AI Avatars Escape the Uncanny Valley

Setting aside the uncanny valley, my understanding is that people like seeing people not because they look like people but because it conveys effort and investment on the part of another human being. We want to feel like we're worth someone else's time, in every context.

We don't like to interact with people just because that's the only way we can interact but because humans are human. AI avatars are the exact opposite: a statement of disinterest. If you don't care enough about my business to be on a sales call with me, why would I bother speaking to an AI avatar you send in your place? What's a thank you message without a human being actually taking the time to record it?

AI avatars seem a lot like crypto: they're a neat technology solving the wrong problem. The "inefficiency" of humans interacting with humans is the fundamental component of communication. I guess it's a lot like LLMs: instead of producing less content that is more valuable / thoughtful per unit, we're producing a lot more content that is much less valuable / thoughtful per unit. AI avatars will create more vacuous communication, not enable our communication to be more thoughtful.

Maybe human behavior will change because of this, maybe the next generation that grows up interacting with AI avatars won't have this same feeling that speaking to an actual human means something.

So far the largest applications I've seen of AI avatars are deepfake-ish Musk videos on YouTube trying to scam people with some shitcoin, and pornography (as always).

> We expect this space will give rise to multiple billion-dollar companies,

The most miraculous thing about this wave of AI is how democratized it is and how resistant it has been to anyone making money of it.

So it's another case of "let's make a solution for nonexisting problem" in the world of AI?

We all hear promises of how AI will save the world and bring about technological innovation and solutions for hard problems, and what we get is replacement for a job that can be done much cheaper by a human being.

I am getting major crypto vibes here, promises of greatness and bringing nothing but unnecessary fluff.

The more AI stuff I see makes me want to purchase a cabin somewhere in remote Montana and start writing a journal.

Doesn't seem to be any discernable improvement on the Zoom Cat Lawyer.

From TFA:

> Can’t you just generate an image of a face, animate it, and add a voiceover? Not quite. The challenge isn’t just nailing the lip sync — it’s making facial expressions and body language move in tandem. It would be weird if your mouth opened in surprise, but your cheeks and chin didn’t budge!

Starting here:

- "generate an image of a face, animate it, and add a voiceover?"

Tried this at [0]. Here's an example visual output:

- https://visualmic.com/example-animation-4.gif

Judge for yourself. You can see the mouth and eyebrows move in response to voice volume, and the eyes shift and blink according to settings. But no cheek movement, no head tilt, and no face shape change.

I think TFA is sort of right.

I'm not sure that face cap and AI are 100% needed, and most of the tools for making great VR models seem either pretty complicated or sort of privacy invading. But, better translating voice input into face changes does seem sort of needed.

There's "virtual Youtuber" (vtuber) software too, but that too seems some combination of complicated/clunky, resource intensive, and/or in signup-required land. [EDIT: Surely, there is a good front end at OpenLive3D [1], but making the .VRM model for it, e.g., with VRoid Studio [2] is where things seem start to get a little more time/energy-intensive.]

I'm not against pseudonymous avatars, but is there a third path? It should be easy and open, no? Gonna have to trawl through the suggestions of a16z on this one.

  [0] https://visualmic.com
  [1] https://openlive3d.com
  [2] https://vroid.com/en/studio

When can we escape a16z ramming worse than useless concepts down our throats and telling us how great they are in a sanctimonious fashion?

That will be the greatest advancement of the millennium for humankind.

This is besides the point, but the linked Monoverse / Unanswered Oddities videos are so good. Would definitely recommend, besides being hilariously funny it's a testament to what the average person can accomplish now if they actually want to put time into making good AI-based content.

This is just a personal preference thing - but I dislike any attempts at injecting artificial personhood - if they make an agentic personal assistant or whatever, I don't want to think of it as a person but a rather sophisticated tool. My phone uses AI to retouch the pictures it takes with its camera, but I don't have to exchange pleasantries with iPhone Clippy for it to do its thing - it just pretty much does.

I don't want artificial buddies, or servants or whatever except maybe in video games.

« This has unlocked advertising for businesses that could never afford traditional ad production »

So glad to see SV focusing on first-order problems after all

I as an artist will be happy to see all these useless talking head "content creators" get replaced by AI Avatars-

So then only real artists and ai will remain- these "content creator" human avatar golems were always transitory anyway- software can chase algo's/trends better than any human so good riddance-

A16Z should rebrand as a ballistics company with the rate they've been shooting themselves in the foot recently.