GaggiX

42 ms·

781.
by GaggiX 2y ago
>Tesla probably gets cells from the factory at round $80-$90/kWh. Where I live there is a warehouse where you can withdraw new MANYI Lifepo4 cells at around 97€/kwh (a single cell) after contacting the seller on Alibaba, so I&#
782.
by GaggiX 2y ago
Even non-tech people had big doubts about Glaze or similar (like nightshade) before the paper and this article were published, I think now only in total denial would you believe the claims of the Glaze team.
783.
by GaggiX 2y ago
I disagree, I don't see how a model that is trained with multimodality (input and output) and that can explore the environment will acquire creativity, I can see being more creative, but that suddenly these specific characteristics sho
784.
by GaggiX 2y ago
A human can be less or more creative, there is always some creativity, but for a machine it is controversial to say if there is any creativity at all. Going from not very creative to very creative seems to be a much easier task than creatin
785.
by GaggiX 2y ago
In the comparison between the models only Rodin seems to produce clean topology, hopefully in the future we will see a model with the strength of both, hopefully from Meta as Rodin is a commercial model.
786.
by GaggiX 2y ago
Food from McDonald's is as real as healthy food, it's food, there's nothing fundamentally different, just lower quality, if that's the point, then you're saying there's also nothing fundamentally different from
787.
by GaggiX 2y ago
If AI has no creativity then it shouldn't be able to compete in the slightest with actual humans, having creativity is such a fundamental aspect of creation that you would think that something that doesn't have it shouldn't e
788.
by GaggiX 2y ago
Information about our world is fundamental to create something recognizable for humans, otherwise it would be just noise, that doesn't mean you just copy the interface of the Apple Meteo app.
789.
by GaggiX 2y ago
The AI should be designing something new, not just spitting out the training dataset (because at this point, just google it). The problem here is that they are relying on third party models that they have no idea what they were trained on.
790.
by GaggiX 2y ago
Just to clarify this paper was released in April 2023.
791.
by GaggiX 2y ago
Some interesting technology behind the voice recognition: https://github.com/futo-org/whisper-acft
792.
by GaggiX 2y ago
Well good news you don't need lithium to build solar panels.
793.
by GaggiX 2y ago
>All optimizations inevitably end up just being funneled into larger and larger models. Well of course if you're trying to beat the SOTA, bigger sizes allow for better models, but not everyone is trying to use or train the latest SO
794.
by GaggiX 2y ago
I should have specified that by "runs directly in your browser" I meant running it locally on the machine. EDIT: You still have to rely on a server to index the web pages of interest.
795.
by GaggiX 2y ago
I believe that sooner or later we will see something similar to Perplexity that runs directly in your browser, opens the pages for you, and answers your question. Unless it has already been done.
796.
by GaggiX 2y ago
Not really, we know absolutely nothing about Claude 3.5 Sonnet, except that it's an LLM.
797.
by GaggiX 2y ago
>How do they know it's better? Probably just evaluation on benchmarks.
798.
by GaggiX 2y ago
It's written in the article, the critic makes mistakes, but it's better than not having it.
799.
by GaggiX 2y ago
>or is there another reason Claude Sonnet 3.5 is so much better at coding than GPT-4o? It's impossible to say because these models are proprietary.
800.
by GaggiX 2y ago
Yes of course, as always, it's very possible that just scaling solved the problem, but the fact that the model is so good makes me wonder if they actually did something different and pre-trained the model on image tokens as well.
801.
by GaggiX 2y ago
The incredible ability of Claude 3.5 Sonnet to create coherent SVG makes me wonder if the LLM was not just pretrained on text. Vision capabilities are usually added later using a vision encoder that does not affect the LLM's knowledge
802.
by GaggiX 2y ago
Still, the gCO₂eq/kWh is 10 times higher than in France when the sun is strong, and much, much worse when it's not.
803.
by GaggiX 2y ago
This is literally just the description under the video, no idea how it should have saved me an hour, more like 10 seconds.
804.
by GaggiX 2y ago
>and capitalise on it solely for themselves. It benefits millions of people who use these models every day, even the most powerful ones like GPT-4o and Claude 3.5 Sonnet are available for free to the public (with some rate limit).
805.
Adversarial Perturbations Cannot Reliably Protect Artists from Generative AI(arxiv.org)
4 points by GaggiX 2y ago|1 comments
806.
by GaggiX 2y ago
No this has nothing to do with OpenAI Q*.
807.
by GaggiX 2y ago
The limit varies based on demand, now with the release of the new model it's very limited, but it wouldn't be like this forever. Honestly, like I did with GPT-4 and GPT-4-turbo, I would just use the API on an alternative frontend.
808.
Gen-3 Alpha(runwayml.com)
5 points by GaggiX 2y ago|1 comments
809.
by GaggiX 2y ago
I remember seeing a video from WSJ about the drive-thru AI chatbot from https://presto.com/ being used in a fastfood restaurant, to be honest it works quite well, I wonder if McDonald's will switch to them. The service
810.
by GaggiX 2y ago
>Captcha seems to be the only solution Not for long.

More ›