leodriesch

That's what I really like about Flux Kontext, it has similar editing capabilities to the multimodal models, but doesn't mess up the details. The editing with gpt-image-1 only really works for complete style changes like "make this ghibli", but not adding glasses to a photorealistic image and have it retain all the details.

I got SAM 1 to work with MPS device on my MacBook Pro M1, don’t know if it works with this one too.

I ran it with 3040x3040px images on my MacBook M1 Pro in about 9 seconds + 200ms or so for the masking.

I’d guess testing hardware is same as training hardware, so A100. If it was on a mobile device they would have definitely said that.

I think it’s fair to leave it out in the on-device model comparison. 3b is much smaller than 8b, it is obviously not going to be as good as llama 3 if they did not make groundbreaking advancements with the technology.

The model is always wrong, since it predicts a propability distribution over all possible tokens, but the target has 100% possibility for one token and 0 for all others.

I am really impressed by the Apple Maps implementation. I think it also uses textured polygons, but does so in a very good looking way and at 120 fps on an iPhone, showing even a whole city in textured 3d.

Brave Search also has one https://brave.com/search/api/

How does this compare to MLX? As far as I understand MLX is equivalent to PyTorch but optimized for Apple Silicon.

Is this meant for training MLX models in a distributed manner? Or what is its purpose?

I'd say most are thinking of Midjourneys success in image generation when talking about this kind of progress.

I’d say talent? Outside of OpenAI no team has been able to release a model as capable as GPT-4, and I’m unsure if the CIA has been prioritizing LLM experts in their hiring.

Could you link me to a finetune optimized for function calling? I was looking for one a few weeks ago but did not find any.

The purple icon in the shared transcript indicates GPT-4.

It’s a movie about a humanlike personal AI companion.

https://m.imdb.com/title/tt1798709/

Their privacy policy lists OpenAI as one of their partners for data processing, which indicates that this is happening not on your device, and data is also shared with third parties.

For me this is the main counterargument against apps like these. I want to feel free to post any information into this without thinking about who may read or use it.

Local is the only way to go for software like this in my opinion.

This is just a prompting hack you can use with any LLM, not exclusive to Claude. But I do like the fact that they include these tricks in their documentation.

For an AI agent that should navigate a computer (which is Adepts use case IIRC) it should work, as it only has to output commands.

Support for web extensions as an example (Safari extensions)

I don’t think this would suffice since the regulations would apply to EVERY iOS device, not just a specific iPhone.

I am wondering how much value social network data like Twitter or Discord really has when compared to something like Wikipedia or textbooks.

It is mostly unformatted, contains slang and is often far from the truth. Is this data really all that useful for training an LLM?

This is a difficult question I think independent of the restrictions that OpenAI imposes on GPT-4.

The model does not know what it knows, that’s why it sometimes hallucinates instead of saying it doesn’t know. But to answer the latest event it knows, it has to know which events it knows.

macOS native dictation is not as good as Whisper in terms of accuracy, however that is probably going to change with macOS Sonoma since they will switch the model for speech recognition to a better one (Transformer based iirc).

The linter plugin for hook dependencies is pretty great, I very very rarely need to specify a dependency array different from the one the linter suggests.

Iteration speeds, I could see myself using a slower framework to ship faster.

See usage of React Native for example, it certainly produces apps worse than native ones, however many businesses still choose to work with it because you can improve the developer experience and ship faster.

They seem to have closed down comments on it, I'm seeing

> An owner of this repository has limited the ability to comment to users that have contributed to this repository in the past.

Not a native speaker, but I read that as it is here to stay, not that it is pointless.

I'm interested in how the transformer based speech recognition from iOS 17 will perform compared to Whisper. I guess it will work more "real-time" like the current dictation on iOS/macOS, but I'm unsure as I am not on the beta right now.

It is fairly normal to build web servers in a way so they can be scaled horizontally (more instances rather than larger instances). So they can just have more containers run their Django servers and distribute the load between them.

Eye and hand tracking are supposedly the only way to interact with the device, so unless your app is not interactive at all, you will use them.

OP works for Vercel so I assume they had early access.