Live data from Hacker News

Meta AI announces Massive Multilingual Speech code, models for 1000+ languages

github.com

51–60 of 232 posts

Re: Meta AI announces Massive Multilingual Speech code, models for 1000+ languages

#51

Earlier quoted context omitted.

Based on the availability of STT, TTS, and translation models available to download in that github repo, a real-life babelfish is 'only' some glue code away. Wild times we're living in.

AFAIK STT is still very bad without speaker-specific fine-tuning, so it's not going to be a literal babelfish (translating in the ear of the receiver), but it could make you speak many languages.

Your interlocutor's earpiece could beam yours a delta for a finetuned model of their voice before they open their mouth. Except not compatible across iMessage users and whatsapp users or some other predictable silicon valley negative sum power play like that.

Re: Meta AI announces Massive Multilingual Speech code, models for 1000+ languages

#52
post #15

Earlier quoted context omitted.

It's not VR, it's Metaverse; VR might involve experiences that are not bullshit.

The metaverse, if we define it as an information layer that exists in parallel with the physical 'stuff' universe and is a seamless, effortless, and essential part of what we experience as reality will be an enormous part of our future. Meta might just be a few centuries ahead of the curve, which is just as bad as being a few centuries late.

Centuries really? I think Meta's going to be just on time.

Re: Meta AI announces Massive Multilingual Speech code, models for 1000+ languages

#53

Earlier quoted context omitted.

according to their blog post[1], MMS achieves ~half the error rate on words, while supporting 11x more languages. pretty impressive. [1] https://ai.facebook.com/blog/multilingual-model-speech-recog...

I wonder what the performance is on English specifically. Edit: Just checked the paper, it seems to be worse[1][2] but feel free to correct me. I feel like they should've just taken the Whipser architecture, scaled it, and scaled the dataset as they did. [1] Page: https://i.imgur.com/bq15Tno.png [2] Paper: https://scontent.fcai19-5.fna.fbcdn.net/v/t39.8562-6/3488279...

My guess is wav2vec performs better on low resource than whisper.

Re: Meta AI announces Massive Multilingual Speech code, models for 1000+ languages

#54
Imagine if we used these types of models for like 500 years and it locked their vocabulary in time, disallowing any further language blending; then somehow the servers turned off and nobody could communicate across language barriers anymore.

Someone should write that down in some sort of short-story involving a really tall structure.

Re: Meta AI announces Massive Multilingual Speech code, models for 1000+ languages

#55

meta is doing more for open ai than openai

I don't think anyone is too surprised by the fact that OpenAI went closed. As ironic as it may be, it was pretty obviously the only way they were going to continue to exist.

On the other hand, it definitely underscores:

- How blatantly companies exploit the "open source" concept with no remorse or even tacit acknowledgement; for many startups, open source is a great way to buy some goodwill and get some customers, then once they have what they want, close the doors and start asking for rent money. Nothing wrong with doing closed source or commercial software, but would OpenAI still have gained relevance if they hadn't started the way they did with the name they did?

- How little anyone gives a shit; we all watched it happen, but apparently nobody really cares enough. Investors, customers, the general public, apparently all is fair in making money.

I'm not suggesting OpenAI is especially evil, definitely not. In fact, the most depressing thing is that doing this sort of bait and switch mechanic is so commonplace and accepted now that it wasn't news or even interesting. It's just what we expect. Anything for a dollar.

But maybe I still seem like I'm just being whiny. Okay, fair enough. But look at what's happening now; OpenAI wants heavy regulation on AI, particularly they want to curtail and probably just ban open source models, through whatever proxy necessary, using whatever tactics are needed to scare people into it. They may or may not get what they want, but I'm going to guess that if they do get it, ~nobody will care, and OpenAI will be raking in record profits while open source AI technology gets pushed underground.

Oh I'd love to be wrong, but then again, it's not like there's anything particularly novel about this strategy. It's basically textbook at this point.

Re: Meta AI announces Massive Multilingual Speech code, models for 1000+ languages

#56
post #36

The problem with all these model releases is they have no demos or even video of it working. It’s all just download it and run it, like it’s an app.

They're intended for researchers/professionals not consumers, and I'm not sure how a video is going to be helpful?

And the issue with a live demo is that these are resource-intensive, they're not just webpages. It's an entire project to figure out how to host them, scale them to handle peaks, pay for them, implement rate-limiting, and so forth.

For the intended audience, download-and-run-it doesn't seem like an issue at all. I don't see how any questions are going to be answered by a video.

Re: Meta AI announces Massive Multilingual Speech code, models for 1000+ languages

#58
post #36

The problem with all these model releases is they have no demos or even video of it working. It’s all just download it and run it, like it’s an app.

Here's the blog post with video of it working: https://ai.facebook.com/blog/multilingual-model-speech-recog...

Re: Meta AI announces Massive Multilingual Speech code, models for 1000+ languages

#59
post #39
post #36

The problem with all these model releases is they have no demos or even video of it working. It’s all just download it and run it, like it’s an app.

I'd argue that having a "download and run" approach is so much better than videos or demos. Why do you think this is a problem?

Why should we have any marketing materials, ever? Why do we show pictures of products? Sometimes people want to see the capabilities before they completely dive in and spend their time working on something.

Re: Meta AI announces Massive Multilingual Speech code, models for 1000+ languages

#60
post #30

meta is doing more for open ai than openai

this is absolutely true, with pytorch, llama, ton of models in vision and now with this !

We're in upside-down world, Meta is now cool; actually I always admired their open source projects, starting with React and PyTorch.
Post reply on HN