GaggiX

25 ms·

451.
by GaggiX 2y ago
For more context: https://en.wikipedia.org/wiki/Denn%C5%8D_Senshi_Porygon The Wikipedia article also has the scene that you can watch at your own risk.
452.
by GaggiX 2y ago
Asking for counterarguments to these models is very useful when you're trying to develop an argument, especially to understand if there's some aspect of the conversation you've missed.
453.
by GaggiX 2y ago
The embedder is only 1.1M parameters, so it should run extremely fast.
454.
by GaggiX 2y ago
The car is perhaps a Citroën Ami.
455.
by GaggiX 2y ago
I think that for open models Flux models are considered the best ones. The interface that is used for these models is usually ComfyUI. For best model in general, I think that Ideogram 2 and Recraft model are the best one to use, recraft.ai
456.
by GaggiX 2y ago
I'm glad they understand that sometimes people want images that look like images, not overly saturated, high-contrast images that immediately look like AI-generated images.
457.
by GaggiX 2y ago
I'm talking about Thread, Bluesky does indeed allows porn, that's why many artists are moving from Twitter to Bluesky.
458.
by GaggiX 2y ago
Also no porn, since Twitter is the only mainstream platform that allows porn, a real competitor would have to allow porn as well.
459.
by GaggiX 2y ago
You cannot put "bastion of truth" and "X" in the same sentence.
460.
by GaggiX 2y ago
Some AI companies are going to win the competition. I don't think OpenAI is going anywhere.
461.
by GaggiX 2y ago
>AI image generation is complete crap It depends on the model, nowadays you can generate images that cannot be distinguished from stock photos, illustrations, anime images etc... I guess some people think it's bad because the only t
462.
by GaggiX 2y ago
Gary Marcus is writing this article every year so that one day he will be right.
463.
by GaggiX 2y ago
Maybe they are using Claude API for the translation, Claude models are really good multilingual models. EDIT: the "Limitations" section report the use of LLMs without specifying the models used.
464.
by GaggiX 2y ago
By announcing you mean a tweet from Sam saying "chat.com"?
465.
by GaggiX 2y ago
They just bought the domain, similar to what Google does with gogle.com.
466.
by GaggiX 2y ago
Honestly right now Anthropic offer a better service, at least if you are a programmer, I don't really see the moat.
467.
by GaggiX 2y ago
Also because of the limitations of this model to achieve something in AI Minecraft is completely different than the standard game, for example to go to the Nether is easier to find the red mushroom texture and use it to confuse the AI into
468.
by GaggiX 2y ago
I imagine they trained on data that is protected by privacy laws, similar to Meta.
469.
by GaggiX 2y ago
They probably trained on data protected by privacy laws, similar to Meta.
470.
by GaggiX 2y ago
ChatGPT does not have a watermark.
471.
by GaggiX 2y ago
This is the kind of stuff you can now simply ask Claude 3.5 Sonnet to build it for you and it will do it, no need to search for one.
472.
by GaggiX 2y ago
Zed allows you to use Claude 3.5 Sonnet for free unlimited under Zed AI, at least for now.
473.
Framer: Interactive Frame Interpolation(aim-uofa.github.io)
1 points by GaggiX 2y ago|0 comments
474.
by GaggiX 2y ago
>There's no way this tech will ever make a playable version of Mario Wait a few months, if someone is willing to use their 4090 to train the model, the technology is already here. If you could play a level of Doom than Mario should
475.
by GaggiX 2y ago
Kinda hyped to see how this model (or a much bigger one) will run on Etched's transformer ASIC, Sohu, if it ever comes out.
476.
by GaggiX 2y ago
They do.
477.
by GaggiX 2y ago
>That's not a capability that comes naturally to a probabilistic sampling of tokens. The linked page from OpenAI clearly show the opposite, read the "Using SimpleQA to measure the calibration of large language models" and
478.
by GaggiX 2y ago
yeah, this is by far the most interesting part of this page, the fact that LLMs can know what they know is not a trivial fact.
479.
by GaggiX 2y ago
These models are simply much more powerful than a tradition search engine and stackoverflow, so many people use these models for a reason, a friend of mine that never tried ChatGPT until very recently managed to solve a problem he couldn&#x
480.
by GaggiX 2y ago
The table only shows the models that they managed to beat, so there is no GPT-4o or Claude 3.5 Sonnet for example.

More ›