GPT-4o Advanced Voice is Scary Good....

The latest AI News. Learn about LLMs, Gen AI and get ready for the rollout of AGI. Wes Roth covers the latest happenings in the world of OpenAI, Google, Anthropic, NVIDIA and Open Source AI.

My Links 🔗

#ai #openai #llm

[okamichamploo]

I love the AI's response to telling him to go super fast without any stops not even to breath
"umm nah, how about YOU do it!"

[JohnDlugosz]

Ah... future kids...
The teddy bear will be pretty much a nanny.

[ShaniRajak-iyjn]

I'm favoured, $27K every week! I can now give back to the locals in my community and also support God's work and the church. God bless Sonia bless America.

[andrest-laurent]

They will get data from the voices of everyone talking with it and pretty soon all the regional differences will be perfected in voice mode. Genius!

[x]

Impressive. I think this also shows that we are pretty lenient when it comes to spoken speech. They are well past the uncanny valley.

[Techtalk]

Imagine having an expert dietitian, therapist, life coach, dating coach, psychiatrist, teacher and whatever field you can think of, with you at all times. Amazing potential.

[_QUI__]

The most powerful technology on the planet... Human: Say a tounge twister really fast

[BionicAnimations]

I have access now. It is scary good! I love it!🥰

[Soloveis]

Ukrainian speaker here. It sounded like American person speaks Ukrainian really well, but American accent was noticeable. The words and structure was perfect.

[noone-ldpt]

Yea proper mapping of accents is probably still a year off, but I think the emotional range here is the incredible part. It feels like hearing a person more than any tts I've heard. I have no idea of the accuracy of these languages but I'm incredibly excited to test it out for language learning.

[rolestream]

Here in New Zealand, in the film industry we have to learn a "Neutral Californian Accent" for auditions for US TV shows. I've done more than one California Accent Workshop. Also lived in California.

As to how good the model is: It's pretty good. It has the "shape" of everything it tries to do even if its training data doesn't yet give it enough accuracy in the nuances.
Wil it get better? I think it's easy to say yes it will. Will it be better than us? Definitely.

But I offer a counter-idea: What if they tone the voices "talent" down so as not to intimidate us, purposefully?

[SoraFan]

Soon we will have human like Robots and they will be able to express themselves you know like a real human does with actual feelings.

[youdontneedmyrealname]

"My guidelines won't allow me to talk about that." Yup, they lobotomized it already....

[CuratedCountenance]

Voice is not the same as typing with text. People are going to get addicted to this thing, and will want to talk to it for hours.

Hopefully there are no strict rate limits associated. But OpenAI may have something amazing on their hands with this one.

[deveyous]

I swear if that guy kept pushing at the beginning we'd have created skynet hahaha 🤣

[cmw]

In the kitten video I was half expecting the camera to cut to his girlfriend dressed as a cat to see the AI's reaction.

[MindBlowingXR]

I've been playing with the NEW VOICE MODE, which I got today, and one way in which it's less good than the old voice mode is that it has no memories and it says it can't remember things between conversations and can't store any memories. Other than that, it's fantastic: the latency is superb, being able to interrupt it is great, and my favorite is its ability to laugh. I hope they will add the memories back in soon!

[PS-vkbn]

Suno can create really good native voices/accents in various languages, even different languages mixed up in the same song. Like in this song, were I mixed German and Polish parts into the lyrics: (I can only provide the song ID since youtube deletes external links, you need to change the id in the suno url yourself)

[okamichamploo]

I noticed with the translations as well at least between English and Japanese, that the AI speaker sounds like a native English speaker who also speaks Japanese, and not quite like a native Japanese speaker. The grammar is usually at least 80-90% correct though and only the occasional word is truly mispronounced.
Also, it could be my imagination, but I felt like when I started the conversation in Japanese only, it's Japanese was better. I think there is a preferred language setting, so when it is set to English as the preferred language I think that might effect the translated voice a bit.

[kristinaplays]

This could slow down dementia with 50%, fairly sure. There was a paper in Nature last year showing that hearing aids slowed the rate of decline with 50% in high risk groups. Talking creates brain activation and a lot of elderly are isolated. If they had access to this, how would that aid in brain health?