless the organization opts in

I am wondering how this works out for YouTube with Premium. It removes ads and is 12€/mo.

I mostly read HN on my iPhone using the website with the zoom cranked up a little bit and I find it works better than the variety of native clients available.

Perhaps it won't take information from the training data but rather take the content from the pages a regular search would return and convert it to a nice answer to your query.

This seems to be the approach https://www.perplexity.ai/ is taking.

Compression makes it blurry to the point it’s unreadable on my phone, could you post the original text?

The readme does not seem to be geared towards people not familiar with the topic.

My questions:

- Is this on the run on consumer GPU scale, or run on 8 A100 scale or you can’t run it yourself ever scale? - How does it compare to other language models in quality/abilities? - What is the training data?

So rendering the HUD locally and a 3d scene on a remote server?

In my experience Safari is the best browser on iOS, because it is the only one that supports extensions.

It also has the best UI/UX and speed of any browser on iOS, but that’s subjective.

Framer Motion automatically corrects box-shadow and border-radius [0], this article leaves this details I guess.

[0]: https://www.framer.com/docs/layout-animations/###border-radi...

If the neighbors are also Motion components you would just have to "ping" them so they can check for layout changes and animate accordingly.

That would measure the component only on mount, so it would not notice if it’s layout changed.

You need to run it without the dependency array or add all layout-changing props and state to it.

You can enable a classic titlebar (at least on Windows, can't test on other systems right now) by right clicking next to the URL bar and choosing Customize Toolbar. Then theres an option to enable the title bar in the lower left corner.

I'd think about it as a "Linux" car, in the sense that you'd have an open source operating system for cars, the manufacturer would just supply drivers for the hardware (like graphic cards drivers).

GPU drivers for NVIDIA graphics card have been a bit of a hassle for me, I'd love to see a version with (proprietary) NVIDIA drivers included, like Pop!_OS provides.

Question was out of interest, not meant as a discouragement. Sounds nice, will check it out!

How is this different from Cheatsheet?

https://formulae.brew.sh/cask/cheatsheet