INFP: Audio-Driven Interactive Head Generation in Dyadic Conversations | Hacker News Reader

INFP: Audio-Driven Interactive Head Generation in Dyadic Conversations

grisoon.github.io

29 pointsnnx21 comments

Amazing.. unfortunately code and data link back to the project page. But amazing research.

So much grumpiness! This is amazingly cool. The out of band (paintings) look great. Lots of good preemptive facial expressions: people clearing their throats to interrupt.. really awesome.

haha wow they’ll stop at nothing to sell this “agent” crap in generative “AI”.

I wonder how many years it’ll take till the VCs // general public finally realize the marketing sham they have fallen for and the bubble will pop.

Nevertheless, cool project in theory but they say

“Imagine having a conversation with a socially intelligent agent. It can attentively listen to your words and offer visual and linguistic feedback promptly. This seamless interaction allows for multiple rounds of conversation to flow smoothly and naturally. In pursuit of actualizing it, we propose INFP, a novel audio-driven head generation framework for dyadic interaction.”

Do people actually hold conversations with chatbots enough to make this a viable thing? I know this is just the code but it is made by ByteDance so they definitely have a monetary incentive to make it and of course want a ROI, 1. do enough people hold conversations with bots to make this worthwhile and 2. are these hypothetical people okay?

Imo you have to really lack imagination to not see use cases. I can tell you one use case from the top of my head that you may like. AI tutoring, especially for kids. I'd built a voice based language teaching bot for a young cousin of mine and giving the bot the face of a character she likes (e.g. Rey from Star Wars) would make her so much more engaged and excited to learn.

You'd be surprised by the number of people just talking to chat/voice bots. We make voice calling bots, and some people love endlessly talking to our bots for hours. Most of these people are perfectly normal. Check out: https://x.com/deedydas/status/1806352948328583221 .

Technology like this may one day eclipse OnlyFans as the go-to for lonely men. While OnlyFans women are real, AI women can be hyper-real. The conversations will be of better quality too [0].

[0] https://www.vice.com/en/article/onlyfans-management-agency-c...

Invoking Cunningham's Law. I just don't see the usecase for this. Like what are humans fundamentally missing that this is solving?

Pumping out more content faster? I can't even fathom a world where even more content is pumped out.

Avatar video calls? Just turn your video off if you are going to do that I don't need to talk to a puppet I want to talk to you.

Scams, misinformation campaigns, sucking lonely people down an even more lonely hole?

This feels past making a better mousetrap and into making a better window breaker.

Is your felt problem really about lack of positive, worthwhile use cases, or is this thing just so creepy?

Because I'm sure there are less useful yet less hated researches, like giant spherical purified water tanks in China lined with TV-sized vacuum tubes(they're coool).

I think AI outputs just don't meet expectations for human consumption. It's just too crude like how the molten giant from Nausicaa wasn't ready. Their uttarances cause people to flinch, or if people accumulated exposure despite red flags, slowly drive them crazy. And I think that's the real dealbreaker of the big AI launch - beyond generating boilerplaty code.

(I've seen people hypothesizing coming of a mass off-Internet movement that result from developing trust issues between the meat world and mega platformers, that I think we'll believe when that actually happen)

> Their uttarances cause people to flinch, or if people accumulated exposure despite red flags, slowly drive them crazy. And I think that's the real dealbreaker of the big AI launch - beyond generating boilerplaty code.

Following along with that, don't you think it's odd the authors didn't include a single Chinese figure in their examples? Maybe they get weirded out looking at faces they are used to seeing get generated and it's less weird using other races.

It's very obvious that this makes any interaction with an agent more engaging and potentially more effective.

But why do I want my interactions with an agent more engaging?

Going by the content of the conversation demos, I suspect this may one day eclipse OnlyFans and/or telehealth therapy sessions.

Targeted, hyper focused, “relatable” marketing is big business. It’s not a problem worth solving, but it’s worth a lot of money.

If only there was another term for "dyadic conversations", like "dialogues".

Unfortunate choice of name. If you search for you'll get a crap ton of Myers-Briggs pseudoscience (https://en.wikipedia.org/wiki/Myers%E2%80%93Briggs_Type_Indi...).

As I listened to the examples it seemed like the INFP personality type was chosen for much of the conversations.

I thought this was about the INFP personality type.

> INFP denotes our method is Interactive, Natural, Flash and Person-generic.

The shittiest acronym since "GPT".

What's wrong with "GPT" as an acronym?

Nothing, unless you're french-speaking (basically reads as "I just farted")

I thought so too, but I also wonder if the name here is more than just a nod to Meyers-Briggs. The personality types in the conversation demos are plausibly INFP.

Me too!

Reply on news.ycombinator.com