GaggiX

39 ms·

841.
iOS 18 AI boost could be called 'Apple Intelligence'(appleinsider.com)
2 points by GaggiX 2y ago|0 comments
842.
by GaggiX 2y ago
Yes, I sell better bullets.
843.
by GaggiX 2y ago
Go to the "Explorations of capabilities" and explore all the capabilities: https://openai.com/index/hello-gpt-4o/ You cannot have this level of control by prompting Dalle, also GPT-4o isn't using Wh
844.
by GaggiX 2y ago
An important aspect that is not considered in the article is that GPT-4o can generate images by itself (even though the feature is not enable to the public) meaning that it's very likely trained on sequential image tokens and the image
845.
by GaggiX 2y ago
>That’s about the only time I use or want to use “AI” on my phone ever – and I wish I didn’t have to do even that. I hope you have recovered from this truly hard pressing moment in your life.
846.
by GaggiX 2y ago
What's the hydrid in this case? Also isn't Apple going big on AI next week?
847.
by GaggiX 2y ago
I wasn't asking about running the model locally, also for that I have to wait for someone to quantize the model.
848.
by GaggiX 2y ago
Is there a demo or API to test the model? There are so many vision language models these days, it's hard to say which one is better, they also use in many cases different benchmarks.
849.
by GaggiX 2y ago
Going from 79.5% up to 84.2% is a -23% error rate for example, it's quite a huge difference.
850.
by GaggiX 2y ago
You can use it for speculative decoding for example, this would increase the speed of larger models.
851.
by GaggiX 2y ago
>that also makes good use of that full window?! To be honest the "Needle In A Haystack" test is the most trivial test for a model that relies on full attention, it's expected to be easy to pass if the model was trained cor
852.
by GaggiX 2y ago
>But you're a tech person. I'm trying to think of this from the point of view of e.g. a likely potential jury. Courts can call experts to testify on matters requiring specialized knowledge or expertise.
853.
by GaggiX 2y ago
Adversarial loss is used in many cases like when training a VAE, and a VAE can use a transformer architecture.
854.
by GaggiX 2y ago
Yeah that's right, I doubt that a model would generate an image or text so close to a real one to violate copyright law just by pure chance, the image/text space is incredibly large.
855.
by GaggiX 2y ago
I didn't say it takes an exact copy for copyright infringement.
856.
by GaggiX 2y ago
Having exact copies of the samples inside the model weights would be an extremely inefficient use of space, and also it would not generalize, unless it generated a copy so close to the original that it would violate copyright law if used, I
857.
by GaggiX 2y ago
If the model is first pre-trained on unlabeled images, then it takes about 10 labeled images of cats and 10 labeled images of dogs to train a (possibly strong) classifier (example: DINOv2), I doubt humans will do better.
858.
by GaggiX 2y ago
Since LLMs seem to follow the rules of grammar quite well, and this is quite the opposite of "sequential", I guess LLMs do not process informations very sequentially.
859.
by GaggiX 2y ago
Can you share the blog post?
860.
by GaggiX 2y ago
I use it all the time, simple example if you can't remember a word you can describe it to ChatGPT and it will respond with it.
861.
by GaggiX 2y ago
>hopefully this fad will die out soon Are you talking about ChatGPT? That seems pretty delusional.
862.
by GaggiX 2y ago
Well this heymusic sounds quite worse compared to Suno or Udio.
863.
by GaggiX 2y ago
The weights are available: https://huggingface.co/Doubiiu/ToonCrafter/blob/main/model.c... The repo: https://github.com/ToonCrafter/ToonCrafter There is also a demo: https:/&#
864.
ToonCrafter: Generative Cartoon Interpolation(doubiiu.github.io)
3 points by GaggiX 2y ago|1 comments
865.
by GaggiX 2y ago
Yeah, and it kind of died.
866.
by GaggiX 2y ago
NSFW has always been allowed on Twitter or you would have to erase half of the artist community.
867.
by GaggiX 2y ago
If you know Twitter already or it's like the opposite.
868.
by GaggiX 2y ago
Claiming extremely realistic deepfake is the all-in-one solution when a compromising photo/video of you is leaked online. Not specifically related to this case, but something similar has already happened at least one before: https:&#x
869.
by GaggiX 2y ago
The models tested are extremely small, a few thousand parameters and the performance is of course not great, I don't think we can extrapolate much from this. I don't understand why they chose such small models when you can train m
870.
by GaggiX 2y ago
Can it understand how you feel by the intonation of your voice, can it recognize the animal by the sound? If not, then it's probably not end-to-end, ChatGPT already had this mode for months where they simply use a STT and TTS to let yo

More ›