4 ms·

There is a new trend emerging in computing with the advent of Machine learning and that is, it is no longer sufficient to have talent/knowledge/passion/amazing

There is a new trend emerging in computing with the advent of Machine learning and that is, it is no longer sufficient to have talent/knowledge/passion/amazing team to create something. You need very deep pockets too. I've been waiting for a WebspeechAPI implementation in Firefox for many years now but the only reason why I don't think it's happening is because maybe Mozilla lacks the funds to train the models for speech-to-text/text-to-speech, etc. This gap will grow more and more as new machine learning features are introduced (e.g. natural lang page searching, better translation, etc) which would be very easy for Google to add to it's browser but very expensive for open source projects.

gggtt 6y ago
I am not convinced this is true. The reason Mozilla has not text-to-speech is because the guy behind RNNoise and LPCSpeech (JM Valin) has been bought by Amazon so that they can build even more privacy invasive Alexa & friends device. So yeah, in a way this an issue of money, but not because you can't afford GPUs & Data, but because you can't buy the right people
- eternalny1 6y ago
  > The reason Mozilla has not text-to-speech is because the guy behind RNNoise and LPCSpeech (JM Valin) has been bought by Amazon That is part of the problem. With smaller projects you are at risk of the "hit by a bus" problem, where major/complex technology is dependent on a single individual or a very small team.
fctorial 6y ago
They could ask their users to donate cpu/gpu if this is an issue.
- preommr 6y ago
  I am pretty sure that Google provides a lot of those services at a loss, or at the very least, they can synergize with other projects they have in way that justifies the cost. Either way, people will choose the free option rather than the option that requires them to pay even a nominal fee, particularly for software that people have gotten used to getting for free.
shakna 6y ago
> I've been waiting for a WebspeechAPI implementation in Firefox for many years now but the only reason why I don't think it's happening is because maybe Mozilla lacks the funds to train the models for speech-to-text/text-to-speech, etc. The Web Speech API is still in "Draft" spec, but Mozilla have been working steadily on it and finished the work for SGML parsing back in January. There are actually flags you can change to enable it on most platforms, its just they aren't considered "stable" yet. Mozilla's DeepSpeech engine is also not that far off from reaching a level of stability. You can help them with it here [0] if you're interested. [0] https://commonvoice.mozilla.org/en https://commonvoice.mozilla.org/en

gggtt 6y ago

I am not convinced this is true. The reason Mozilla has not text-to-speech is because the guy behind RNNoise and LPCSpeech (JM Valin) has been bought by Amazon so that they can build even more privacy invasive Alexa & friends device. So yeah, in a way this an issue of money, but not because you can't afford GPUs & Data, but because you can't buy the right people

eternalny1 6y ago
> The reason Mozilla has not text-to-speech is because the guy behind RNNoise and LPCSpeech (JM Valin) has been bought by Amazon That is part of the problem. With smaller projects you are at risk of the "hit by a bus" problem, where major/complex technology is dependent on a single individual or a very small team.

fctorial 6y ago

They could ask their users to donate cpu/gpu if this is an issue.

preommr 6y ago
I am pretty sure that Google provides a lot of those services at a loss, or at the very least, they can synergize with other projects they have in way that justifies the cost. Either way, people will choose the free option rather than the option that requires them to pay even a nominal fee, particularly for software that people have gotten used to getting for free.

shakna 6y ago

> I've been waiting for a WebspeechAPI implementation in Firefox for many years now but the only reason why I don't think it's happening is because maybe Mozilla lacks the funds to train the models for speech-to-text/text-to-speech, etc. The Web Speech API is still in "Draft" spec, but Mozilla have been working steadily on it and finished the work for SGML parsing back in January. There are actually flags you can change to enable it on most platforms, its just they aren't considered "stable" yet. Mozilla's DeepSpeech engine is also not that far off from reaching a level of stability. You can help them with it here [0] if you're interested. [0] https://commonvoice.mozilla.org/en https://commonvoice.mozilla.org/en