3 ms·

Roast My Idea

1 points by thothofc 2y ago

Hello everyone, please be honest about the this idea. Im building a multilanguage social platform with a Speech to Speech model. Why? There are over 7000 languages, 40% of the world population only speaks their native langugage and 75% don't speak english. This will solve the communication barrier and expand the reach of content. Looking for angels btw.

MilnerRoute 2y ago
Sooner or later, someone is going to ask you: how are you going to moderate it? But your angel investors will want to know: how are you going to monetize it? (Advertisers will probably want their ads targeted by region -- but also by keywords...) But you could really get people excited if you had some novel approach to all the standard issues people bring up around social media. (Foreign influence campaigns/disinformation, CSAM, etc.)
- thothofc 2y ago
  Moderaration is made via machine learning models spam comments, nsfw and deepfakes can be easly solved. I am really into this project and exacly thinking abouth these things ahead on. There is something I found contradictory about angel investors and vc space, if you pay attention to any of these podcasts they are always saying to just ship things at the basis and then add the complex stuff later bc what you must worry about is getting the users and promoting the product. But at the same time they come up with all those questions about it.
numpad0 2y ago
Couple reasons I can think of: - GPT translations aren't that better than regular old MT, it still "sound robotic" and starts talking unrelated things that share literal expressions just like anything before it. - Not all languages has same word orders, and some people think as they speak; this means babelfish is plain not possible without a time machine. - Actually, try running some French or German TV or YouTube or TikTok content through that pipeline and see if you like it. You can automate the process later, ask ChatGPT how or outsource to penguins in Antarctica for a bucket of fish, just see if the concept of "funny videos from somewhere else with AI voiceover" works at all.
- thothofc 2y ago
  [dead]

Roast My Idea

1 points by thothofc 2y ago

MilnerRoute 2y ago

Sooner or later, someone is going to ask you: how are you going to moderate it? But your angel investors will want to know: how are you going to monetize it? (Advertisers will probably want their ads targeted by region -- but also by keywords...) But you could really get people excited if you had some novel approach to all the standard issues people bring up around social media. (Foreign influence campaigns/disinformation, CSAM, etc.)

thothofc 2y ago
Moderaration is made via machine learning models spam comments, nsfw and deepfakes can be easly solved. I am really into this project and exacly thinking abouth these things ahead on. There is something I found contradictory about angel investors and vc space, if you pay attention to any of these podcasts they are always saying to just ship things at the basis and then add the complex stuff later bc what you must worry about is getting the users and promoting the product. But at the same time they come up with all those questions about it.

numpad0 2y ago

Couple reasons I can think of: - GPT translations aren't that better than regular old MT, it still "sound robotic" and starts talking unrelated things that share literal expressions just like anything before it. - Not all languages has same word orders, and some people think as they speak; this means babelfish is plain not possible without a time machine. - Actually, try running some French or German TV or YouTube or TikTok content through that pipeline and see if you like it. You can automate the process later, ask ChatGPT how or outsource to penguins in Antarctica for a bucket of fish, just see if the concept of "funny videos from somewhere else with AI voiceover" works at all.

thothofc 2y ago
[dead]