Would this enable a model to learn concepts in one language and generate answers about it in another, as long as it learns general translations between them?
My educated guess:
Not more than any other LLM.
The text-latent encoder and latent-text decoder just find am more efficient representation of the tokens, but it's more of a compression instead of turning words/sentences into abstract concepts.
There will be residuals of the input language be in there.
I don’t think for this approach it sounds like, this is related to the large concept model: https://arxiv.org/abs/2412.08821, where the latent space is SONAR, which is very much designed for this purpose. You learn SONAR embeddings so that every sentence with the same semantic meaning gets mapped to the same latent representation. So you can have e.g. a French SONAR encoder and a Finnish SONAR encoder, trained separately with large scale corpi of paired sentences with the same meaning (basically the same thing you would use for learning translation models directly, but for SONAR you don’t need to train a single model per pair of languages). The LCM then works in this language-agnostic SONAR space which means it does (in principle) learn concepts from texts or speech in all supported languages
Im not sure how the depth estimation alone translates into the view synthesis, but the current implementation on-device is definitely not convincing for literally any portrait photographs I have seen.
True stereoscopic captures are convincing statically, but don't provide the parallax.
Good monocular depth estimation is crucial if you want to make a 3D representation from a single image. Ordinarily you have images from several camera poses and can create the gaussian splats using triangulation, with a single image you have to guess z position for them.
For selfies, I think iPhones with Face ID use the TrueDepth camera hardware to measure Z position. That’s not full camera resolution, but it will definitely help.
That's how EU's digital wallet is supposed to work:
> The selective disclosure of attributes will allow you to only share the specific information requested by a service provider, without revealing extra information.
> For example, with the selective disclosure of attributes you could choose to share your date of birth, but without revealing any other identifying details that could be used for profiling.
You want to know what the global impression of the US is right now? Here's a translated quote from a newspaper today, from a source in our military:
> – The US has the most qualified intelligence organizations in the world at its disposal. Both the CIA and the FBI have been politicized under the current regime. I find it difficult to see how we will be able to maintain the trusting cooperation we have had with the US in the past after this.
The actions of the current administration speaks far louder than any font ever could, and it's tearing down decades of good will and trust.
> AN OFFICIAL WEBSITE OF THE UNITED STATES GOVERNMENT
> What's the biggest brand in the world? If you said Trump, you're not wrong.
This is beyond satire by now, it reminds me of Idi Amin and his official title:
His full self-bestowed title ultimately became: "His Excellency, President for Life, Field Marshal Al Hadji Doctor Idi Amin Dada, VC, DSO, MC, CBE, Lord of All the Beasts of the Earth and Fishes of the Seas and Conqueror of the British Empire in Africa in General and Uganda in Particular"
5G does not mean shorter waves/higher frequencies, that's just a common deployment. In Sweden we have 5G on the 700 MHz band, 5900 MHz, and several others in between.
reply