GaggiX

29 ms·

421.
iGenius Releases Colosseum 355B(igenius.ai)
2 points by GaggiX 2y ago|1 comments
422.
by GaggiX 2y ago
As always, I find it hilarious that some people believe that these companies will train their flagship model on uncurated data, and that text generated by a Markov chain will not be filtered out.
423.
by GaggiX 2y ago
There is also this TTS: https://github.com/rhasspy/piper that is pretty good (depending on the language) and extremely fast, would be cool to change the script to user Piper instead of Kokoro in case you want to use a
424.
by GaggiX 2y ago
Is there a reason to choose VGG16 over more modern models?
425.
by GaggiX 2y ago
I don't understand why you are so aggressive ahah, gibberish is easy to recognize I'm sorry, you don't need to be mad about it ahah
426.
by GaggiX 2y ago
An easy way is to give the model the URL of the page so it can value the content based on the reputation of the source, of course the model doesn't know future events, but gibberish is gibberish, and that's quite easy to filter, e
427.
by GaggiX 2y ago
The dataset is curated, very likely with a previously trained model, so gibberish is not going to do anything.
428.
by GaggiX 2y ago
Kinda happy that Evince doesn't start executing JS when opening a PDF.
429.
by GaggiX 2y ago
You don't need to wait, you can use: https://github.com/m-bain/whisperX right now for STT with timestamps and diarization.
430.
by GaggiX 2y ago
I wonder what STT model that is fast enough and accurate VLC could be use for the application.
431.
by GaggiX 2y ago
(2023)
432.
by GaggiX 2y ago
>local inference hardware (NPU on Arm SoC). Okay the battle is already lost from the beginning.
433.
by GaggiX 2y ago
Are you going to use OpenAI API or maybe setup a Meta model on an NVIDIA GPU? Ahah Edit: I found it funny to buy hardware/compute to only fund what you are trying to stop.
434.
by GaggiX 2y ago
I hate LLM companies, I guess I'm going to use OpenAI API to "obfuscate" the content or maybe I will buy an NVIDIA GPU to run a llama model, mhm maybe on GPU cloud.
435.
by GaggiX 2y ago
I imagine these companies today are curing their data with LLMs, this stuff isn't going to do anything.
436.
by GaggiX 2y ago
The model used is RWKV 169M v4.
437.
by GaggiX 2y ago
It would be cool to see these benchmarks on the newly released Jetson Orin Nano Super, like faster-whisper.
438.
by GaggiX 2y ago
Humans and AIs are different, the next benchmark would be build so that it emphasize the weak points of current AI models where a human is expected to perform better, but I guess you can also make a benchmark that is the opposite, where hum
439.
by GaggiX 2y ago
It would be really cool to have a model like this but multilingual, it would really help with things like moderation.
440.
by GaggiX 2y ago
Something very cool about the Georgian language is that its family, the Kartvelian family, has no known relation to any other language family. It's one of the world's primary language families. Its origin, including the alphabet,
441.
by GaggiX 2y ago
If only a mad man would make a Photopea/Photoshop clone open source, then everyone (who has the skills) would be able to not only use a decent open source image editor, but one that can be fully customized to your needs.
442.
by GaggiX 2y ago
Youtube captions have improved massively in recent years, they are flawless in most cases, sometimes a few errors (almost entirely in reporting numbers). I think that the biggest problem is that the subtitles do not distinguish between the
443.
by GaggiX 2y ago
>I want AI to do my dishes and laundry Buy a dishwasher and a washing machine.
444.
by GaggiX 2y ago
Even if memorization is not required, you still need to map every move on the board one by one under time pressure, an advance amateur would still make illegal moves.
445.
by GaggiX 2y ago
It has been for the last several months now.
446.
by GaggiX 2y ago
It often happens when the players play blondfold chess, as in this case.
447.
by GaggiX 2y ago
You should not finetune the models on the strongest setting of Stockfish as the move will not be understandable unless you really dig deep into the position and the model would not be able to find a pattern to make sense of it, instead I su
448.
by GaggiX 2y ago
I can confirm that an advanced amateur can play illegal moves by playing blindfold chess as shown in this article.
449.
by GaggiX 2y ago
>considering the size of the model. We don't know the size of Claude 3.5 Sonnet or any other Anthropic model.
450.
by GaggiX 2y ago
I was thinking about the same thing, also as an accessibility feature for your phone or software for your PC (I imagine this last one already exists).

More ›