Maybe you can log all the traffic to and from the proprietary models and fine tune a local model each weekend? It's probably against their terms of service, but it's not like they care where their training data comes from anyway.
Local models are relatively small, it seems wasteful to try and keep them as generalists. Fine tuning on your specific coding should make for better use of their limited parameter count.
This is not true, 5G has multiple positioning improvements that are not related to higher frequencies. 5G has something called LMF (Location Management Function) that handles positioning of user clients through multiple means, like round trip time, angle of arrival, and dedicated 5G positioning reference signals.
i swore 5G used much higher frequencies (and is therefore blocked by so many more things that don't affect 4G and below.) I'm glad I'm wrong, thank you.
There are other affluent countries that doesn't do nearly as well, so there's more to it than that.
> Qatar’s EV Market reported an impressive surge, with YTD sales up to September up by 119.6%. However, it remains under 2% of total light vehicle sales, with demand still lagging behind. The government has reaffirmed its commitment to scale up EV adoption in the future, establishing the goal to reach an EV share of 10 percent of domestic sales by 2030
> In 2024, electric or plug-in hybrid cars made up 28% of new registrations in Switzerland (compared with 30% in 2023). This was the first setback for such vehicles after steady growth since 2015.
Even China has more incentive than Switzerland... I guess the issue is even with tax credits, the super-rich can still splurge and buy the gas-guzzling SUVs, and it seems it's a country full of G-Wagens, with people with money who need to show off.
In Shenzhen, the government made gas taxis have higher meter prices, so obviously passengers will pick electric taxis. E-scooters must be cheaper to run, so they're popular with all the delivery riders, and normal commuters there...
Norway has a slightly higher GDP (PPP) per capita than Switzerland, so saying it's just because they're rich enough to buy gas-guzzling G-wagens doesn't seem to be the answer.
Before taxes... And as AI just educated me, GDP is the monetary amount the country earns from its production, but Norway's oil earnings are heavily invested by the government.
Perhaps Switzerland is more known by the "filthy rich"/Nouveau riche crowd, and therefore attracts them...
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.
Local models are relatively small, it seems wasteful to try and keep them as generalists. Fine tuning on your specific coding should make for better use of their limited parameter count.
reply