Kyutai Labs July 3rd multimodal Moshi model announcement compact

No video

Kyutai Labs July 3rd multimodal Moshi model announcement compact

Рет қаралды 6,338

multimodal ai art

Күн бұрын

A compact 10min version of the 50min live announcement: www.youtube.co...

Пікірлер: 30

@arturhellmann9138 Ай бұрын

the thing is, sometimes the response time is faster than that of a human... That makes it feel strange sometime.

@dazdazfzf Ай бұрын

Exactly

@frankpoe7272 Ай бұрын

Yes

@piedraangular9190 Ай бұрын

Europe start to compite. Great news!

@disabledlifestyle Ай бұрын

hello very interesting. I am totally blind and I would love to use this technology. Does anyone know when will this be available for iPhone or Windows computers and how much it’s going to cost to be able to talk to the certification intelligence thank you.

@TheidiotAmongUs Ай бұрын

It looks like they will release the models open source so it will be free.

@MrKohlenstoff Ай бұрын

@@TheidiotAmongUs While this is true, I'm not sure if the model will be lightweight enough for people to run it on their own devices locally? Because if it requires hosting in a data center, then the open source nature does not necessarily help end users to get free access to the actual working AI (although it still makes that more likely to happen in some way). But not sure what is the case here.

@ssw4m Ай бұрын

They did say that it can run on device. I'm not sure what the hardware requirements will be, maybe GPU or just a modern CPU with enough RAM.

@disabledlifestyle Ай бұрын

@@ssw4m thank you friends for the replies now I wonder when will be available to try it out. I am excited about this.

@bitodin Ай бұрын

@@disabledlifestyle Have you tried ChatGPT audio conversations features?

@KoreanBunny1 Ай бұрын

특이점이 온다!!

@SuperMelon2008 Ай бұрын

Take THAT United States! (said with heavy french accent)

@wisterier Ай бұрын

Amazing

@ibrahimkamissoko3968 Ай бұрын

It's so Amazing

@bigdaddyb04 Ай бұрын

Sounds too robotic. Atleast it’s available tho unlike ClosedAI

@lte6983 Ай бұрын

Tried the demo and it’s pretty bad. Got stuck repeating the same thing while interrupting and ignoring me

@yannickhs7100 Ай бұрын

Voice is good, but underlying model is like GPT-2

@dazdazfzf Ай бұрын

rag dispo ?

@User-actSpacing Ай бұрын

Is this demo fake? AI is responding even before person finishes their question. Feels prerecorded.

@agnesslovehealz Ай бұрын

Actually others done it too live and just like that try it and see

@CappaBeta Ай бұрын

It's replying to the first half of the sentence, it's so fast it has a habit of interrupting. Others have demonstrated this live, it's not fake. You can also try it. It's also not nearly as intelligent as the latest text models, due to its compactness and latency requirements.

@maitrecontroleprincipal1704 Ай бұрын

Ai anticipation

@ki-music Ай бұрын

Jonas Andrulis wird heulen

@user-dn8cw8oc6s Ай бұрын

Du BUZZ.. Rien de neuf avec une compilation de technos existantes et matures!! chatGPT-4o est plus fluide..

@Nutricare-yr3fe Ай бұрын

It's just the beginning, bro

@Nicolas-dy6bk Ай бұрын

Du BUZZ peut être mais au moins la techno est bien réelle et déjà en ligne à pouvoir être utilisée. Ce n'est pas parfait mais au moins ils ont le mérite de la rendre disponible gratuitement. On attend toujours GPT-4o Vocal et Sora...

@cyrilyon Ай бұрын

@@Nicolas-dy6bk Chaque conversation est limitée à 5 minutes, puisque Moshi se met rapidement à raconter n’importe quoi (le LLM qu’il utilise, Helium, a aussi été développé par Kyutai).