Comments URL: https://news.ycombinator.com/item?id=48699803

Points: 4

# Comments: 0

is this a good idea: having an online directory of our questions with your answers? An ISS interview directory was posted in Show HN recently, however it only includes established astronauts…

Comments URL: https://news.ycombinator.com/item?id=48224401

Points: 2

# Comments: 0

Visitors can answer any question they like, and if they want to be attributed, they can add their name.

tell us what you think!

Comments URL: https://news.ycombinator.com/item?id=48217882

Points: 2

# Comments: 0

Comments URL: https://news.ycombinator.com/item?id=48181240

Points: 2

# Comments: 3

A Golden Speaker= corrected accent, similar voice identity= easier to mimic

Step 1. Record themselves reading a given sentence (phonetic pangram) aloud. Step 2. Have the Golden Speaker system generate a Golden Speaker for them Step 3. Use their Golden Speaker to read aloud any text or convert their speech in real time

"Baby Frankie" has...terminated as of yesterday. The new iteration is called "Frank". (Frank/ie is growing up, you see)

@ https://loxai.tech!!

We generate a "Golden Speaker" for you -> You record yourself saying a sentence (a phonetic pangram), we keep your accent which represents your voice identity but we correct your pronunciation and then you have a Golden Speaker system. This may be in the form of a TTS or a real-time accent conversion technology.

the above has a provisional patent attached to it, but obviously that doesn't mean anything. & because I'm 16 and know very little so far, I'd very much appreciate any input on the above idea (it's currently a waitlist @ https://loxai.tech). esp if anyone's in linguistics or AI, lmk your opinion!

In 1990, the first research paper regarding Golden Speakers came out proving their efficacy in helping individuals build native-like accent in a target foreign language. A "Golden Speaker" refers to another person with proper pronunciation but a similar voice identity.

With today's technology, a user's "Golden Speaker" can be generated-> You record yourself saying a sentence (which would be a phonetic pangram), we keep your accent which represents your voice identity but we correct your pronunciation and then you have a Golden Speaker system. This may be in the form of a TTS or a real-time accent conversion technology.

furthermore, if you would like to get early access to a real-time accent conversion technology to be able to go from Point A (foreign accent) to B (native-like accent) faster, you can sign up at loxai.tech.

would very much appreciate input on this idea especially from those in AI and linguistics

Comments URL: https://news.ycombinator.com/item?id=48114622

Points: 1

# Comments: 1

- Individuality and ego are human needs. In the past year, there's been a boom in using AI to personalise products to consumers. A take on this: are companies really doing "individuality as horizontal growth" as consumers would like them to? Do consumers really need businesses to tell them that they're individuals? Are personalised feeds actually honouring their individuality, or are they just grouping consumers into demographics that homogenise them and strip away their individuality? And do businesses realise that consumers aren't blind to this? (reference at the rise & mission of Bluesky).

- Above connects to backlash from consumers in response to AI slop. AI slop is a feature of this era of conformity-> so look ahead to the era of individuality (reference historical back-and-forths between these two concepts- ie 1910s were the tail of social conformity and featured great social unrest until it was followed by the 1920s)

Also:

- I am assuming Hacker News is visited by a lot of software devs: one-shot learning and zero-shot learning could be something to check out! (reference Meta currently building Large Concept Models)

That's all. I hope you got value out of this post and have a good day!

(long note if you appreciated this and want to see an example of the above concepts and judge some ideas for yourself:

The above was written by me, Audrey Fei, I'm 16:

- I've been looking at the speech area of AI (text and image has already had its moment). currently have a patent pending for a real-time accent conversion technology (rooted in the idea of a “Golden Speaker” which dates back to 1990; also the patent is for the specific architecture of the System, which is different from standard architectures expected in the industry).

it's at loxai.tech (I'm trying to raise awareness to this tech so there's a waitlist page). the central idea behind it is that: at least in the subfield of speech processing, Golden Speakers are contrarian to the norm of speech synthesis (where speaker identities are lost, instead of preserved). also it uses one-shot learning (a MAML) in its structure

- there's a MVP that led to the above (neutboom.com). It centers around natural language acquisition (differentiates itself from the rest of the language learning market). It's might be a bit clunky in places (you can just tell me about anything that's ugly) but hopefully it's interesting or adds some perspective (at least personally I feel that translation is a middleman.....and why would you need a middleman especially when it overcomplicates the process of language learning? input on this would be appreciated).

)

Anyway I hope you have a creative rest of your day!

Comments URL: https://news.ycombinator.com/item?id=48057908

Points: 2

# Comments: 2